We study the question of which social choice functions from an abstract type space to a set of outcomes are truthful, i.e., implementable by truthful mechanisms, when utilities are quasi-linear. For convex domains, our main theorem characterizes truthful social choice functions as those...