Theory Change, Structural Realism, and the Relativised A Priori

Abstract

In this paper I claim that Quinean naturalist accounts of science, that deny that there are any a priori statements in scientific frameworks, cannot account for the foundational role of certain classes of statements in scientific practice. In this I follow Michael Friedman who claims that certain a priori statements must be presupposed in order to formulate empirical hypotheses. I also show that Friedman’s account, in spite of his claims to the contrary, is compatible with a type of non‐Quinean naturalism that I sketch. Finally I also show that Friedman’s account needs amending because it cannot provide a rational account of theory change. I accomplish this by arguing for a structural realist view of theory change. I show how this view fits well with an account like Friedman’s and helps it deal with the problem of theory change and in retaining its superiority over Quinean naturalism.

Acknowledgements

I would like to thank Corey Mulvihill, Jon Tsou and Idil Boran for their useful comments on early drafts. I would also like to thank the philosophy faculty of Carleton University for their comments and for inviting me to present an early draft as part of their 2005 colloquia series. I also thank the attendees of the 2005 Canadian Philosophical Association Congress, where I also presented a draft. I would also like to thank the editor and two anonymous referees for their very extensive comments on earlier drafts.

Notes

[1] Although a review of the current literature on SR is beyond the scope of this paper, some further comment is warranted on the version I follow. Worrall’s (1989) view is that science gives us knowledge of structure and not nature. This has been subject to much criticism because, according to Psillos (1995, 2001) and others, it includes a difficult to sustain distinction between structure and nature. Worrall’s ‘nature’, according to Psillos, becomes undefinable, and this causes Psillos to doubt the plausibility of SR. Chakravartty (1998) accepts this criticism and avoids the difficulty by identifying objects and their nature in terms of their properties that are used in laboratory interactions. Ladyman (1998) on the other hand proposes a much stronger version of SR and asserts that there are no objects at all, only structure. However, this view is highly problematic. As Psillos (2001) and I also (McArthur 2006) note, it tries to maintain a notion of isomorphism without the idea of paired objects. Cao (2003) has pointed out that by defining everything in terms of structure it can stifle the quest for deeper structures and causes. Furthermore, I have argued that since it makes all relations internal, it suffers from regress problems (McArthur 2006). Given all this, I have chosen a version of SR that seems to lack the problems of Ladyman’s view and also avoids the difficulties Psillos has noted with Worrall’s position. SR is controversial, of course, as Psillos’s writings attest, but it is entertained in a sufficiently wide way to be a candidate for my purposes, and I think the version I have chosen to follow avoids most of the criticism current in the SR literature.