Hidden variable models for quantum mechanics can have local parts

We criticize Colbeck and Renner's (CR's) statement that "any hidden variable model can only be compatible with quantum mechanics if its local part is trivial" [Phys. Rev. Lett. 101, 050403 (2008)]. We note that CR's attempt to divide a nonlocal hidden variable model into a "local part" and a "nonlocal part" contains a restriction on the latter. This restriction implies that the division is really into a "local part" and a "nonsignaling nonlocal part." CR's nonsignaling requirement on the "nonlocal part" cannot be physically motivated, since the hidden variables cannot be accessed by experimenters. Nor is it a natural mathematical generalization from the local hidden variable case, since it is simple to make a generalization without CR's requirement. We give an explicit nonlocal hidden variable model that, in the case the restriction is not enforced, contains nontrivial local hidden variables.