The potential energy surfaces for the reactions of silyl radical (SiH3)-Si-. with ethylene and propylene are studied using both the spin-projected MP2 method (PMP2) with the 6-31G* basis set and the QCISD(T) method with the more accurate 6-311G** basis. For both reactions it is found that the channels leading to the olefin double-bond addition are highly favored with respect to the reaction pathways associated with hydrogen abstraction. These results agree with recent experiments and rule out the hypothesis that hydrogen abstraction and not double addition can be the primary process in the case of propylene. However, in the comparison between ethylene and propylene, it is found that the activation energy for the addition is slightly lower in the latter case than in the former suggesting that in gas phase alkyl substitution activates olefins toward addition by (SiH3)-Si-.. The good agreement between the PMP2 and the QCISD(T) results and the experimental results (activation energies and reaction enthalpies) indicates that a PMP2 approach with a basis set of double-zeta quality plus polarization functions (6-31G*) can provide a reliable description for this class of reactions. A simple diabatic model is used to rationalise these computational results. This model indicates the pi --> pi* triplet excitation energy as the key factor which determines the trend of the gas-phase activation barrier on passing from ethylene to propylene.