We have studied the proline-directed, Pd-catalyzed enantioselective hydrogenation of isophorone in the liquid state using a variety of methods. Our results unambiguously reveal the true reaction pathway and demonstrate that all earlier mechanistic hypotheses are wrong: although a proline/isophorone condensation product is formed, it is merely a spectator and not a key reaction intermediate in subsequent heterogeneous hydrogenation. Enantioselectivity is the result of kinetic resolution-a process that occurs homogeneously in solution and not at the metal surface. Racemic 3,3,5-trimethylcyclohexanone (TMCH) is produced by initial heterogeneous hydrogenation of isophorone; proline then reacts homogeneously, preferentially with one enantiomer of TMCH, leaving an excess of the other. Thus in complete contrast to the case of ketoester asymmetric hydrogenation, the metal surface is not involved in the crucial enantio-differentiation step. The mechanism we propose also explains why the maximum attainable yield of enantiopure TMCH cannot exceed 50%.