Abstract
August 16, 1997 David Lewis2 has long defended an account of scientific law acceptable even to an empiricist with significant metaphysical scruples. On this account, the laws are defined to be the consequences of the best system for axiomitizing all occurrent fact. Here "best system" means the set of sentences which yields the best combination of strength of descriptive content 3 with simplicity of exposition. And occurrent facts, the facts to be systematized, are roughly the particular facts about a localized space-time region that are non-modal, non-dispositional, and non-causal. Scientists providing or attempting to provide laws are plausibly seen as giving general principles that unify a body of data. Thus they organize or systematize the arrangement of occurrences. For this reason, Lewis's account has the important merits of providing contact with actual scientific practice while making sense of the standard philosophical conception that laws should be general but more than mere accidental generalizations. However, Lewis has long known about a potential problem with this account, a problem involving chance and credence.4 In a recent series of articles he, Michael Thau, and Ned Hall have developed a new formulation of the relationship between chance and credence which solves the problem. However, I will argue that these articles leave untouched and even exacerbate a closely related and more fundamental problem with the best system account, the problem of nomic necessity. Laws are supposed to be more than true; in some sense they must be true. Yet a principle's membership in the best systematization for one world seems to say nothing about its necessity, i.e., its truth at other worlds. I close by briefly describing how an alternative empiricist account may remove both problems