The purpose of this article is to update and defend syntax-based (conserved DNA-sequence motifs) gene concepts. I show how syntax-based concepts have been and can be extended to accommodate complex cases of genome expression, regulation, and processing. In response to difficult cases and causal parity objections, I argue that the syntax-based approach fleshes out a deflationary concept that defines genes in terms of sequences and organizational features of the genome that contribute to a phenotype.

This essay examines the claim “path dependence entails irreversibility” from the point of view of evolutionary biology. I argue that evolutionary irreversibility possesses many faces, sometimes conflicting with path dependence. I propose an account of path dependence that does not rely on irreversibility and explain why it more naturally coexists with the notion of (contingent) irreversibility developed by the Belgian paleontologist Louis Dollo. However, I argue that we should not conceive of this relationship as necessary.

There is still a great deal of debate over what counts as a constraint and about how to assess experimentally the relative importance of constraints and selection in evolutionary history. I will argue that the notion of a constraint on variation, and thus the selection-constraint distinction, depends on two specifications: (1) what counts as a variant—constraints limit or bias the production of what? and (2) range of assessment—over what range of times or conditions is the variation assessed? Specifications 1 and 2 help us to understand empirical work on the relative importance of constraint and selection in evolution.

Analysis of two key ways of characterizing genes—as causes of phenotypic effects and as genomic DNA sequences—has yielded widespread pessimism that they can be united in a coherent gene concept. This raises important questions about the epistemology of genomic research: If analysis of a genome sequence cannot yield information about genes defined both in terms of their products and their DNA sequence, then what could we learn from it? I investigate basic tools of genomic analysis, argue that they do not reflect the application of either gene concept, and clarify how we learn from genomic research.

The theory of mind (ToM) deficit associated with autism spectrum disorder has been a central topic in the debate about the modularity of the mind. In a series of papers, Philip Gerrans and Valerie Stone argue that positing a ToM module does not best explain the deficits exhibited by individuals with autism. In this article, I first criticize Gerrans and Stone's account. Second, I discuss various studies of individuals with autism and argue that they are best explained by positing a higher-level, domain-specific ToM module.

In practice, scientists must convey data in a “representational style” (e.g., as a numerical array or visual representation). Various authors seek to explain the epistemic role of scientific visual representation in terms of formal conventions (e.g., Goodman, Perini, and Kulvicki). Goodman also tends to dismiss the epistemic relevance of human cognition. My position is that visual conventions are nonarbitrary, in that they play to scientists’ cognitive abilities and limitations. My account draws on Perini's formal analysis, scientific case studies, and empirical literature on global pattern detection in neurotypicals, autistics, and dyslexics.

This article encourages the use of explicit methods in linguistics by attempting to estimate the size of a linguistic data set. Such estimations are difficult because redundant data can easily pad the data set. To address this, I offer some explicit operationalizations of the data and their features. For linguistic data, negative associations do not indicate true redundancy, and yet for many measures they can be mathematically impossible to ignore. It is proven that this troublesome phenomenon has positive Lebesgue measure and is monotonically increasing and that these two features hold robustly in four different ways.

One of the obstacles facing massive modularity is how a pervasively modular mind might generate non-domain-specific thoughts by integrating the content produced by various domain-specific modules. Peter Carruthers has recently argued that the operations of the language faculty are constitutive of the process by which the human mind is able to integrate content from heterogeneous conceptual domains. In this article, I first argue that Carruthers's data do not provide support for either of two possible interpretations of his thesis. In addition, I provide empirical and theoretical reasons for thinking content integration is performed external to the language faculty.

Recently, in attempting to account for explanatory reasoning in probabilistic terms, Bayesians have proposed several measures of the strength of a potential explanation. These candidate measures of “explanatory power” arguably have interesting normative interpretations and consequences. What has not yet been investigated, however, is whether any of these measures are also descriptive of people's actual explanatory judgments. Here I present my own experimental work investigating this question. I argue that one measure in particular is an accurate descriptor of explanatory judgments. Then I briefly point to some implications of this result for the epistemology and the psychology of explanatory reasoning.

Many explanations in physics rely on idealized models of physical systems. These explanations fail to satisfy the conditions of standard accounts of explanation. Recently, some philosophers have claimed that idealizations can be used to underwrite explanation nonetheless but only when they are what have variously been called representational, Galilean, controllable, or harmless idealizations. This article argues that such a half-measure is untenable and that idealizations not of this sort can have explanatory capacities.

Paul Needham has claimed in several recent papers that Dalton's chemical atomism was not explanatory. I respond to his criticism of Dalton by arguing that explanation admits of degrees and that under a view that allows for a spectrum of explanatory value, it is possible to see ample worth in Dalton's atomistic explanations. Furthermore, I argue that even Duhem, who rejected atomism, acknowledged the explanatory worth of Dalton's atomism.

Objective chance, or the “big bad bug” of David Lewis's account of Humean Supervenience forces, as is well known, is a modification of the Principal Principle. Here, I argue that standard assumptions regarding conditional probabilities entail several puzzling consequences for Lewis's New Principle, namely, an apparent requirement to account for the chance of a theory of chance. These problems, I argue, cannot be adequately answered within the received framework, and so I suggest that an interpretation of conditional probabilities in terms of Alan Hájek's recent work might go some distance toward their resolution.

I argue that, in causal inference from many observational studies, the piecemeal collection of data can cause underdetermination, even if arbitrarily large amounts of reliable data are available. Two theorems reveal that, for any variable set V, there are causal theories over V that can be distinguished if and only if all variables are simultaneously measured. These results entail that, a priori, one cannot know which observational studies will be most informative with respect to the true causal theory describing V. Hence, scientific institutions may need to play a larger role in coordinating differing research programs.

P. Kyle Stanford defends the problem of unconceived alternatives, which maintains that scientists are unlikely to conceive of all the scientifically plausible alternatives to the theories they accept. Stanford's argument has been criticized on the grounds that the failure of individual scientists to conceive of relevant alternatives does not entail the failure of science as a corporate body to do so. I consider two replies to this criticism and find both lacking. In the process, I argue that Stanford does not provide evidence that there are likely scientifically plausible unconceived alternatives to scientific theories accepted now and in the future.

I contrast our own evidence for the hypothesis of organic fossil origins with that available in previous centuries, suggesting that the most powerful contemporary evidence consists in a form of projective support whose distinctive features are not well captured by familiar hypothetico-deductive, abductive, or even more recent and more technically sophisticated (e.g., Bayesian) accounts of scientific confirmation. I suggest that such accounts either misrepresent or ignore something important about the heterogeneous ways in which scientific hypotheses can be supported by evidence, and I go on to suggest that the search for any single such account may be misguided in any case.

Recently Halpern and Pearl and Hitchcock have presented influential accounts of actual (token) causation using Bayesian networks. These accounts have been deterministic. Here we present a probabilistic extension to these active path analyses of actual causation. The extension uses “soft” interventions to set distributions rather than just single values. The resulting account can handle at least as wide a range of examples as the original accounts, without assuming determinism.

This article addresses the question of whether we should conceive of mechanisms as productive of change in a regular way. I argue that, if mechanisms are characterized as fully regular, on the one hand, then not enough processes will count as mechanisms for them to be interesting or useful. If no appeal to regularity is made at all in their characterization, on the other hand, then mechanisms can no longer be useful for grounding prediction and supporting intervention strategies. I conclude that, if the New Mechanistic Philosophy is to be successful, a stochastic characterization of mechanisms must be adopted.

Philosophers of science have insisted that evidence of underlying mechanisms is required to support claims about the effects of medical interventions. Yet evidence about mechanisms does not feature on dominant evidence-based medicine “hierarchies.” After arguing that only inferences from mechanisms (“mechanistic reasoning”)—not mechanisms themselves—count as evidence, I argue for a middle ground. Mechanistic reasoning is not required to establish causation when we have high-quality controlled studies; moreover, mechanistic reasoning is more problematic than has been assumed. Yet where the problems can be overcome, mechanistic reasoning can and should be used as evidence.

Machamer, Darden, and Craver's account of the nature and role of mechanisms in the special sciences has been very influential. Unfortunately, a confusing array of ontic, epistemic, and pragmatic distinctions is required to individuate their mechanisms, mechanism schemata, and mechanism sketches. I diagnose this as a conflation of token-level causal relations with type-level relations. I propose instead that a mechanism is an abstraction that relates entity types and activity types on the model of a directed graph. Mechanisms have an ontic status distinct from the causal chains of token entities and token activities that instantiate them.