It is well-known that the ability of the Vickrey-Clarke-Groves (VCG) mechanism to implement efficient outcomes for private value choice problems does not extend to interdependent value problems. When an agent's type affects other agents' utilities, it may not be incentive compatible for him to truthfully reveal his type when faced with VCG payments. We show that when agents are informationally small, there exist small modifications to the VCG transfers that restore incentive compatibility. We further show that truthful revelation is an approximate ex post equilibrium. Lastly, we show that in replicated settings aggregate payments sufficient to induce truthful revelation go to zero.