The model proposed by Cossee for Ziegler-Natta polymerization of propene is re-examined by computing the non-bonded energies for all sets of possible internal coordinates of the atoms at the catalytic site. For catalytic sites in the bulk of (110) surfaces of TiCl3-α, it is possible to explain the regiospecificity of the reaction but, at variance with Cossee, we see no evidence that the surface conditions the methyl group of the olefin to protrude out of the crystal. The stereospecificity may arise however from the fixed, chiral orientation of the first carbon-carbon chain bond, due to interactions with the surface. Further computations seem to indicate that the polymerization could occur with much less steric hindrances on edges or reliefs of the surfaces, the relative positions of olefin and chain being exchanged in respect to the Cossee model. © 1979.