Abstract
It is shown that Bell's proof of violation of local realism in phase space is incorrect. Using Bell's approach, a violation can be derived also for nonnegative Wigner distributions. The error is found to lie in the use of an unnormalizable Wigner function.