How are scientific concepts formed? This question was central for positivist and operationalist philosophers concerned to root scientific thought directly in experience (e.g., Hempel 1952, Bridgman 1927). Although the positivist program has been abandoned, the current interest in the philosophy of scientific discovery shows the need for a theory of conceptual development. This paper offers a theory of how new concepts can arise, not by abstraction from experience or by definition, but by conceptual combination. Such combination produces a new concept as a non-linear, non-definitional amalgam of existing concepts. After proposing rules intended to account for a variety of mundane cases of conceptual combination, the paper presents several illustrations of how scientific concepts have arisen through combination.

According to a strict empiricist position (Hume 1739), concepts are abstracted from experience or formed by definition. You form the concept of a dog by experiencing a number of examples of dogs and extracting the common features.

Several 17th- and 18th-century natural philosophers and methodologists aimed to establish a body of positive scientific results which were more than hypothetical or conjectural. They made exaggerated claims for the truth and empirical pedigree of their scientific conclusions. Since the mid-19th century, investigators have been more circumspect; and, increasingly in the 20th century, reflective scientists have become explicit fallibilists. They frequently acknowledge that current results are unlikely to endure.

Seemingly in keeping with the working scientists own attitude, many, perhaps most, contemporary philosophers of science—in the wake of dramatic revolutions in several sciences and philosophical attacks on the quest for certainty—now take for granted the hypothetical, the conjectural nature of all Interesting scientific claims, particularly deep theoretical claims. According to these philosophers, the lesson we have learned from 19th-century methodologists is that 17th-century discovery programs, which aimed to justify theories by “deduction from the phenomena”, were only utopian dreams.

Could there have been a more down to earth philosopher of science than Charles Saunders Peirce? His later years are characterized by metaphysical opinions that are speculative. Yet throughout his career, his view of scientific method was quintessentially pragmatic. For instance, he reckons the testability of scientific hypotheses by these oft-repeated standards: “money, time, energy, thought” (7.220). He came to the philosophy of science from the laboratory by a route involving a dispute with leading scientists, carried on in the scientific journals and congresses of his day, and reported to the International Geodetic Committee. The dispute concerned the soundness of the design of certain experiments carried out by Peirce and others to measure the strength of the gravitational field at the Earth's surface. (Peirce's own account of these events is given in 7.1 - 7.20.

Cognitivism is now rather clearly the dominant approach in psychology. Philosophers such as Putnam (1975), Dennett (1978), Lycan (1981), and Cummins (1983) have supported the cognitivist strategy by proposing that mental states are to be defined functionally in terms of their interactions with other mental states. One of the most prominent features of the cognitivist-functionalist position is the autonomy it is thought to bestow upon psychology. Psychology, as viewed from this perspective, describes the processing of mental representations within the mind-brain and does not need to concern itself with the neurological activity that underlies its operation.

The arguments that are raised to support the autonomy of psychology from neurophysiology are arguments against a strong reductionist position that makes psychology dispensable (a view held, for example, by P.S. Churchland (1980)). The arguments are designed to show that there are regularities in nature that are not captured in the laws of physics.

In an interesting recent paper, Geroch (1984) has sought to state the content of, “and to some extent to advocate” a view toward quantum mechanics which is at least closely akin to the view advanced by Everett, known popularly (but, arguably, misleadingly) as the many-worlds interpretation. While sensitive to objections along the lines of that by Stein quoted above, Geroch believes that these can adequately be answered. The key to answering such objections is to argue that physics is not responsible for accounting for our experience of physical phenomena—it is enough that it account for the physical phenomena themselves. It is this last argument that I wish to examine in this paper, both in general, and specifically in its application to Everett's interpretation of quantum mechanics.

What are the eplstemolog leal virtues that we value In scientific theories? In The Scientific Image. Bas van Fraassen (1980) argues that the preeminent virtue Is empirical adequacy: the capacity of a theory to “save the phenomena”. In an Intuitive sense a theory Is empirically adequate if “What It says about the observable things and events in this world is true.” This point of view, according to which the eplstemologlcal aim of science is to save the phenomena, van Fraassen calls constructive empiricism. In an earlier “constructive critique” of constructive empiricism (1983), I argued that there were serious problems In van Fraassen's proposal for extending the notion of empirical adequacy to statistical theories. In that paper I suggested an alternative definition for the empirical adequacy of a simple statistical hypothesis: roughly speaking, the idea Is that a simple statistical hypothesis Is empirically adequate provided that there does not exist another simple hypothesis which has a higher likelihood relative to the set of all phenomena (past, present, and future) In the actual world.

Here follows a puzzle for those attempting to explain inductive inference. Suppose one hypothesis, H, implies another, I. Suppose further we obtain evidence, E, deemed sufficient to warrant the rejection of I. Is this evidence sufficient to warrant the rejection of H as well?

It is very tempting to answer this question affirmatively. The inference in question appears to be justified by a single use of modus tollens or denying the consequent:

A testament to the obvious plausibility of this sort of inference is the ubiquity of its endorsement in treatises on inductive inference. See, for example, Popper (Popper 1935, p. 76), Hempel (Hempel 1945, p. 31), Hacking (Hacking 1976, p. 33), or Savage (Savage 1954, p. 32). Indeed, a referee reporting on another paper called the following “rule of implication”:

(+) If hypothesis H implies hypothesis I, then evidence sufficient to warrant the rejection of I, in turn warrants the rejection of H.

The idea that events in nature are, somehow, governed by natural laws, and that to understand those laws would be a key to understanding the events—is a very old one. Whether abstracted from a religious context, or arrived at through independent reflection about nature, this idea first became prominent in the early days of what we now call science. Since the time geometry was worked out in axiomatic form, some 2,300 years ago, the combination of its deductive power, together with the understanding it yielded of the fundamental physical realm of space, I think it is fair to say geometry has stood as the very model of what we ought to expect laws of nature to look like, as well as how they ought to function.

This phenomenon can be seen in the work of Euclid himself, especially that in mechanics (Clagett 1955).

Accounts of the structure of scientific theories in twentieth century philosophy of science have tended to focus on epistemological questions to the near exclusion of the metaphysical consequences of their solutions. This emphasis can be understood as resulting from the interest and concerns of the logical positivist movement and continued by proponents of the so-called “received view” of theories (Suppe 1977). One area where this division of interest has been a barrier to progress is understanding the growth of scientific knowledge.

While philosophers as diverse as Nagel, Hempel, Kuhn, and Lakatos have attempted to explain the historical course of science, they have largely built their views on foundations designed to explicate the relation between theories (paradigms, scientific research programs) and the world. Thus, the received view's model of progress via theory (or indeed, whole science) reduction (e.g., Nagel 1961) rests on its largely syntactic/logical account of theory structure.

Bas van Fraassen has formulated a semantical or model theoretic analysis of the structure of scientific theories. He contrasts his semantical approach with the syntactic approach of the logical positivists and argues that his theory is preferable on a number of grounds. My aims in this paper will be threefold. First, I will present a brief description of van Fraassen's approach. Secondly, I will compare his theory with that of the logical positivists and, in so doing, identify the virtues that van Fraassen believes recommend, his theory over the positivist theory. Finally, I will raise an objection that is positivist in spirit which suggests that, epistemically, his account is indistinguishable from classical historicism.

For van Fraassen, “To present a theory is to specify a family of structures, its models; and secondly, to specify certain parts of these models (the empirical substructures) as candidates for the direct representation of observable phenomena.

One of the central problems of Kuhn's theory of scientific development is its reductionistic and local approach, i.e., its application of one basic pattern of scientific development—Kuhnian revolutions or K-revolutions—to only one area of science, the physical sciences. But the question arises as to why significant developmental episodes in the history of physics and chemistry only should be taken as paradigm cases, especially inasmuch as Kuhn himself claims to be describing and explaining the development of science generally. As Kuhn says, scientific revolutions “are at the heart of the most significant episodes of scientific development.“(Kuhn 1962, p. 140).

I shall take a somewhat radical position. First, I shall argue that the concepts of paradigm shift and scientific revolution are asymmetrical. That is, paradigm shifts represent a genus, and scientific revolutions are species within that genus; and whereas all K-revolutions are paradigm shifts, the converse does not hold.

There is a conflict between the intuition of scientists, and even many nonscientists, who see science as making progress, as being, in fact, the paradigm case of what it means to have progress in knowledge, and the writings of many philosophers of science who may be labeled ‘historicists’. For the most part the historicists do not deny progress outright. However, they have undercut previously accepted accounts of how science progresses.

Thomas Kuhn (1962) proposed that there are theories which are not only incompatible but also semantically incommensurable in order to explain historical evidence that scientists who hold consecutive theories often fail to come to terms with each other, being unable to resolve differences by appeal to evidence, authority or convention. Despite Kuhn's objections, this thesis has generally been interpreted by friends and foes alike so as to preclude direct rational communication across revolutionary divides in science. In this paper, I will sketch a weaker form of incommensurability which allows eventual comparison of incommensurable theories, but is consistent with Kuhn's model of science.

According to Kuhn, mature science normally operates within a disciplinary matrix which determines the nature of both scientific problems and their solutions. Such a matrix is composed of an articulated formalism, together with a set of standard problems called “exemplars”, and a set of preferred analogies for extending the exemplars to new applications.

A number of evaluations of the general prospects of Carnap's program of inductive logic have been presented on different occasions (e.g., Adler 1980, Jeffrey 1973, Kuipers 1984a, Lakatos 1968, Niiniluoto 1982). Niiniluoto's and my own account are the most positive about the internal strength of the program, provided one includes open systems of inductive probability, i.e., systems with non-zero probabilities for contingent universal generalizations. As is well-known, this was done for the first time by Hintikka.

This paper will not add a new evaluation to the list. Instead, I will show that the internal heuristic of Carnap's restricted program is already strong enough to solve one of its major problems, i.e., the problem of integrating analogy influences in systems of inductive probability.

Carnap has given considerable attention to the problem of analogy; to be precise, to what he called analogy by similarity, or analogy influences based on a distance function between the predicates.

In their classic work on the logic of explanation Hempel and Oppenheim (1948) claim that to explain the phenomena in the world of our experience is to answer the question “why?”, rather than the question “what?”. But there is a source of embarrassment, viz., the underdeveloped logic of why-questions itself. In this paper I want to explore some recent advances which, I shall argue, help us to overcome some of the embarrassment.

One way in which the logic of questions could help is proposed by Jaakko Hintikka. According to Hintikka (1981a) we can look upon scientific inquiry as a series of questions put to nature. For instance a scientist who tests a theory first derives its consequences C1,...,Cn and then asks nature questions of the form “Is CQ true?”. The inquirer (querier) here attempts to induce from nature one of the conclusive answers “Co is true” or “Co is not true”.

It has been suggested by Larry Laudan and others that scientific inquiry be conceptualized as a problem-solving and question-answering activity. This approach will be called here the interrogative conception of scientific inquiry. The logic of such a conception of scientific inquiry has not been studied systematically, however. In this paper, some of the main aspects of the logic on which the interrogative conception of science is based will. be briefly outlined on certain simplifying assumptions.

It is in fact natural to think of the typical task that confronts a scientist as a question-answer process in two different senses. Our scientist - I shall call her or him the Inquirer - is given a theory T as a starting-point and a question “C or C?” to be answered. Theory T is here assumed to be a truly general theory, i.e., it is assumed not to contain any individual constants.

Among objects and events in the world some have been or eventually will be observed. Others never have been and never will be. Current wisdom has it that among these latter are those that are nonetheless observable. Furthermore, the class of observable, not just observed, phenomena determines the empirical import of science. As van Fraassen tells us, a theory is empirically adequate if “what the theory says about what is observable (by us) it true.” (1980, p. 18). If there are quibbles with this picture, they are with the feasibility of a nonarbitrary or a theory-neutral distinction between the observable and the unobservable, but not with the idea that there are observable yet forever unobserved objects and events.

This, however, is precisely my quibble. Concrete items of the world—events and objects—are either observed or unobserved but, I urge, are not in addition observable or unobservable. This is not because I deny that a nonarbitrary or theory-neutral distinction can be drawn.