History of science, it has been argued, has benefited philosophers of science primarily by forcing them into greater contact with “real science.” In this paper I argue that additional major benefits arise from the importance of specifically historical considerations within philosophy of science. Loci for specifically historical investigations include: (1) making and evaluating rational reconstructions of particular theories and explanations, (2) estimating the degree of support earned by particular theories and theoretical claims, and (3) evaluating proposed philosophical norms for the evaluation of the degree of support for theories and the worth of explanations. More generally, I argue that theories develop and change structure with time, that (like biological species) they are historical entities. Accordingly, both the identification and the evaluation of theories are essentially historical in character.

This paper analyzes the generation and function of hitherto ignored or misrepresented interfield theories, theories which bridge two fields of science. Interfield theories are likely to be generated when two fields share an interest in explaining different aspects of the same phenomenon and when background knowledge already exists relating the two fields. The interfield theory functions to provide a solution to a characteristic type of theoretical problem: how are the relations between fields to be explained? In solving this problem, the interfield theory may provide answers to questions which arise in one field but cannot be answered within it alone, may focus attention on domain items not previously considered important, and may predict new domain items for one or both fields. Implications of this analysis for the problems of reduction and the unity and progress of science are mentioned.

Rudolf Carnap defended two quite different critiques of traditional philosophy: in addition to the much discussed verifiability criterion, he also proposed a critique based upon “formalizability.” Formalizability rests upon the principle of tolerance plus an acceptance of a linguistic methodology. Standard interpreters of Carnap (e.g., [7] and [8]) assume that the principle of tolerance (and, hence, formalizability) gains its argumentative support from verificationism. Carnap, in fact, kept the two critiques separate and independent. Indeed, verificationism is even, in spirit, inconsistent with tolerance. If the formalizability approach is emphasized, traditional metaphysics is reconstructed, not banished. Philosophical disputes remain rationally decidable, but metatheoretical in nature. Two results follow: Carnap's metaphilosophy cannot be rejected merely on the basis of rejections of verifiability. Second, Carnap's conclusion that all philosophy concerns language provides no reason for despair.

Simon Kochen and Ernst Specker's well-known argument ([6]) against hidden variable theories for quantum mechanics is also an argument against the possibility of quantum systems having, simultaneously, precise values for all of the dynamical quantities associated with such systems. Devices for defeating the argument were in the literature even before its publication, but recently Arthur Fine has raised a new difficulty ([4]). Fine points out that Kochen and Specker's argument requires the following principles:

In 1924, Bohr, Kramers and Slater tried to introduce into microphysics conservation principles that hold only on the average. This attempt was abandoned in the light of the Compton-Simon experiment (and its later refinements). Since that time, except for a moment of doubt in 1936, it has been thought that the classical conservation laws (for energy and momentum) hold in quantum theory for each individual interaction, in a way that yields the classical exchange-and-balance of momentum (or energy) familiar from the laws of elastic collisions. It has been thought, that is, that in each individual “collision” what one part of the total system loses in linear momentum (say) another part gains, so as to maintain the same total amount afterwards as before. To those familiar with discussions of the interpretation of quantum theory, however, it will be apparent that the very concepts needed to express this idea of an elastic collision are generally not admitted in the theory. For one needs the idea that both before and after collision the total system has a well-defined value for the conserved quantity and, moreover, that the various component parts of the system have well-defined values for those quantities out of which the conserved one is composed. But in the case of spin, for example, it is customary to say that although total spin may be conserved in certain interactions one cannot attribute values to the separate components of spin (in various directions) because the associated operators do not commute. (Thus the whole sum is not even the sum of its parts.) Moreover, one does not generally refer to the values of quantities except in eigenstates. Hence, in general, one would not refer to the value of the conserved quantity before and after interaction, except for interactions initiated in eigenstates.

The debate between Glymour ([2]) and Fine ([1]) hinges in part on a comparison of the width of the incoming wave packet in momentum space with the angles intercepted by the detectors in the Cross-Ramsey experiment. As Fine argues, it follows from the quantum formalism that the initial dispersion will be conserved in Compton scattering, and he allows that the Sum Rule is constrained by the statistical results of quantum mechanics. The Sum Rule may fail, but it will not fail in any way that violates quantum mechanics. Thus most of the time the individual momentum value does not change more than the width of the initial wave packet: usually the value of Qγe at t must equal the value of Qγ at t′ plus the value of Qe at t′ ± Δp/2, where Δp represents the dispersion in momentum for the initial gamma ray. Thus in order to refute Fine's thesis, the detectors must discriminate finely enough to pick out changeovers of individual values within Δp.

This paper takes a critical look at theory-free, statistical methodologies for processing and interpreting data taken from respondents answering a set of dichotomous (yes-no) questions. The basic issue concerns to what extent theoretical conclusions based on such analyses are invariant under a class of “informationally equivalent” question transformations. First the notion of Boolean equivalence of two question sets is discussed. Then Lazarsfeld's latent structure analysis is considered in detail. It is discovered that the best fitting latent model depends on which one of the many informationally equivalent question sets is used. This fact raises a number of methodological problems and pitfalls with latent structure analysis. Related problems with other methodologies are briefly discussed.