I compare two kinds of holism about values: G.E. Moore's 'organic unities', and Jonathan Dancy's 'value holism'. I propose a simple formal model for representing evaluations of parts and wholes. I then define two conditions, additivism and invariabilism, which together imply a third, atomism. Since atomism is absurd, we must reject one of the former two conditions. This is where Moore and Dancy part company: whereas Moore rejects additivism, Dancy rejects invariabilism. I argue that Moore's view is more plausible. Invariabilism (...) ought to be retained because (a) it eliminates the needless multiplication of values inherent in variable evaluations, and (b) it preserves a certain necessary connection between values and reasons, which Dancy himself endorses. (shrink)