Do we need norms in a naturalised epistemology?

Epistemic rationality (or science) is traditionally associated with the following of epistemic norms. Though it is common to eschew the idea that such norms can be specified as part of an priori foundation for science, many supporters of so-called naturalised epistemology retain the goal of finding genuine epistemic norms. I argue that, for two significant subvarieties of naturalized epistemology, we cannot make sense of the norms they propose as essential to the scientific enterprise. The two subvarieties fall within the category of what I call anti-foundationalism since they deny that epistemic norms can be seen as any kind of foundation for science; instead they develop and change with scientific knowledge. The first derives from John Stuart Mill. He argues that science starts with an assumption that straight enumerative induction is a reliable inference rule, and gradually produces more and more refined inductive canons in relation to the observational success of different kinds of hypotheses. The problem for this account is this: If these canons are going to count as genuine norms, one must assume that the process of refining hypotheses in relation to observation in general constitutes a rational method for gaining ever truer beliefs. But if this is assumed, why should norms be of use in guiding the formation of scientific hypotheses in addition to observation? The other subvariety of anti-foundationalism is often termed instrumentalism; its most developed form is in Larry Laudan's normative naturalism. The fundamental idea is that the norms for science take the form of hypothetical imperatives: 'If your cognitive aim is A, employ method M', which are justified in the same way ordinary empirical statements are. I argue that hypothetical imperatives are (i) not needed in science and