We prove that, for a Lipschitz function on R-n, n [.ge.] 2, the approximate and the Clarke subdifferentials can differ everywhere. This completely answers a question by A.D. loffe, which was partially answered by G.
Katriel.