Abstract

Pierre Duhem's well-known argument for the underdetermination of theory choice by evidence is often cited in current discussion of scientific realism and antirealism. I argue that the familiar account of it that participants draw on is incomplete. After studying Duhem's philosophy of scientific language, which he regards as both abstract and symbolic, I conclude that, for Duhem, underdetermination rests on the observation that instruments are ubiquitous in mathematical sciences. This ties auxiliary assumptions to the use of instruments and completes the familiar account of underdetermination. Even when complete, however, this account has no obvious implications for the debates about realism, nor does it settle how Duhem's own philosophy should be interpreted. An interpretive problem arises because there is textual support for both antirealist and realist readings of his work. Accordingly, in the recent literature he has been interpreted both as a fictionalist and as a structural realist. I argue that neither of these readings will do. First, the reasoning that might motivate a fictionalist interpretation assumes a certain semantical view that Duhem would not accept. To develop his pragmatic alternative, I address his conception of theory-ladenness and his account of how scientists come to understand phenomena. Second, the realist conviction that mitigates his apparent instrumentalism---specifically, the belief that physical theory is approaching a logically unified and natural classification---does not implicate him as a structural realist. For Duhem, natural classifications are ontologically significant and preserved over conceptual revolutions because they pick out essential features of real things and order phenomena accordingly. To properly classify Duhem's philosophy, I study his realist and antirealist claims and point out that they are different. In his view, scientific reasoning leads straight to antirealism but feelings, hunches, and intuitions motivate, without justifying, a kind of realism. I argue that Duhem's approach is characteristic of motivational realism