Advocates have sought to prove that underdetermination obtains because all theories have empirical equivalents. But algorithms for generating empirical equivalents simply exchange underdetermination for familiar philosophical chestnuts, while the few convincing examples of empirical equivalents will not support the desired sweeping conclusions. Nonetheless, underdetermination does not depend on empirical equivalents: our warrant for current theories is equally undermined by presently unconceived alternatives as well-confirmed merely by the existing evidence, so long as this transient predicament recurs for each theory and body of evidence we consider. The historical record supports the claim that this recurrent, transient underdetermination predicament is our own.

This paper examines in detail two paths that lead to Structural Realism (SR), viz. a substantive philosophical position which asserts that only the structure of the world is knowable. The upward path is any attempt to begin with empiricist premises and reach a sustainable realist position. (It has been advocated by Russell, Weyl, and Maxwell among others.) The downward path is any attempt to start from realist premises and construct a weaker realist position. (It has been recently advocated by Worrall, French, and Ladyman.) This paper unravels and criticizes the metaphysical presuppositions of both paths to SR. It questions its very possibility as a substantive—and viable—realist thesis.

Arthur Fine and André Kukla have argued that realism and instrumentalism are indifferent with respect to scientific practice. I argue that this claim is ambiguous. One interpretation is that for any practice, the fact that that practice yields predictively successful theories is evidentially indifferent between scientific realism and instrumentalism. On the second construal, the claim is that for any practice, adoption of that practice by a scientist is indifferent between their being a realist or instrumentalist. I argue that there are no good arguments for the indifference claim under the second interpretation, and good reasons to think that it is false.

“Scientific theories are maps of the natural world.” This metaphor is often used as part of a deflationary argument for a weak but relatively global version of scientific realism, a version that recognizes the place of conventions, goals, and contingencies in scientific representations, while maintaining that they are typically true in a clear and literal sense. By examining, in a naturalistic way, some relationships between maps and what they map, we question the scope and value of realist construals of maps—and by extension of scientific representations. Deflationary philosophy of science requires more variegated stances.

Many discussions between realists and non-realists have centered on the issue of reference, especially whether there is referential stability during theory change. In this paper, I shall summarize the debate, sketching the problems that remain within the two opposing positions, and show that both have ended on their own slippery slope, sliding away from their original position toward that of their opponents. In the search for a viable intermediate position, I shall then suggest an account of reference which, to a degree, follows the causal theory in explaining reference as carving the world at its joints. Contrary to the causal theory, however, I submit that this world is a phenomenal world whose variable joints exist only in a historical process in which they are transmitted gradually from one generation to the next. According to this account, the joints of the phenomenal world are constituted by family resemblance, where bundles of features that span bounded areas in perceptual space underlie the joints. Furthermore, the integrity of the cognitive process by which these joints are recognized depends on a transmission process by which new generations are presented with given joints and bundles by the preceding generation. Contrary to a traditional realist account, this heritage from the preceding generation may be transformed into new joints and bundles before transmission to new generations. This permits a continuous process of referential change in which the joints and bundles at different stages in the development of a theory can be connected by chains-of-reasoning.

A variety of recent philosophical discussions, particularly on topics relating to complexity, have begun to reemploy the concept of ‘emergence’. Although multiple concepts of ‘emergence’ are available, little effort has been made to systematically distinguish them. In this paper, I provide a taxonomy of higher-order properties that (inter alia) distinguishes three classes of emergent properties: (1) ontologically basic properties of complex entities, such as the mythical vital properties, (2) fully configurational properties, such as mental properties as they are conceived of by functionalists and computationalists, and (3) highly configurational/holistic properties, such as the higher-level patterns characteristic of complex dynamical systems. Or more simply: emergence as ontological liberality, emergence as multiple realizability, and emergence as interactive complexity.

The main thrust of the paper involves a theoretical and philosophical analysis of the claim made in September 1999 that atomic orbitals have been directly imaged for the first time. After a brief account of the recent claims the paper reviews the development of the orbit and later orbital concepts and analyzes the theoretical status of atomic orbitals. The conclusion is that contrary to these claims, atomic orbitals have not in fact been observed. The non-referring nature of modern atomic orbitals is discussed in the context of Laudan's writings on realism, the success of theories, and whether or not scientific terms refer. I conclude that the failure to observe orbitals is a good prima facie case for divorcing the success of theories from the question of whether their central terms refer. The added relevance of this case is that it concerns a current and highly successful theory. Finally, the relevance of this ‘floating model’ to contemporary discussions on scientific models is briefly considered.

According to Hugh Mellor in Real Time II (1998, Ch. 12), assuming the logical independence of causal facts and the ‘law of large numbers’, causal loops are impossible because if they were possible they would produce inconsistent sets of frequencies. I clarify the argument, and argue that it would be preferable to abandon the relevant independence assumption in the case of causal loops.

The existence of “undermining futures” appears to show that a contradiction can be deduced from the conjunction of Humean supervenience (HS) about chance and the Principal Principle. A number of strategies for rescuing HS from this problem have been proposed recently. In this paper, a novel way of defending HS from the threat is presented, and it is argued that this defense has advantages not shared by others. In particular, it requires no revisionism about chance, and it is equally available to defenders of HS who hold HS to be necessary and those who hold it to be contingent.

The Principle of Indifference, which dictates that we ought to assign two outcomes equal probability in the absence of known reasons to do otherwise, is vulnerable to well-known objections. Nevertheless, the appeal of the principle, and of symmetry-based assignments of equal probability, persists. We show that, relative to a given class of symmetries satisfying certain properties, we are justified in calling certain outcomes equally probable, and more generally, in defining what we call relative probabilities. Relative probabilities are useful in providing a generalized approach to conditionalization. The technique is illustrated by application to simple examples.

A Bayesian account of independent evidential support is outlined. This account is partly inspired by the work of C. S. Peirce. I show that a large class of quantitative Bayesian measures of confirmation satisfy some basic desiderata suggested by Peirce for adequate accounts of independent evidence. I argue that, by considering further natural constraints on a probabilistic account of independent evidence, all but a very small class of Bayesian measures of confirmation can be ruled out. In closing, another application of my account to the problem of evidential diversity is also discussed.

Two Bayesian approaches to choosing between statistical models are contrasted. One of these is an approach which Bayesian statisticians regularly use for motivating the use of AIC, BIC, and other similar model selection criteria, and the other one is a new approach which has recently been proposed by Bandyopadhayay, Boik, and Basu. The latter approach is criticized, and the basic ideas of the former approach are presented in a way that makes them accessible to a philosophical audience. It is also pointed out that the former approach establishes a new, philosophically interesting connection between the notions of simplicity and informativeness.

Critics of Bayesianism often assert that scientists are not Bayesians. The widespread use of Bayesian statistics in the field of radiocarbon calibration is discussed in relation to this charge. This case study illustrates the willingness of scientists to use Bayesian statistics when the approach offers some advantage, while continuing to use orthodox methods in other contexts. The case of radiocarbon calibration, therefore, suggests a picture of statistical practice in science as eclectic and pragmatic rather than rigidly adhering to any one theoretical position.

Garber (1983) and Jeffrey (1991, 1995) have both proposed solutions to the old evidence problem. Jeffrey's solution, based on a new probability revision method called reparation, has been generalized to the case of uncertain old evidence and probabilistic new explanation in Wagner 1997, 1999. The present paper reformulates some of the latter work, highlighting the central role of Bayes factors and their associated uniformity principle, and extending the analysis to the case in which an hypothesis bears on a countable family of evidentiary propositions. This extension shows that no Garber-type approach is capable of reproducing the results of generalized reparation.

We use evidence from cognitive psychology and the history of science to examine the issue of the theory-ladenness of perceptual observation. This evidence shows that perception is theory-laden, but that it is only strongly theory-laden when the perceptual evidence is ambiguous or degraded, or when it requires a difficult perceptual judgment. We argue that debates about the theory-ladenness issue have focused too narrowly on the issue of perceptual experience, and that a full account of the scientific process requires an examination of theory-ladenness in attention, perception, data interpretation, data production, memory, and scientific communication. We conclude that the evidence for theory-ladenness does not lead to a relativist account of scientific knowledge.

In this paper I argue for the cognitive impenetrability of perception by undermining the argument from reentrant pathways. To do that I will adduce psychological and neuropsychological evidence showing that (a) early vision processing is not affected by our knowledge about specific objects and events, and (b) that the role of the descending pathways is to enable the early-vision processing modules to participate in higher-level visual or cognitive functions. My thesis is that a part of observation, which I will call perception, is bottom-up and theory neutral. As such, perception could play the role of common ground on which a naturalized epistemology can be built and relativism avoided.

Many recent cognitive studies reveal that human cognition is inherently perceptual, sharing systems with perception at both the conceptual and the neural levels. This paper introduces Barsalou's theory of perceptual symbols and explores its implications for philosophy of science. If perceptual symbols lie in the heart of conceptual processing, the process of attribute selection during concept representation, which is critical for defining similarity and thus for comparing taxonomies, can no longer be determined solely by background beliefs. The analogous nature of perceptual symbols and the spatial nature of intraconceptual relations impose new constraints on attribute selection. These constraints help people with different background beliefs select compatible attributes, which constitute a common “platform” for taxonomy comparison.

Recent discussion of the statistical character of evolutionary theory has centered around two positions: (1) Determinism combined with the claim that the statistical character is eliminable, a subjective interpretation of probability, and instrumentalism; (2) Indeterminism combined with the claim that the statistical character is ineliminable, a propensity interpretation of probability, and realism. I point out some internal problems in these positions and show that the relationship between determinism, eliminability, realism, and the interpretation of probability is more complex than previously assumed in this debate. Furthermore, I take some initial steps towards a more adequate account of the statistical character of evolutionary theory.

This paper uses a number of examples of diverse types and functions of models in evolutionary biology to argue that the demarcation between theory and practice, or “theory model” and “data model,” is often difficult to make. It is shown how both mathematical and laboratory models function as plausibility arguments, existence proofs, and refutations in the investigation of questions about the pattern and process of evolutionary history. I consider the consequences of this for the semantic approach to theories and theory confirmation. The paper attempts to reconcile the insights of both critics and advocates of the semantic approach to theories.