Discussions of unified science frequently suppose that the various scientific theories should be combined into one unified theory, and it is usually supposed that this should be done by successive reductions of the various theories to some fundamental theory. Yet, there has been little systematic study of the characteristics of unified theories, and little foundational support for the use of reductions as a unifying procedure.

In this paper I: (a) briefly review some of my previous work on microreductions, (b) state some conditions which are necessary in order for a theory to be unified, (c) argue that when certain identities exist between the elements in the domains of two theories, then the only satisfactory way to combine these two theories into one unified theory is by a microreduction, and (d) indicate briefly some further applications and consequences of this work.

That reductions of theories and unifications of science frequently occur by means of identifications is a widely accepted hypothesis of methodology. Usually we are concerned with the micro-reductions of things, in which an entity is identified with a structured aggregate of smaller constituents. And, it is frequently alleged, we must also take into account further identifications in which the attributes of the macro-object, expressed by predicates in the macro-theory, are identified with attributes of the aggregate of micro-entities differently expressed by predicates of the micro- or reducing theory.

I am fundamentally in strong agreement with this view, but intend here to explore one case of inter-theoretic reduction where some important qualifications of this account of reduction are in order. Philosophical scepticism is frequently encountered as to the possibility of ever making coherent sense of the notion of attribute identification. This scepticism often shows itself in the opposition to the identificatory account of the reduction of the theory of mental states to that of neurological states.

Although doubtlessly aimed at later developments in physics, Einstein's famous remark, if interpreted so as to include classical statistical mechanics, nicely captures the spirit of work on the early quantum theory by men like Bohr and Ehrenfest, who, despite their conviction that the classical theories failed, nevertheless mined their riches to the fullest in the development of the new theory. In this paper I try to exhibit and characterize the patterns of reasoning involved in Ehrenfest's attempts to generalize Planck's early theory of the linear harmonic oscillator, and I then employ the same historical case as a basis for arguing that the reduction of problems to problems is an important phenomenon which cannot be fully understood in terms of the reduction of theories.

Galileo has been seen, from the philosophical point of view, alternately as a Platonist whose rationalist insights enabled him to read the book of nature because it was written in ‘the language of mathematics,’ and as an experimentalist who used the hypothetico-deductive methods of modern science to establish his new results empirically (McTighe, 1967; Settle, 1967; Drake, 1970; Shapere, 1974). Both of these views present difficulties. In this paper I shall make use of recent historical research to argue that neither is correct, that the method utilized by Galileo was neither Platonist nor hypothetico-deductivist, but was basically Aristotelian and Archimedean in character. This method, moreover, was not merely that of classical antiquity, but it had been emended and rejuvenated in the sixteenth century, and then not by Greek humanist Aristotelians or by Latin Averroists but rather by scholastic authors of the Collegio Romano whose own inspiration derived mainly from Thomas Aquinas.

Notable among the by-products of the late nineteenth-century controversies over the atomic theory of matter was the first full-scale contribution to the philosophy of science written in America, John Bernard Stallo's The Concepts and Theories of Modem Physics. The work of a lawyer and jurist largely self-trained in philosophy, its publication in 1881 provoked considerable (and mostly hostile) reaction, for Stallo attacked atomism and mechanism at a time when scientists were still divided into vociferously opposing camps over the question of whether or not the ultimate structure of matter was atomic - indeed, were in sharp disagreement over even the utility of the hypothetical concept ‘atom’ in physics and chemistry.

Evér since the early 1800s, when Dalton had transformed the corpuscular hypothesis from a commonplace of the prevailing natural philosophy into a working part of chemistry, the fortunes of atomism had risen and fallen as competing atomic and non-atomic theories were matched against the accumulating experimental evidence.

Two doubtful philosophical doctrines have contributed strongly to the development of IQ tests, and now play a major role in defending IQ tests as measures of intelligence: they are operationism and fictionalism.

The application of operationism in psychometrics is in such doctrines as the following: one can avoid answering the question whether IQ tests measure intelligence simply by defining the word ‘intelligence’ as what IQ tests measure. Obviously, such a definition does not avoid the issue. For we can now reasonably ask whether people who accept the operational definition use the word ‘intelligence’ to refer to the same quantity as that referred to by people who reject the operational definition or have never considered it. Many operational definitions of ‘intelligence’ would be clearly unsatisfactory. Consider, for example, an operational definition which stipulates that a person's intelligence is the number of pounds indicated when the person is placed on a scale.

Scientific revolutions are no everyday affair. So it is of some importance that Fortune (Alexander, 1972) and several other major magazines have recently proclaimed that we are in the midst of a major Kuhnian revolution in the social sciences, one that has significant implications for social policy. According to these magazines, ‘environmentalist’ theories, which assume that equalization of human environments and opportunities will increase equality of achievement between individuals, groups, and races, are in ‘crisis'. The crisis exists because the egalitarian reform programs of the 1960's, which relied on such theories and thus constituted tests of them, failed to equalize achievement. In the face of this crisis, Fortune suggests, scientists are welcoming evidence from the study of ducks, baboons, and humans which points to “a basic intractability in human nature, a resistance to being guided and molded for improving society” (Alexander, 1972, p. 132).

Although my two colleagues on this panel are bona fide philosophers who clearly have a bona fide interest in the ongoing controversy over IQ and the educability of children (which should mean that the topic requires no further defense), I nonetheless feel constrained to say a word in defense of our topic as a viable subject for discussion in a forum devoted to the Philosophy of Science. Perhaps this is because, as the representative of the discipline which created this mess, I fear you might associate me with it. You might feel that bad science is a bad subject for study by philosophers of science, and furthermore that anyone associated with such a discipline would be the last person to have anything intelligent to say about philosophy.

While it is true that we do not need a philosophy of bad science, a kind of Gresham's Law seems sometimes to operate, in which bad science drives out the good.

It may sound strange, if not heretical to suggest, as I wish to do, that there is no such thing as the rationality of science. At best we can talk about rationalities of science. Both historically and contemporarily we have used different criteria for explaining and justifying the alleged rationality of science. By giving different criteria for the justification of rationality, we ipso facto constitute different scopes for rationality.

This immediately poses the question. If so, which of these criteria are more “justifiable” than others ? Which rationality is preferable to other rationalities ? Now the point is that in justifying a set of criteria which would elucidate for us a given concept of rationality, in other words, in outlining the scope of a given rationality, we do so by resorting to some kind of rationality. And it would be strange if the hidden rationality we resort to, in order to justify the explicit one, would be much different from the explicit one.

In several recent papers ([7], [8]) Bas van Fraassen has suggested that the structure of a scientific theory might be more appropriately represented by his “semantic view of theories” than by the traditional “syntactic view.” Under the syntactic view, to characterize a theory one provides “a finite list of sentences given to count as axioms, plus a finite set of syntactic transformations, of an effective character, given to generate the set of all theorems from these axioms.” ([8], p. 305) The set of all such theorems is identified with the theory, and contains the set of all claims made by the theory about physical systems. Under the semantic approach, however, one “does not view a physical theory as… a kind of Principia Mathematica cum nonlogical postulates.” ([7], p. 337) Rather, “all the resources of mathematical English” are used to construct a theoretical framework, and “the theoretical reasoning of the physicist is viewed as ordinary mathematical reasoning concerning this framework.”

Philosophers working in the logical empiricist tradition, such as Rudolf Carnap and Carl Hempel, have held that theoretical terms get their meanings from their connections with observational terms in the postulates of scientific theories. They have tried to explain exactly how this happens, in the sense of giving a logical reconstruction of it. For various reasons, to be discussed below, the belief arose that it is impossible in general to give explicit definitions of theoretical terms. Attention thus shifted to showing how correspondence rules relating theoretical and observational terms give ‘partial interpretations’ to the theoretical terms.

Three objections can be raised against the way in which this enterprise was conceived and carried out. These objections involve what will here be called (1) ‘The Problem of the Theoretical-Observational Dichotomy', (2) ‘The Meaning-Magic Problem', and (3) ‘The Problem of Unwanted Interpretations'. The purpose of this article is to set forth the second and third of these objections and develop a solution to them.

Feyerabend and others have been defending for some time the thesis that even observation sentences depend for their meanings on the thegries in which they play a role. According to the most radical version of this thesis, if two theories contain incompatible theoretical statements, then even if the observation terms and sentences associated with the two theories are the same, these terms and sentences mean something different, according as they are associated with one theory or the other. This claim has made it hard to see how, on this view, any two theories could be said to compete, or be incompatible, or be comparable on the basis of observation.

The whole dispute over this claim seems highly suspect to begin with. For look at the philosophical terms in which it is couched. On the one hand there is the notion of the “meaning” of a term or sentence, whose status as a useful tool of philosophical analysis has been in serious question for the past two or three decades.

What relevance - if any - does philosophy of science have for science education? Unfortunately, this question has been largely unexplored. To be sure a great deal has been written on the philosophy of science; perhaps even more has been written in science education. However, surprisingly little has been written on the relation between the two areas. In this paper I will suggest some ways in which philosophy of science can have relevance for science education.

Before one considers how philosophy of science can have relevance to science education it is important to consider what is meant by ‘philosophy of science’. Some contemporary philosophers of science have distinguished two different approaches to philosophy of science: analytic and speculative. However in recent years a view of the philosophy of science has become popular that does not seem easily to fit into either the analytic or speculative categories.

Without question, one of the most important cluster of issues in recent philosophy of science has centered around the attack on the rigid positivist distinction between theory and observation or between a theoretical language and an observational language. Kuhn, Feyerabend, Hanson, Toulmin, and Polanyi are all names closely associated with one version or another of this attack. All have argued that observational categories are essentially theory-determined and there is no determinate observational base, or neutral observational language. Thus at least the positivist account of the objectivity of scientific knowledge would seem to be seriously threatened by the thesis of the theory-ladenness of observation. For without an independently accessible observational base against which to test scientific theories, wherein would objectivity consist ?

A number of philosophers of science have rallied to the defense of objectivity against the threat posed by the thesis of the theory-ladenness of observation. One of the earliest defenses and still one of the most reasonable and persuasive was offered by Israel Schemer in his book, Science and Subjectivity.

Since I have had the privilege of reading the remarks of my fellow-symposiasts before writing out my own, I should like to begin by making the briefest of comments on each of their contributions. Professor Petrie raises the Kantian-like question of how science education is possible – or, more precisely, how, if all observation is theory-laden, science education is possible. His intriguing answer is that “Science education is possible with the aid of the humanistic tool of metaphor.” I find it difficult to asses this formulation without further examples of its concrete application to actual pedagogical problems in the natural sciences.

According to the backing law account of causation a singular causal claim is to be analyzed (or “justified”) by reference to a suitable nomic theory which, together with the given singular statement describing a cause, deductively supports or explains the statement describing the effect. This backing law (or deductive-nomological) account of singular causation has recently become the target of several kinds of criticism. First, the possibility of giving a detailed and elaborate account of the required nomic or explanatory backing has been doubted. Secondly, it has been claimed that the deductive-nomological account is bound to lead to unacceptable ontological difficulties (see Kim [1969] and [1973a]). Thirdly, it has been argued that the backing law account fails to give a satisfactory analysis of certain conditional or counterfactual aspects of causation (see Lewis [1973b]). Fourthly, it has been claimed that this approach does not work at least within the social and historical sciences as there are no laws and theories of the required sort to be found there.

Hempel in Aspects of Scientific Explanation and Other Essays tells us that “all scientific explanation involves, explicitly or by implication, a subsumption of its subject matter under general regularities; that it seeks to provide a systematic understanding of empirical phenomena by showing that they fit into a nomic nexus.” This I take to be an informal statement of his covering-law, “model” of scientific explanation. In defense of his “model” against apparent examples of scientific explanations which do not fit it, Hempel's almost instinctive reaction has been to patch them up in some way so that they do conform. This, I believe, is the wrong way to defend the covering-law “model”. In the three parts of this brief essay I shall do the following: (1) show that Hempel's attempt to patch up a certain version of so-called “rational” explanation does not succeed of its purpose, (2) generalize that result with respect to all dispositional explanations, and (3) reflect momentarily on philosophic method and another way to defend the covering-law “model”.

The following considerations are offered in response to the critical observations and constructive proposals set forth by Laird Addis, in his paper ‘On Defending the Covering-Law “Model”’, concerning my explications of “rational” and dispositional explanation and concerning the claims associated with the covering-law model of explanation.

1. My main objection to Dray's construal of rational explanation, as characterized by the first schema in Addis's paper, was not that it is at odds with the covering-law model, but that the second explanans sentence, ‘In a situation of type C, the appropriate thing to do is x’, expresses a norm and therefore cannot possibly explain why A did in fact do x: to do that, we need, not a normative sentence, but a descriptive one, roughly to the effect that A was disposed to act in accordance with the normative principle.