In this article, I explain how a variant of David Miller's argument concerning the language dependence of the accuracy of predictions can be applied to Joyce's notion of the accuracy of “estimates of numerical truth-values” (i.e., Joycean credences). This leads to a potential problem for Joyce's accuracy-dominance-based argument for the conclusion that credences (understood as “estimates of numerical truth-values” in Joyce's sense) should obey the probability calculus.

Acidity provides an interesting example of an everyday concept that developed fully into a scientific one; it is one of the oldest concepts in chemistry and remains an important one. However, up to now there has been no unity to it. Currently two standard theoretical definitions coexist (Br⊘nsted and Lowry's and Lewis's); the standard laboratory measure of acidity, namely the pH, only corresponds directly to the Br⊘nsted-Lowry concept. The lasting identity of the acidity concept in modern chemistry is based on the persistence of the quotidian concept. This is suggestive for considerations of other scientific concepts.

This essay unpacks a seeming paradox: a concept used to formulate, promote, and legitimate oppressive ideologies—a concept used to formulate mistaken, because they were typological, biological theories about human diversity—is, it seems, the same concept that now promises to deliver wonderful, socially sensitized, innovative results in social and genetic epidemiology. But how could that be? How could scientists expect a concept as problematic as ordinary race to deliver useful scientific results? I propose that there is a process for retranslating Ballungen race concepts in appropriate ways to make them fit and work within social scientific and bioscientific contexts.

In 1894, William Bateson objected to the terms “heredity” and “inheritance” in biology, on grounds of contamination with misleading notions from the everyday world. Yet after the rediscovery of Mendel's work in the spring of 1900, Bateson promoted that work as disclosing the “principles of heredity.” For historians of science, Bateson's change of mind provides a new angle on these terms at a crucial moment in their history. For philosophers of science, the case can serve as a reminder of the potential of Putnam's hypothesis of a division of linguistic labor for analyzing the semantic lives of scientific kind terms.

There is a growing recognition that fictions have a number of legitimate functions in science, even when it comes to scientific explanation. However, the question then arises, what distinguishes an explanatory fiction from a nonexplanatory one? Here I examine two cases—one in which there is a consensus in the scientific community that the fiction is explanatory and another in which the fiction is not explanatory. I shall show how my account of “model explanations” is able to explain this asymmetry, and argue that realism—of a more subtle form—does have a role in distinguishing explanatory from nonexplanatory fictions.

Some philosophers of science—myself included—appeal to fiction as an interpretation of the practice of modeling. This raises the specter of an incompatibility with realism, since fiction making is essentially nontruth regulated. I argue that the prima facie conflict can be resolved in two ways, each involving a distinct notion of fiction and a corresponding formulation of realism. The main goal of the essay is to describe these two packages. I comment also on how to choose between them.

The relation between models and the world is mediated by experimental procedures generating data that are used as evidence to evaluate the model. Data can serve as empirical evidence, for or against, only if they result from reliable experimental procedures. The aim of this article is to discuss the role of relevance judgments in the evaluation of reliability and to clarify the conditions under which reliability can be a strictly empirical matter. It is argued that reliability is a strictly empirical issue only in the restricted case in which the claim under test/investigation is about a data-generating procedure.

Inquiries into the nature of scientific modeling have tended to focus their attention on mathematical models and, relatedly, to think of nonconcrete models as mathematical structures. The arguments of this article are arguments for rethinking both tendencies. Nonmathematical models play an important role in the sciences, and our account of scientific modeling must accommodate that fact. One key to making such accommodations, moreover, is to recognize that one kind of thing we use the term ‘model’ to refer to is a collection of propositions.

A scientific theory offers models for the phenomena in its domain; these models involve theoretical quantities, and a model's structure is the set of relations it imposes on these quantities. A fundamental demand in scientific practice is for those quantities to be clearly and feasibly related to measurement. This demand for empirical grounding can be articulated by displaying the theory-dependent criteria for a procedure to count as a measurement and for identifying the quantity it measures.

Although most philosophical accounts about model/world relations focus on structural mappings such as isomorphism, similarity has long been discussed as an alternative account. Despite its attractions, proponents of the similarity view have not provided detailed accounts of what it means that a model is similar to a real-world target system. This article gives the outlines of such an account, drawing on the work of Amos Tversky.

Robustness concepts are often invoked to manage two obstacles confronting models of ecological systems: complexity and uncertainty. The intuitive idea is that any result derived from many idealized but credible models is thereby made more reliable or is better confirmed. An appropriate basis for this inference has proven elusive. Here, several representations of robustness analysis are vetted, paying particular attention to complex models of ecosystems and the global climate. The claim that robustness is itself confirmatory because robustness analysis employs a Bayesian variety-of-evidence argument is criticized, but recent overwhelming pessimism about robustness may have a silver lining.

The use of power to infer null hypotheses from negative results has recently come under severe attack. In this article, I show that the power of a test can justify accepting the null hypothesis. This argument also gives us a new powerful reason for not treating p-values and power as measures of the strength of evidence.

Many philosophers argue that the emphasis on commercializing scientific research—and particularly on patenting the results of research—is both epistemically and socially detrimental, in part because it inhibits the flow of information. One of the most important of these criticisms is the “tragedy of the anticommons” thesis. Some have attempted to test this thesis empirically, and many have argued that these empirical tests effectively falsify the thesis. I argue that they neither falsify nor disconfirm the thesis because they do not actually test it. Additionally, I argue that there is other evidence that actually supports the thesis.

The proponents of the entrepreneurial university have claimed that it implies adjustments in the normative structure of science. In this article, I will critically examine whether a qualitatively new kind of academic ethos can emerge from the commercialization of academic research. The traditional conception of norms of science as institutionalized imperatives is distinguished from the constructivist conception of norms as strategic or ideological resources. An empirical case study on the commercialization of the research of one academic language-technology group is presented. The case study does not support the constructivist conclusion that the norms of science are malleable at will.

Ghostwriting scandals are pervasive in industry-funded biomedical research, and most responses to them have presumed that they represent a sharp transgression of the norms of scientific authorship. I argue that in fact, ghostwriting represents a continuous extension of current socially accepted authorship practices. I claim that the radically collaborative, decentralized, interdisciplinary research that forms the gold standard in medicine is in an important sense unauthored, and that this poses a serious problem in applied social epistemology. It is no easy matter to find procedural or architectural solutions that can secure epistemic trust and accountability for collaborative publishing in biomedicine.

Psychometrically oriented researchers construe low interrater reliability measures for expert peer reviewers as damning for the practice of peer review. I argue that this perspective overlooks different forms of normatively appropriate disagreement among reviewers. Of special interest are Kuhnian questions about the extent to which variance in reviewer ratings can be accounted for by normatively appropriate disagreements about how to interpret and apply evaluative criteria within disciplines during times of normal science. Until these empirical-cum-philosophical analyses are done, it will remain unclear the extent to which low interrater reliability measures represent reasonable disagreement rather than arbitrary differences between reviewers.