Sympathetic commentators on Kant’s account of physical knowledge agree that while philosophy of science has much to gain from Kant’s notion of constitutive a priori principles, Kant’s conceptual and methodological resources are inconsistent with the possibility of scientific change. In this article, I argue that this received view is lacking since Kant’s claim that a unique set of a priori principles structures our knowledge for all time is not central to his account of the constitutive a priori. Two underemphasized points (...) of Kant’s theory bear this out. First, Kant applies ‘a priori’ widely to include non-truth-evaluable elements of knowledge. As such, Kant primarily understands the necessity that attaches to these elements in light of the constitutive role they play in knowledge rather than in terms of truth for all time. Second, Kant uses two methods to establish the existence of constitutive a priori principles: the analytic and the synthetic. On my interpretation, while scientific change has discredited the analytic method, the synthetic method remains viable. In this way, I offer a new perspective on the ways in which Kant’s theory of the constitutive a priori can ground Neo-Kantian philosophies of science like those of Hans Reichenbach and Michael Friedman. (shrink)