It is an old question why in quite a few Indo-european languages adpositions govern a dative case to yield a locative meaning and an accusative case to force a directional interpretation. While I have no real answer to this question, I would like to present some ideas as to how we might approach this question. The core of the idea is that the dative case is not intrinsically linked to locative meaning but that it is the default case in oblique domains, including adpositional phrases, at least in those languages that have a case system rich enough to distinguish oblique cases from others. The accusative case, on the other hand, is closely associated with measurement. And this, I would like to argue, is why at least certain directional adpositional constructions involve the accusative case.