To understand the behavior of a complex system, one must understand the interactions among its parts. Doing so is difficult for nondecomposable systems, in which the interactions strongly influence the short term behavior of the parts. Science's principal tool for dealing with nondecomposable systems is a variety of probabilistic analysis that I call EPA. I show that EPA's power derives from an assumption that appears to be false of nondecomposable complex systems, in virtue of their very nondecomposability. Yet EPA is extremely successful. I aim to find an interpretation of EPA's assumption that is consistent with, indeed that explains, its success.

A number of arguments have been given to show that the modal interpretation of ordinary nonrelativistic quantum mechanics cannot be consistently extended to the relativistic setting. We find these arguments inconclusive. However, there is a prima facie reason to think that a tension exists between the modal interpretation and relativistic invariance; namely, the best candidate for a modal interpretation adapted to relativistic quantum field theory, a prescription due to Rob Clifton (2000), comes out trivial when applied to a number of systems of physical interest. However, it is far from clear whether this difficulty for the modal interpretation is traceable to relativistic invariance per se or to the infinite number of degrees of freedom involved. In any case, the proponents of the modal interpretation have work to do.

Probability theory is important not least because of its relevance for decision making, which also means: its relevance for the single case. The frequency theory of probability on its own is irrelevant in the single case. However, Howson and Urbach argue that Bayesianism can solve the frequentist's problem: frequentist-probability information is relevant to Bayesians (although to nobody else). The present paper shows that Howson and Urbach's solution cannot work, and indeed that no Bayesian solution can work. There is no way to make frequentist probability relevant.

In three recent papers, Wayne Myrvold (1996, 2003) and Timothy McGrew (2003) have developed Bayesian accounts of the virtue of unification. In his account, McGrew demonstrates that, ceteris paribus, a hypothesis that unifies its evidence will have a higher posterior probability than a hypothesis that does not. Myrvold, on the other hand, offers a specific measure of unification that can be applied to individual hypotheses. He argues that one must account for this measure in order to calculate correctly the degree of confirmation that a hypothesis receives from its evidence. Using the probability calculus, I prove that the two accounts of unification require the same underlying inequality; thus, McGrew and Myrvold have accounted for unification in fundamentally identical probabilistic terms. I then evaluate five putative counterexamples to this account and show that these examples, far from disqualifying it, serve to clarify our notion of unification by disentangling it from a host of other concepts.

The famous ethological maxim known as ‘Morgan's Canon’ continues to be an object of interpretive controversy. I reconsider Morgan's Canon in light of two questions: First, what did Morgan intend? Second, is this, or perhaps some reinterpretation of the Canon, useful within cognitive ethology? As for the first issue, Morgan's distinction between higher and lower faculties is suggestive of an early supervenience concept. As for the second issue, the Canon, both in its original form and in various recent rereadings, offers nothing useful to cognitive ethology.