The thesis defends orthodoxy about knowledge by arguing against shifty views of knowledge (or 'know'). According to shifty views of this sort, pragmatic factors, such as stakes or the cost of error, affect whether one can truly ascribe 'knowledge' to someone. The thesis distinguishes knowledge and certainty, and it argues in favour of (context-sensitive) certainty norms for assertion and action: a subject S can appropriately assert a proposition p and act on this proposition p in context C only if p is certain for S in context C. These norms are established on the basis of linguistic and conversational data, but also on the basis of more principled considerations. The thesis shows that if we embrace such norms, the main arguments in favour of shifty views of knowledge (or 'know') can be dismissed. New criticisms of rival orthodox approaches are also provided.