Three accounts relating meaning and context are compared: a classical or static one as proposed by Stalnaker, a contextual or dynamic one as proposed in Dynamic Semantics, and a massively contextual one, defended here. On the last view, meaning and interpretation is a matter of a change of an epistemic context by means of an inductive inference, thus of pragmatics. As in dynamic semantics, meaning is a matter of epistemic state change. But now it is construed normatively, and it is the contextual change that explicates meaning, not meaning that explicates why a context changes in the way it does. Meaning is contextual on this approach because the justification of inductive inferences depends on contextual parameters (such as a partition of answers, or a degree of caution with respect to the risk of incurring error, etc.) for whose assessment no objective standards can be given. Contextuality is not a feature of language per se, and questions of contextual change are not primarily linguistic ones.