ABSTRACT
The question of how to deal with auxiliary hypotheses in confi rmation theory is still live, and has recently been discussed by Strevens (2001, 2005) and Fitelson and Waterman (2005, 2007).1 In essence, the core problem is as follows. How do we measure the (dis)confi rmation of a theory, as distinct from the (dis)confi rmation of the theory plus the auxiliary hypotheses (and/or other data) used to render it predictive?
