The system presented by the author in The Logical Foundations of Statistical Inference (Kyburg 1974) suffered from certain technical difficulties, and from a major practical difficulty; it was hard to be sure, in discussing examples and applications, when you had got hold of the right reference class. The present paper, concerned mainly with the characterization of randomness, resolves the technical difficulties and provides a well structured framework for the choice of a reference class. The definition of randomness that leads to this framework is simplified and clarified in a number of respects. It resolves certain puzzles raised by S. Spielman and W. Harper in their contributions to Profiles: Henry E. Kyburg, Jr. and Isaac Levi (R. Bogdan (ed.) 1982).

Plausibly, when we adopt a probabilistic standpoint any measure Cb(h, e) of the degree to which evidence e confirms (or is evidentially relevant to) hypothesis (or theory) h relative to background knowledge b should meet these five desiderata:

(1) Cb(h,e) > 0 when P(h/eb) > P(h/b; Cb(h,e) < 0 when P(h/eb) < P(h/b); Cb(h,e) = 0 when P(h/eb) = P(h/b).

(2) Cb(h,e) is some function of the values P(·/b) and P(·/·b) assume on the at most sixteen truth-functional combinations of e and h.

(3) If P(e/hb) < P(f/hb) and P(e/b) = P(f/b) then Cb(h,e) ≤ Cb(h,f); if P(e/hb) = P(f/hb) and P(e/b) < P(f/b) then Cb(h,e) ≥ Cb(h,f).

(4) Cb(h,ef) – Cb(h,eg) is fully determined by Cb(h,e) and Cbe(h,f) – Cbe(h,g); if Cb(h,ef) = 0 then Cb(h,e) + Cbe(h,f) = 0.

(5) If P(e/hb) = P(e/tb) then Cb(h,e) = Cb(t,e).

Biological science does not try to distinguish between health and disease. Biology is concerned with the interaction between living organisms and their environment. What we call health or disease is quite irrelevant.

These reactions between the individual and his environment are complex. The individual and his surroundings form an integrated system which we can arbitrarily divide into two parts. There is an “external” component, by which we mean such factors as light, heat, percentage of oxygen in the air, quantity of minerals or vitamins in food, micro-organisms in food or air, and so on. These can induce changes in what we arbitrarily call the “internal” component. Here we include such crude factors as anatomical structures, or finer details like composition of intercellular fluid, or secretions of glands, or changes of electrical potential in nerve or muscle.

While many philosophers of science have accorded special evidential significance to tests whose results are “novel facts”, there continues to be disagreement over both the definition of novelty and why it should matter. The view of novelty favored by Giere, Lakatos, Worrall and many others is that of use-novelty: An accordance between evidence e and hypothesis h provides a genuine test of h only if e is not used in h's construction. I argue that what lies behind the intuition that novelty matters is the deeper intuition that severe tests matter. I set out a criterion of severity akin to the notion of a test's power in Neyman-Pearson statistics. I argue that tests which are use-novel may fail to be severe, and tests that are severe may fail to be use-novel. I discuss the 1919 eclipse data as a severe test of Einstein's law of gravity.

We statisticians, with our specific concern for uncertainty, are even more liable than other practical men to encounter philosophy, whether we like it or not. For my part, I like it comparatively well. That is why the honor of opening this session of discussion has come to me, though my background makes my knowledge and idiom somewhat different from your own.

Preanalytically, we are all scientific realists. But both philosophers and scientists become uncomfortable when forced into analysis. In the case of scientists, this discomfort often arises from practical difficulties in setting out a carefully described set of objects which adequately account for the phenomena with which they are concerned. This paper offers a set of representative examples of these difficulties for contemporary physicists. These examples challenge the traditional realist vision of mature scientific activity as struggling toward an ontologically well-defined world picture. They challenge antirealist alternatives as well.

We examine, from the partial structures perspective, two forms of applicability of mathematics: at the “bottom” level, the applicability of theoretical structures to the “appearances”, and at the “top” level, the applicability of mathematical to physical theories. We argue that, to accommodate these two forms of applicability, the partial structures approach needs to be extended to include a notion of “partial homomorphism”. As a case study, we present London's analysis of the superfluid behavior of liquid helium in terms of Bose-Einstein statistics. This involved both the introduction of group theory at the top level, and some modeling at the “phenomenological” level, and thus provides a nice example of the relationships we are interested in. We conclude with a discussion of the “autonomy” of London's model.

The trouble with the idea of measurement is its seeming clarity, its obviousness, its implicit claim to finality in any inquisotory discourse. Its status in philosophy of science is taken to be utterly primitive; hence the difficulties it embodies, if any, tend to escape detection and scrutiny. Yet it cannot be primitive in the sense of being exempt from analysis; for if it were every measurement would require to be simply accepted as a protocol of truth, and one should never ask which of two conflicting measurements is correct, or preferable. Such questions are continually being asked, and their propriety in science indicates that even measurement, with its implication of simplicity and adroitness, points beyond itself to other matters of importance on which it relies for validation.

Real kinds or categories, according to conventional wisdom, enter into lawlike generalizations, while nominal kinds do not. Thus, gold but not jewelry is a real kind. However, by such a criterion, few if any kinds or systems of classification employed in the social science are real, for the social sciences offer, at best, only restricted generalizations. Thus, according to conventional wisdom, race and class are on a par with telephone area codes and postal zones; all are nominal rather than real. I propose an account of real kinds that recognizes the current reality of race but not zip codes and shows how a kind can be both constructed and real. One virtue of such an understanding of realism is the light shed on our current practice of racial classification. Race is not a real biological kind but neither is race a myth or illusion. However, the question of whether a social kind is real is separate from whether the category is legitimate. W. E. B. Du Bois maintained that while there are no biological races, race is real and should be conserved. My aim, in this paper, is not to argue for the legitimacy or conservation of race but to defend Du Bois's idea that kinds of people can be both made up and real and provide an understanding of realism that does justice to the social sciences.

Several articles have recently appeared arguing that there really are no viable alternatives to mechanistic explanation in the biological sciences (Kaplan and Bechtel; Kaplan and Craver). We argue that mechanistic explanation is defined by localization and decomposition. We argue further that systems neuroscience contains explanations that violate both localization and decomposition. We conclude that the mechanistic model of explanation needs to either stretch to now include explanations wherein localization or decomposition fail or acknowledge that there are counterexamples to mechanistic explanation in the biological sciences.

1. The problem. The concept of confirmation of an hypothesis by empirical evidence is of fundamental importance in the methodology of empirical science. For, first of all, a sentence cannot even be considered as expressing an empirical hypothesis at all unless it is theoretically capable of confirmation or disconfirmation, i.e. unless the kind of evidence can be characterized whose occurrence would confirm, or disconfirm, the sentence in question. And secondly, the acceptance or rejection of a sentence which does represent an empirical hypothesis is determined, in scientific procedure, by the degree to which it is confirmed by relevant evidence.

Thought experiments in science are merely picturesque argumentation. I support this view in various ways, including the claim that it follows from the fact that thought experiments can err but can still be used reliably. The view is defended against alternatives proposed by my cosymposiasts.

This article examines ontological/dynamical aspects of emergence, specifically the micro-macro relation in cases of universal behavior. I discuss superconductivity as an emergent phenomenon, showing why microphysical features such as Cooper pairing are not necessary for deriving characteristic properties such as infinite conductivity. I claim that the difficulties surrounding the thermodynamic limit in explaining phase transitions can be countered by showing how renormalization group techniques facilitate an understanding of the physics behind the mathematics, enabling us to clarify epistemic and ontological aspects of emergence. I close with a discussion of the impact of these issues for questions concerning natural kinds.

Robustness is a common platitude: hypotheses are better supported with evidence generated by multiple techniques that rely on different background assumptions. Robustness has been put to numerous epistemic tasks, including the demarcation of artifacts from real entities, countering the “experimenter's regress,” and resolving evidential discordance. Despite the frequency of appeals to robustness, the notion itself has received scant critique. Arguments based on robustness can give incorrect conclusions. More worrying is that although robustness may be valuable in ideal evidential circumstances (i.e., when evidence is concordant), often when a variety of evidence is available from multiple techniques, the evidence is discordant.

It is shown that quantum mechanics cannot be formulated as a stochastic theory involving a probability distribution function of position and momentum. This is done by showing that the most general distribution function which yields the proper quantum mechanical marginal distributions cannot consistently be used to predict the expectations of observables if phase space integration is used. Implications relating to the possibility of establishing a “hidden” variable theory of quantum mechanics are discussed.

A Bayesian account of the virtue of unification is given. On this account, the ability of a theory to unify disparate phenomena consists in the ability of the theory to render such phenomena informationally relevant to each other. It is shown that such ability contributes to the evidential support of the theory, and hence that preference for theories that unify the phenomena need not, on a Bayesian account, be built into the prior probabilities of theories.

In this article I assess the problems and prospects of a microstructural approach to chemical substances. Saul Kripke and Hilary Putnam famously claimed that to be gold is to have atomic number 79 and to be water is to be H2O. I relate the first claim to the concept of element in the history of chemistry, arguing that the reference of element names is determined by atomic number. Compounds are more difficult: water is so complex and heterogeneous at the molecular level that `water is H2O’ seems false under some interpretations. I sketch a response to this problem.

In this paper we first briefly review Bell's (1964, 1966) Theorem to see how it invalidates any deterministic “hidden variable” account of the apparent indeterminacy of quantum mechanics (QM). Then we show that quantum uncertainty, at the level of DNA mutations, can “percolate” up to have major populational effects. Interesting as this point may be it does not show any autonomous indeterminism of the evolutionary process. In the next two sections we investigate drift and natural selection as the locus of autonomous biological indeterminacy. Here we conclude that the population-level indeterminacy of natural selection and drift are ultimately based on the assumption of a fundamental indeterminacy at the level of the lives and deaths of individual organisms. The following section examines this assumption and defends it from the determinists' attack. Then we show that, even if one rejects the assumption, there is still an important reason why one might think evolutionary theory (ET) is autonomously indeterministic. In the concluding section we contrast the arguments we have mounted against a deterministic hidden variable account of ET with the proof of the impossibility of such an account of QM.

The fundamental problem considered is that of the existence of a joint probability distribution for momentum and position at a given instant. The philosophical interest of this problem is that for the potential energy functions (or Hamiltonians) corresponding to many simple experimental situations, the joint “distribution” derived by the methods of Wigner and Moyal is not a genuine probability distribution at all. The implications of these results for the interpretation of the Heisenberg uncertainty principle are analyzed. The final section consists of some observations concerning the axiomatic foundations of quantum mechanics.

The method of positron emission tomography (PET imaging) illustrates the circular logic popular in subtractive neuroimaging and linear reductive cognitive psychology. Both require that strictly feed-forward, modular, cognitive components exist, before the fact, to justify the inference of particular components from images (or other observables) after the fact. Also, both require a “true” componential theory of cognition and laboratory tasks, before the fact, to guarantee reliable choices for subtractive contrasts. None of these possibilities are likely. Consequently, linear reductive analysis has failed to yield general, reliable, componential accounts.