Donald Davidson suggested that, in attempting to give meaning theories, we should proceed via giving truth theories. For the programme of truth-theoretic semantics to be successful, two tasks need to be accomplished. First, it has to be shown that natural languages are actually amendable to truth theoretic treatment. The second task is to show how we can bridge the gap between a truth theory and a genuine meaning theory. This second task is necessitated by the simple fact that truth theories by themselves are too weak to satisfy the central desideratum for a meaning theory. In a recent paper, Greg Ray suggests that the goal of giving meaning theories can easily be achieved, not by supplementing the truth-theoretic apparatus, but by doing away with it altogether (Ray 2014). In particular, he proposes that what he calls means-that theories meet all the desiderata on meaning theories, and that 'anyone with basically Davidsonian commitments must accept the means-that approach as viable and that it has certain evident virtues'. This paper argues that theories along Ray's lines do not satisfy the central desideratum for meaning theories. Ray's belief to the contrary rests on a common misunderstanding of what this desideratum amounts to. As I indicate in the final section of the paper, there perhaps is a way of substantially amending Ray's approach in order to meet the central desideratum. However, in an ironic twist, this will require giving a full-fledged truth theory for the metalanguage, including its intensional constructions like 'means that' - precisely the project Ray attempted to avoid.
