A methodological critique of the semantic conception of theories

Abstract

A new PhD slated to teach a beginning undergraduate course on scientific reasoning recently asked me to recommend topics. I launched into a description of my baby- Popper-plus-statistics class-give them enough deductive logic to understand the Duhemian problem, do the Galileo case study, use the notion of severe test to introduce a bit of probability theory, then segue to the problem of testing statistical hypotheses. My interlocutor was looking impatient. But Im a strong adherent of the Semantic Conception of theories, he said. I cant teach all that stuff about trying to falsify bold conjectures. This was not a moment for proselytizing, so I loaned him a copy of Gieres textbook, which is based on the Semantic Conception, and sent him happily on his way. However, this episode raises an interesting question, one that takes on some urgency as the Semantic Conception of scientific theories (SC) seems well on its way to becoming the new received view: What accounts of scientific method, confirmation and explanation does the SC support? A major motivation of the Semantic Conception for philosophers was to replace the awkward syntactic account of theories proposed by the positivists, who favoured a formal axiomatic system accompanied by correspondence rules or meaning postulates. But since Popper never gave an account of meaning and was never worried about the problem of how to interpret theoretical terms, it might seem that there should be no inherent tension between Popperian methodology and an account of science that views theories as sets of models. Because the Semantic Conception liberates the content of a theory from any particular linguistic formulation, this move might appear congenial to those in agreement with the Popperian dictum, Words dont matter. Furthermore, the Semantic Conceptions emphasis on mapping structures that reside in the world also seems to mesh with Poppers anti-essentialism. Yet I will argue that the overall approach to scientific inquiry that accompanies the SC approach is antithetical to a Popperian account of scientific methodology, which is intended to maximize the role of criticism. Moreover the methodological glosses that commonly accompany expositions of the Semantic Conception are either antithetical to commonly accepted norms of scientific inquiry or hopelessly ad hoc. The points I wish to make are concerned neither with the technical details of the SC nor the more idiosyncratic aspects of Popperian methodology, such as his views on induction. Furthermore, for the purposes of this paper I will not worry about whether we should view our best scientific theories as being true, approximately true or merely empirically adequate. My intention is rather to dramatize the methodological differences between an approach to science that focuses on model building and one that construes scientific inquiry as a search for true or approximately true or empirically adequate generalisations. Most of my claims are straightforward and do not hinge on the particular variant of the Semantic Conception one adopts as long as that conception views scientific theories as a class of models of relational structures, not as universal generalisations. Glymour gave this characterization of the SC: (T)he product of science in this view is not so much knowledge of general propositions as an understanding of systems of models and how to embed various classes of phenomena within these models. (Salmon, p. 122) Thus for my purposes, van Fraassen is not a target simply because his SC account includes generalisations as part of the theory. He says a theory is empirically adequate exactly when all appearances are isomorphic to empirical substructures in at least one of its models (Boyd, p. 192). Neither am I concerned here about the SC approach to a theory such as cosmology which has only one intended application. The da Costa and French book arrived too late for me to evaluate it carefully, but it may also evade the criticisms that follow. As these exceptions make clear, my quarrel here is less with the SC per se than with the methodology associated with it especially by writers who take the disunity of science to be a virtue. Ron Giere is the SC theorist who has paid most attention to methodological issues. Therefore I will generally illustrate my argument using examples from his writings, which have the additional virtues of being clear and extensive. To my mind, his is an especially interesting version of the Semantic Conception, in part because he takes seriously the goal of accounting for our best examples of scientific practice. Nevertheless, I find that it does not hang together as well as a methodology that construes theories as statements. © 2006 Springer.