The sensitive dependence on initial conditions (SDIC) associated with nonlinear models imposes limitations on the models’ predictive power. We draw attention to an additional limitation than has been underappreciated, namely, structural model error (SME). A model has SME if the model dynamics differ from the dynamics in the target system. If a nonlinear model has only the slightest SME, then its ability to generate decision-relevant predictions is compromised. Given a perfect model, we can take the effects of SDIC into account by substituting probabilistic predictions for point predictions. This route is foreclosed in the case of SME, which puts us in a worse epistemic situation than SDIC.

I defend the view that many biological categories are defined by homology against a series of arguments designed to show that all biological categories are defined, at least in part, by selected function. I show that categories of homology are ‘abnormality inclusive’—something often alleged to be unique to selected function categories. I show that classifications by selected function are logically dependent on classifications by homology, but not vice-versa. Finally, I reject the view that biologists must use considerations of selected function to abstract away from variation and pathology to form a canonical description of a class of biological systems.

The fact that the names of biological species refer independently of identifying descriptions does not support the view of Ghiselin and Hull that species are individuals. Species may be regarded as natural kinds whose members share an essence which distinguishes them from the members of other species and accounts for the fact that they are reproductively isolated from the members of other species. Because evolutionary theory requires that species be spatiotemporally localized their names cannot occur in scientific laws. If natural kind status is denied to species on this ground, it must also be denied to most classes of concrete entities which are now accorded such status.

The first part of this article finds Craver’s mutual manipulability theory (MM) of constitution inadequate, as it definitionally ties constitution to the feasibility of ideal experiments, which, however, are unrealizable in principle. As an alternative, the second part develops an abductive theory of constitution (NDC), which exploits the fact that phenomena and their constituents are unbreakably coupled via common causes. The best explanation for this fact is the existence of an additional dependence relation, namely, constitution. NDC has important ramifications for constitutional discovery—most notably, that there is no experimentum crucis for constitution, not even under ideal discovery circumstances.

In “Counterfactual Dependence and Time's Arrow”, David Lewis defends an analysis of counterfactuals intended to yield the asymmetry of counterfactual dependence: that later affairs depend counterfactually on earlier ones, and not the other way around. I argue that careful attention to the dynamical properties of thermodynamically irreversible processes shows that in many ordinary cases, Lewis's analysis fails to yield this asymmetry. Furthermore, the analysis fails in an instructive way: it teaches us something about the connection between the asymmetry of overdetermination and the asymmetry of entropy.

In Epistemic Cultures (1999), Karin Knorr Cetina argues that different scientific fields exhibit different epistemic cultures. She claims that in high energy physics (HEP) individual persons are displaced as epistemic subjects in favor of experiments themselves. In molecular biology (MB), by contrast, individual persons remain the primary epistemic subjects. Using Ed Hutchins’ (1995) account of navigation aboard a traditional US Navy ship as a prototype, I argue that both HEP and MB exhibit forms of distributed cognition. That is, in both fields cognition is distributed across individual persons and complex artifacts. The cognitive system producing the knowledge is heterogeneous. Nevertheless, in both fields we can reserve epistemic agency for the human components of these systems. We do not need to postulate new distributed cognitive agents, let alone ones exhibiting new forms of consciousness.

My Representing and Intervening (1983) concludes with what it calls an experimental argument for scientific realism about entities. The argument is evidently inapplicable to extragalactic astrophysics, but leaves open the possibility that there might be other grounds for scientific realism in that domain. Here I argue for antirealism in astrophysics, although not for any particular kind of antirealism. The argument is conducted by a detailed examination of some current research. It parallels the last chapter of (1983). Both represent the methodological opinion that abstract or semantic realism/antirealism debates are empty, and typically lead to confused or wrong conclusions because they pay so little attention to the details of a science.

Reasoning that compares two objects or situations to draw conclusions about previously unknown properties of one of them has traditionally been taken to be ampliative and probabilistic. I propose that it is apodeictic reasoning from a premise about isomorphic structures that is often uncertain, but which we may have good reasons to believe. I characterize the structures and their isomorphism, describe patterns of reasoning appropriate to them, and discuss some complications not immediately obvious.

Many scientific models lack an established representation relation to actual targets and instead refer to merely possible processes, background conditions, and results. This article shows how such models can be appraised. On the basis of the discussion of how-possibly explanations, five types of learning opportunities are distinguished. For each of these types, an example—from economics, biology, psychology, and sociology—is discussed. Contexts and purposes are identified in which the use of a model offers a genuine opportunity to learn. These learning opportunities offer novel justifications for modeling practices that fall between the cracks of standard representationalist appraisals of models.

Two major reasons feminists are concerned with science relate to science's social effects: that science can be a powerful ally in the struggle for equality for women; and that all too frequently science has been a generator and perpetuator of inequality. This concern with the social effects of science leads feminists to a different mode of appraising science from the purely epistemic one prized by most contemporary philosophers of science. The upshot, I suggest, is a new program for philosophy of science, a program for a socially responsible philosophy of science.

This paper develops an account of explanation in biology which does not involve appeal to laws of nature, at least as traditionally conceived. Explanatory generalizations in biology must satisfy a requirement that I call invariance, but need not satisfy most of the other standard criteria for lawfulness. Once this point is recognized, there is little motivation for regarding such generalizations as laws of nature. Some of the differences between invariance and the related notions of stability and resiliency, due respectively to Sandra Mitchell and Brian Skyrms, are explored.

Richard Jeffrey's generalization of Bayes' rule of conditioning follows, within the theory of belief functions, from Dempster's rule of combination and the rule of minimal extension. Both Jeffrey's rule and the theory of belief functions can and should be construed constructively, rather than normatively or descriptively. The theory of belief functions gives a more thorough analysis of how beliefs might be constructed than Jeffrey's rule does. The inadequacy of Bayesian conditioning is much more general than Jeffrey's examples of uncertain perception might suggest. The “parameter α” that Hartry Field has introduced into Jeffrey's rule corresponds to the “weight of evidence” of the theory of belief functions.

David Lewis (1979) defends the thesis of the asymmetry of overdetermination: later affairs are seldom overdetermined by earlier affairs, but earlier affairs are usually overdetermined by later affairs. Recently, Carol Cleland (2002) has argued that since the distinctive methodologies of historical science and experimental science exploit different aspects of this asymmetry, the methodology of historical science is just as good, epistemically speaking, as that of experimental science. This paper shows, first, that Cleland's epistemological conclusion does not follow from the thesis of the asymmetry of overdetermination, because overdetermination (in Lewis's sense) is compatible with epistemic underdetermination. The paper also shows, contra Cleland, that there is at least one interesting sense in which historical science is epistemically inferior to experimental science, after all, because local underdetermination problems are more widespread in historical than in experimental science.

What sort of claims do scientific models make and how do these claims then underwrite empirical successes such as explanations and reliable policy interventions? In this paper I propose answers to these questions for the class of models used throughout the social and biological sciences, namely idealized deductive ones with a causal interpretation. I argue that the two main existing accounts misrepresent how these models are actually used, and propose a new account.

This article suggests a ‘best alternative’ justification of induction (in the sense of Reichenbach) which is based on meta-induction. The meta-inductivist applies the principle of induction to all competing prediction methods which are accessible to her. It is demonstrated, and illustrated by computer simulations, that there exist meta-inductivistic prediction strategies whose success is approximately optimal among all accessible prediction methods in arbitrary possible worlds, and which dominate the success of every noninductive prediction strategy. The proposed justification of meta-induction is mathematically analytical. It implies, however, an a posteriori justification of object-induction based on the experiences in our world.

One of the tasks with which communication engineers are presented is that of devising a mechanism by which a significant sequence of words, a message, produced by somebody, the sender of the message, is reproduced at some other place, with the shortest practical time lag. The reproduction must be such that the receiver of the message will be able to understand what the sender meant by his message, at least, if he knows the sender's language (or, perhaps, his specific use of the language). The following illustration is typical: A writes on a sheet of paper “I love you” and wishes that B, 3000 miles away, should become aware of the full content of this message, with little delay and at a low cost. There will be institutions, in a capitalistic society, which will compete with each other in providing A, for a price, with the required service. Those companies which perform these services most satisfactorily, i.e., with an overall better combination of faithfulness, time lag, and cost, will get the job. The executives of these companies will hire engineers and put them to work on improving this overall combination.

Van Fraassen (1989) argues that Inference to the Best Explanation is incoherent in the sense that adopting it as a rule for belief change will make one susceptible to a dynamic Dutch book. The present paper argues against this. A strategy is described that allows us to infer to the best explanation free of charge.

A popular and plausible response against Laudan's “pessimistic induction” has been what I call “preservative realism,” which argues that there have actually been enough elements of scientific knowledge preserved through major theory-change processes, and that those elements can be accepted realistically. This paper argues against preservative realism, in particular through a critical review of Psillos's argument concerning the case of the caloric theory of heat. Contrary to his argument, the historical record of the caloric theory reveals that beliefs about the properties of material caloric, rejected by subsequent theories, were indeed central to the successes of the caloric theory. Therefore caloric remains a favorable case for Laudan. Further, I argue that even confirmed cases of preservation do not warrant an inference to truth.