In one version of the debate between scepticism and scientific realism, the sceptic denies and the realist affirms the ability of science to home in on the unknown truth. Germane to this debate is the familiar claim that probabilistic updating by Bayesian conditionalization will almost surely arrive at the truth so long as one does not “close the door” on a hypothesis by assigning it probability zero. In this paper, we observe, on behalf of the sceptic, that there exists a coherent agent who in fact fails on an uncountable set of possible worlds to find the truth about a simple, universal hypothesis H even though he “keeps the door open” concerning H until H is logically refuted and even though there is a trivial, logically driven method that converges to the truth about H no matter what; that nonetheless, this agent “almost surely” arrives at the truth in the usual, Bayesian sense because the uncountable set of circumstances in which he fails is assigned probability zero by the agent; and that the guarantee that the set of all circumstances in which the agent fails has probability zero hinges on the axiom of countable additivity.

Inferential statistical tests—such as analysis of variance, t-tests, chi-square and Wilcoxin signed ranks—now constitute a principal class of methods for the testing of scientific hypotheses. In particular, inferential statistics (when properly applied) are normally understood, for better or worse, as warranting a theoretical (typically causal) inference from an observed sample to an unobserved part of the population. These methods have been applied with great effect in domains such as population genetics, mechanics, cognitive, perceptual and social psychology, economics, and sociology, and it is not often that scientists in these fields eschew causal explanation in favor of the statement of brute correlations among properties. However, the appropriateness of a particular statistical test, and the reliability of the inference from sample to population, are subject to the application of certain statistical principles and concepts.

The realism-instrumentalism debate (RID) is often seen as the core debate in philosophy of science, yet increasing numbers of philosophers have followed Fine's (1984) lead in questioning the point of this debate. Fine's largely unchallenged rendition of RID makes it is easy to see why. RID appears doomed to stalemate. According to Fine, realists and instrumentalists are both trying to account for the string of progressive episodes in the history of science. The two sides are said to agree on what those episodes are and that they constitute progress. However, it seems that every realist story of one such episode can be matched by an instrumentalist one—and vice versa. As Fine sees it, any such story is merely an attempt to extract surplus philosophical value from the historical labors of scientists. It only adds a misleading air of inevitability to their original efforts.

I shall discuss two recently suggested quantum experiments. These experiments lead to paradoxical situations. I will argue that in the framework of a particular Many Worlds Interpretation of quantum theory the paradoxes do not arise.

The first experiment is the “interaction-free measurement” (Elitzur and Vaidman 1993). The experimental group at Innsbruck headed by Prof. Zeilinger is working now on its realization. Before describing the experiment let me discuss a general question: “How do we know that there is an object in a region of space?”

The simplest case is when the object itself causes some physical changes outside the region:

• i) The object is charged, so there is a field outside. We can measure this field.

• ii) The object yields a potential outside. Aharonov and Bohm have taught us that potential even without field leads to a measurable effect.

• iii) The objects radiates photons or other particles.

Philosophers of science have not paid much attention to recent developments in quantum cosmology. This fact is surprising, since quantum cosmology is replete with conceptual issues involving (e.g.) the fundamental nature of time and space, the interpretation of quantum mechanics, and the ultimate meaning of probability. One notable exception, Quentin Smith, has recently examined the Hartle-Hawking (1983) proposal. Trying to make sense of the view, he resorts to an instrumentalist picture, which treats the proposal as merely a heuristic device for the algorithm responsible for predictions. While we do not examine Smith's account here, we would like to contrast it with the model presented in this note, in which a fully realistic interpretation of quantum cosmology is developed.

Recently there has been a resurgence of interest in the de Broglie-Bohm causal interpretation of quantum mechanics.

Although the relativistic, hot big bang (RHBB) model is generally regarded as “spectacularly successful: In short, it provides a reliable accounting of the history of the Universe from about 0.01 sec after the [big] bang until today, some 15 billion years later” (Turner 1992, 1), it has recently faced serious attacks from a number of physicists who cite long lists of (they claim) critical anomalies (e.g. Arp, et.al. 1990, Burbidge, 1992, Arp and van Flandern, 1992, Lerner, 1993a, Narlikar 1993). In this paper, we shall not attempt to adjudicate on such disputes, nor do we wish to take a position on the normative criteria that should govern science. Rather, our primary concern is to highlight the tensions between the normative criteria accepted by proponents of the RHBB (or demanded of rival theories) and the actual progress of the big bang program.

Bayesians assess the inductive support for theoretical hypotheses on the basis of two sorts of factors, one fairly objective, the other highly subjective. The objective factor consists of the likelihoods or direct inference probabilities that theoretical hypotheses specify for evidential events. This is the means by which evidence affects inductive support. The subjective factor consists of the prior probabilities assigned to the various competing hypotheses. For a Bayesian agent the prior probability of a hypothesis represents how plausible the agent considers the hypothesis to be before the impact of evidence is considered, and Bayesian agents may radically differ in their initial plausibility assessments. Bayes’ formula combines likelihoods with an agent's prior probabilities to produce the agent's posterior probability for each hypothesis. Posterior probabilities represent how plausible an agent considers the hypothesis to be after the evidence is taken into account.

We discuss two general issues concerning diverging sets of Bayesian (conditional) probabilities—divergence of “posteriors”—that can result with increasing evidence. Consider a set of probabilities typically, but not always, based on a set of Bayesian “priors.” Incorporating sets of probabilities, rather than relying on a single probability, is a useful way to provide a rigorous mathematical framework for studying sensitivity and robustness in Classical and Bayesian inference. See: Berger (1984, 1985, 1990); Lavine (1991); Huber and Strassen (1973); Walley (1991); and Wasserman and Kadane (1990). Also, sets of probabilities arise in group decision problems. See: Levi (1982); and Seidenfeld, Kadane, and Schervish (1989). Third, sets of probabilities are one consequence of weakening traditional axioms for uncertainty. See: Good (1952); Smith (1961); Kyburg (1961); Levi (1974); Fishburn (1986); Seidenfeld, Schervish, and Kadane (1990); and Walley (1991).

“Maximizing expected utility (MEU)” is one assumption of (strict) Bayesian decision theory [Savage 1972]. According to the principle of MEU, in a given decision situation, the decision maker should choose one of the alternatives with maximal expected utility [For an excellent discussion of decision theory, see Jeffrey 1990]. However, MEU as the foundation of Bayesian decision theory has been under attack. One counterexample that seems to dispute MEU as a rationality principle is offered by Daniel Ellsberg[Ellsberg 1961]. While discussing what has been dubbed as the Ellsberg paradox, I will consider briefly three decision principles, each of which is different from the MEU principle.

Following a period during which philosophers of science focused on theory to the near exclusion of experiment, a number of philosophers, historians and sociologists of science have, in one way or another, turned their attention to experimentation, instrumentation, and laboratory practices. Considerable work in philosophy of science of the last decade reflects this surge of interest in experiment, as promoted by Ackermann, Cartwright, Franklin, Galison, Giere, Hacking and others. Where has this movement taken us and where do we still have to go?

In asking this question, my focus is on that subset of the experimentalist movement whose members, following Ackermann (1989), I dub the “New Experimentalists”. Although their agendas differ, members of this group share the core thesis that aspects of experiment might offer an important, though largely untapped, resource for addressing key problems in philosophy of science.

When the Neyman-Pearson approach to the testing of statistical hypotheses was introduced in the 1930's and 1940's it was presented with an accompanying philosophy of science. Jerzy Neyman held a strong form of pragmatism. He sought not just to model scientific decision on practical decision, but in a sense to reduce it to practical decision. Neyman-Pearson techniques are used to compare hypotheses in the light of experimental data, and they contain an asymmetry, in that one hypothesis—the “null”—typically gets the benefit of the doubt. For Neyman, the choice of which hypothesis gets this benefit is governed by behavioral considerations. When presented as an account of testing in science this claim was met with vigorous resistance.

Though the philosophy was controversial, the methods became standard. Philosophers and statisticans have produced alternative interpretations of many distinctive features of these tests. But they are not the only ones who have had to give a new rationale for Neyman-Pearson methods.

This paper deals with the mutual relation of speculative considerations, experimental activity, and the attempt to establish “laws of nature.” I shall not give a general analysis, but instead study a historical example. It is taken from the work of Michael Faraday - one of the most original and most successful experimenters.

I shall start by giving a rough sketch of Faraday's view, thereby making clear the particular question on which my paper concentrates. In part three, I analyze an episode of Faraday's actual work, and conclude in part four with a rough analysis of the type of explanation given by Faraday.

Faraday repeatedly emphasizes the necessity of distinguishing “that knowledge which consists of assumption, by which I mean theory and hypothesis, from that which is the knowledge of facts and laws; never raising the former to the dignity or authority of the latter, nor confusing the latter more than is inevitable with the former.”

The purpose of this paper is to provide a systematic account of the role of anomaly in theory change in science. We adopt a naturalist approach to the philosophy of science (e.g., Giere 1985; Maffie 1990) and support our proposal with evidence from the history of science and from the psychology of science.

Previous discussions of the topic of anomaly in theory change have rarely used evidence from psychology. Probably this has been due to the relatively immature state of research in the psychology of science (cf. Gorman forthcoming; Gholson, Shadish, Neimeyer, and Houts 1989). However there is now enough research on the topic of belief change (cf. Chinn and Brewer 1993b) to provide a foundation for theory development in the area of responses to anomaly. Much of the research in this area uses data from undergraduates and children and not scientists (see Dunbar forthcoming, for an exception).

Popper's philosophy is founded on his early conceptions of the problem of demarcation, of its solution, and of the roles of epistemology and methodology. I shall give an account of these, and of his idea of the alternative, ‘naturalistic’ conceptions. I then hope to show that Popper's conceptions are confused.

At the beginning of The Logic of Scientific Discovery Popper tells us that “the task of the logic of scientific discovery, or the logic of knowledge, [is]… to analyse the method of the empirical sciences” (Popper 1959, 27), and that its first job is “to put forward a concept of empirical science, in order to make linguistic usage, now somewhat uncertain, as definite as possible, and in order to draw a clear line of demarcation between science and metaphysical ideas” (1959, 38-39).

Much of the history of science is missing much of the time. The missing history I have in mind is not missing because of lost documents or state secrecy. Nor is it missing because it is obscure. I am talking about central episodes in the history of science and absolutely central issues in the logical structure and evidential support of those episodes. In fact, the episodes are so central and well known, and the missing history (once seen) so obvious, that one feels compelled to hold that there must be a widespread systematic bias among historians and philosophers against seeing certain explanatory patterns.

My explanation for this missing history of science is that there is a strong tendency among historians and philosophers of science toward what I will call “Psychological Predictivism” to distinguish it from “Logical Predictivism.” Logical Predictivism is the position that, if an observed phenomenon provides good evidence for a hypothesis, then that hypothesis (plus unproblematic auxiliary hypotheses), predicts the phenomenon.

In the course of his forceful (1981) attack on “convergent realism”, Larry Laudan attempted to turn an influential pro-realist consideration on its head. Scientific realists have wondered how a theory could enjoy the sort of empirical predictive success exhibited by presently accepted theories in the “mature” sciences, and yet be radically wrong in what it claims is going on “behind” the empirical phenomena. If nothing like the electrons and other particles postulated by current physics exists, how can the theories that there are such things have made the great range of successful predictions that they do about phenomena observed in particle accelerators and the like? Laudan directs such realists to the history of science and to a list of “once successful” theories which are “(by present lights) non-referring”—a list of successful theories which gave a “central” role to notions that, according to theories we accept now, do not exist (1981, 26).

One of the main goals of no-collapse interpretations of quantum mechanics is to show how results of measurements are objectified, or, in a more poignant terminology, how in a measurement situation facts arise. One needs to show how in their dynamical evolution the measured system and, perhaps more importantly, the measuring apparatus acquire, or at least appear to acquire, the definite properties that are otherwise explained by the collapse of the state. This can be seen as a special case of the problem whether quantum mechanics alone can account for the classical behaviour of macroscopic systems; in particular, the permanent definiteness of at least those properties of macroscopic systems that are directly observable, such as the position of the pointer of a measuring apparatus. In other words, whether quantum mechanics can account for macrofacts of the familiar kind.

The decoherence approach to the measurement problem invokes dissipative interactions between a measuring apparatus and its environment to explain, within the context of ‘pure’ quantum mechanics (QM), why such devices appear to possess definite pointer readings. By ‘pure’ QM, I mean Schrödinger evolution with no wavefunction collapse. Several classes of interpretations of pure QM can rely on decoherence. One class is the ‘modal’ interpretations of van Fraassen (1979), Dieks (1989), Healey (1989), and others. Another class is the relative-state and many-world interpretations.

I will argue that decoherence cannot help these interpretations address the general metaphysical challenges raised against them. But decoherence can help pick out a ‘special’ basis that determines which observables receive definite values. I'll explore to what extent decoherence rescues the modal (biorthogonal) basis-selection rule, and Zurek's (environmental interaction) basis-selection rule, from the basis degeneracy problem and the imperfect measurement problem.

Consider the general form of the state of a quantum system, (eq. 1), where the are (non-degenerate) eigenvectors of an operator, A, corresponding to an observable, A. According to the standard interpretation of quantum mechanics (the precise details of which do not concern me here), A does not have a definite value on the system in the state . But now suppose we measure A by coupling the system to a measuring apparatus. The composite system is then (eq. 2) where the are the indicator-states of the apparatus. Again, the standard interpretation tells us that the apparatus is not in a definite indicator-state. But how can this be? Measurements do have definite results. The apparatus is always in a definite state.

I have just reviewed one version of the measurement problem. The modal interpretation of quantum mechanics proposes a solution.

It is clear that many philosophers of science are inclined to be sceptical about the scientific ambitions of economics. Sometimes this scepticism amounts to a forthright denial of economists’ entitlement to be granted the epistemic authority normally claimed by scientists (for example, Dupré 1993). A more modest ambivalence may be signalled by a philosopher's refusal to treat the intuitions of economists as data for the philosophy of science, in the way that (e.g.) physicists’ ontological and epistemological judgements are now routinely treated. The following paper is part of a much larger project that criticizes the basis for this scepticism, at least where microeconomics is concerned. This general project is partly motivated by our belief that philosophers have often viewed microeconomics from a highly distorting set of perspectives. One source of these distortions is a sociological accident: the philosophy of economics, as a part of the philosophy of science, has been largely owned since its outset by Popperians and their intellectual descendants.