Both the most controversial and the most widely publicized aspect of the “new empiricism” of Feyerabend Hanson Kuhn and Toulmin is the thesis (rightly or wrongly) attributed to them that irrationality and subjectivity play a central role in science. The idea that science is pervaded by apt parallels to religious conversion and political propagadizing has achieved truly surprising currency. The epistemological root of this view is the belief that the theory-laden conception of observation which these philosophers espouse effectively prevents any rational comparison of the observational support commanded by competing theories. It has even been pointed out that in the absence of a neutral observational language in terms of which to identify and describe the objects or processes with which rival theories are (supposedly) concerned, there can be no real basis for calling them “rivals”.

It is the intent of this paper to set forth conditions which, if met, would provide a rational basis for believing, even in a context in which observation is theory-laden, that two theories are rivals and that one of them enjoys observational support that is superior to that enjoyed by the other.

For several years now Quine has been proclaiming a solution to problems concerning observation sentences and evidence: “The dislodging of epistemology from its old status of first philosophy has loosed a wave … of epistemological nihilism. This mood is reflected somewhat in the recent tendency of Polanyi, Kuhn, and the late Russell Hanson to belittle the role of evidence and to accentuate cultural relativism …. It is ironical that philosophers, finding “the old epistemology untenable as a whole, should react by repudiating a part which has only now [with Quine's account of observation sentences] moved into clear focus.” (pp. 87-88). Attacks on the notion of observationality by the critics of positivism have been : viewed by many (for example Scheffler, Chapter 1) as seriously undermining the objectivity of science. Quine's claim to have solved the problems, therefore, deserves careful consideration and evaluation.

It is without doubt that the earth sciences were rocked to the core by a revolution in the late sixties with the initial acceptance of Harry Hess’s hypothesis of seafloor spreading and subsequent development and acceptance of plate tectonics. The major aim of this essay is to show precisely why this revolution is not Kuhnian. However, my reasons for claiming that Kuhn’s model fails to apply are at variance with Ruse [27] but in much agreement with R. Laudan [18]. I shall highlight our differences as space permits.

There are two general accounts of the rise of continental drift and plate tectonics which satisfy Kuhn’s model of scientific growth and change. (1) The geosciences prior to the acceptance of seafloor spreading and plate tectonics were in a pre-paradigm state. Earth scientists solved no problems within geological domains, but only argued among themselves about which of the various schools of thought provided the best framework for attacking geological problems.

Our view of the first half of the 20th century has been influenced by what we suppose to have occurred in the middle of that century. It is by now part of the conventional wisdom of the geological community that during the 1950’s and 1960’s a revolution occurred. It is further supposed by many, that before the revolution there was among geologists an uneasiness resulting from the lack of an organizing principle in terms of which accumulating facts could be understood. It is difficult to see our own times in the kind of perspective that historians consider so important an ingredient of ‘good’ history. Geologists now only in their middle years began their careers before the ‘new tectonics’ came to pervade their discipline. They do not look back upon prerevolutionary days as a time waiting to be saved from crisis nor do they see themselves as having collected facts in anticipation of being able to weave them into some pattern on the basis of some as yet undiscovered truth.

The problem we are concerned with in this paper is that of separating out, among the consequences of an empirical theory, those that play a purely linguistic role (the “analytic” consequences) from the rest. This problem we shall refer to as the problem of theoretical analyticity. In a sense, the problem was solved by Carnap's proposal (p. 270) to consider the analytic consequences to be exactly the consequences of a special sentence associated with each theory, along with results of Winnie, which gave further justification to this proposal. Besides Carnap's proposal there have been other attempts to solve the problem of theoretical analyticity (e.g.,). However, all such attempts (including Carnap's) have relied upon a distinction between observation and theoretical terms in the language of empirical theories. Since there has been much debate concerning this distinction, and since there do not seem to be any good reasons why “analyticity” should be tied so closely to it, we shall attempt to attack the problem of theoretical analyticity without assuming an observation-theory distinction among the vocabulary of empirical theories.

The 1960s witnessed a striking change in geology. Since at least the seventeenth century, one of the central problems of the subject had been the origin of the major irregularities of the surface of the globe—continents and oceans, mountain chains and ocean islands—irregularities that were not anticipated by most physical theories. Traditionally these features had usually been explained either as residual traces of events occurring during the very early history of the globe, or as the result of vertical movements of the earth’s crust, caused, for example, by changes in the heat budget. The last two decades have seen an end to all this. The vast majority of geologists now believe that these irregularities largely result from the lateral movement of thin rigid plates covering the earth, a theory now known as “plate tectonics”, but a theory which also has obvious parallels with the hypothesis of continental drift, in which it was postulated that continents can move laterally.

It takes no more than a passing acquaintance with functional localization research in the neurosciences to notice that psychological notions often come into play in the characterization of human brain functions. In the past, such notions were taken primarily from one form or another of associationism; currently, they are adopted from the language of information processing. A rather typical example can be found in a recent article by A.R. Luria (see also). He describes the human brain as made up of three main blocks incorporating basic functions. These functions are described as follows:

What is the relation between a scientific problem and the constraints on its satisfactory solution? To what extent can problems be identified with (or how far are they determined by) such constraints? What types of constraints are there, and how do the various types function to determine problems and their solutions? I wish to consider these questions in the context of a single historical case, the blackbody radiation problem, 1859-1900. Since only the most condensed treatment of even one historical problem is possible, I select one with a familiar and accessible history (see for details). Despite the historical attention it has received, I know of no philosophical treatment of the blackbody case from a problem oriented viewpoint.

More attention to constraints on problem solutions can help philosophers of science in several ways.

The one thing upon which we can all agree is that just over ten years ago a major revolution occurred in the science of geology. Geologists switched from accepting a static earth-picture, to endorsing a vision of an earth with its surface constantly in motion. (Cox [4]; Hallam [12]; Marvin [28]; Wilson [56]). It is true that early in this century the German geologist Alfred Wegener argued that the continents as we today find them have “drifted” to their positions from other positions widely different. However, other than amongst a number of scientists drawn almost exclusively from the Southern Hemisphere, his ideas fell on deaf — or more precisely, contemptuous — ears. Then in the mid 1960’s, almost literally overnight, the geological community swung around and embraced the hypothesis of continental drift, or what we shall see is perhaps more accurately called “plate tectonics”.

Understanding the growth of scientific knowledge is a central problem in philosophy of science today, a part of which takes the form of understanding how new knowledge arises and how it becomes unified with existing knowledge. As a way of tackling this problem, one may ask: how do new fields arise in science? And how is the knowledge in a new field related to that in other fields?

In recent work I characterized a scientific field as an area of science consisting of the following elements: a central problem, a domain of items taken to be facts related to that problem, general explanatory factors and goals providing expectations as to how the problem is to be solved, techniques and methods, and concepts, laws and theories related to the problem which attempt to realize the explanatory goals (p. 44). Also in that work, my coauthor and I discussed interfield theories which serve to link two fields by postulating specific kinds of relations between them.

Quine once claimed that modal logic derived from a use-mention mistake (p. 196). In this paper, I will peddle an, equally outrageous claim: the thesis that subsumption is an essential goal of fundamental explanatory theorizing rests on a failure to distinguish explaining what made x happen from explaining why we believe what we do about what made x happen. I don't mean to suggest that the distinction wasn't noticed (except perhaps by Hume); only that it wasn't vigorously enforced. There have always been philosophers who assimilated a belief to its justification with malice aforethought.

Everyone knows that subsumption is not sufficient for explanation: the length of a pendulum is not explained by appeal to its period and the pendulum law, and emission spectra are not explained by Balmer's formula. But the spirit of the deductive-nomological account of explanation lives on in the belief that subsumption is an essential feature of an explanatory theory.

I became acquainted with Lakatos’s work in 1965 when I started studying at London School of Economics—where Lakatos taught. As his work was developed over the succeeding years until his death in 1974, one thing always puzzled me: his work seemed to contain such conflicting tendencies. He would continue developing his ideas along a progressive line, and suddenly would insert an element which appeared to me quite reactionary. By ‘reactionary’, I should hasten to add, I mean imbued with the spirit of Positivism—a person of different bias might reverse the labels!

When I was given this opportunity to reflect again on Lakatos’s work I did not try to resolve the puzzle by looking into Lakatos’s intellectual history1; rather I attempted to separate clearly those aspects of Lakatos’s work which seemed to me driving in the right and the wrong directions. Both aspects concern the research program which has dominated both philosophy of science and philosophy of mathematics in the twentieth century.

Mathematics offers us a puzzling contrast. On the one hand it is supposed to be the paradigm of certain and final knowledge: not fixed to be sure, but a steadily accumulating coherent body of truths obtained by successive deduction from the most evident truths. By the intricate combination and recombination of elementary steps one is led incontrovertibly from what is trivial and unremarkable to what can be non-trivial and surprising.

On the other hand, the actual development of mathematics reveals a history full of controversy, confusion and even error, marked by periodic reassessments and occasional upheavals. The mathematician at work relies on surprisingly vague intuitions and proceeds by fumbling fits and starts with all too frequent reversals. In this picture the actual historical and individual processes of mathematical discovery appear haphazard and illogical.

The first view is of course the currently conventional one which descends from the classic work of Euclid.

The dispute concerning whether reasons are causes and related controversies continue unabated. The conflict manifests itself in various ways. Reasons are not causes: because reasons are not logically independent of actions; because no inductive evidence is required for establishing a reason for why I do something; and because no generalizations are required to explain why someone does something on the basis of reasons. It is also maintained that historical explanation is not explanation by appeal to covering laws and that empathetic understanding is more than an heuristic device for generating hypotheses to explain why people do what they do. It is further maintained that the social sciences are on the wrong track if they see themselves as more akin to the natural sciences, as opposed to philosophy (conceptual analysis) and the humanities.

The amazing fact is that the contradictory of each of the above stated positions has been seriously maintained.

I once wrote to Imré Lakatos that if he had not already established himself as an accomplished philosopher and provocateur, he could have made a successful career as a Hollywood script writer. I had in mind, at that time, a beautifully written paper he had constructed by pasting together the cuttings from other papers. I had forgotten that his reputation as a dramatist was already established, with the production of his Proofs and Refutations [9]. For that work, it seems to me, is simply a play within a play. The main character, whose part is written in the footnotes, delivers a monologue which is a commentary on the drama, between teacher and students, that occupies center stage. This main character is not named in the play, so for ease of reference I shall call him “Imré”. What Imré points out is that the drama on center stage imitates life, in showing us the essential features of the methodology by virtue of which mathematical science grows.

Sometimes, when an explanandum E is a quantitative claim, we are unable to explain why the specific numerical value mentioned in E is correct. Sometimes an explanans X is available, however, which accounts for a statement E' asserting a value close to that mentioned in E. Although X strictly speaking provides evidence against E, it is in some circumstances seen as approximatively explaining E along with E'. Such an approximative explanation may be defined as follows:

Explanans X is linked to the desired explanandum E only indirectly, by means of E's closeness to E'. Hence, a special problem arises about the conditions under which such an approximative explanation is really acceptable.

Theories of chemistry have come to depend inescapably upon the framework provided by atomic physics. This is a dependency which is anchored upon the ability within quantum theory — in particular since the wave mechanics of Schrödinger — to account for chemical bonding by the pairing of electrons, the stability of their resulting structures, and thereby to provide a basis from which to deal with chemical behavior generally. The optimism generated fifty years ago by the initial attempts to apply this framework1 has bolstered the supposition, within the sciences, that chemistry is exclusively an experimental science, it has no theory except what it borrows from atomic physics. It seems to have established, within philosophy, so secure an example of intertheoretic reduction that however often we hear it cited, it remains a case rarely, if ever, examined.

In his intellectual autobiography, Rudolf Carnap describes a dispute between Otto Neurath and the rest of the members of the Vienna Circle.