Addition of tert-butyl and benzyl radicals to ethene has been studied by semiempirical and ab-initio quantum mechanical methods. The semeempirical approaches fail to reproduce even the trend of experimental reaction rates. At the ab-initio level, unrestricted Hartree-Fock computations with split valence basis sets give results in fair agreement with experiment. Extension of the basis set has little effect, whereas correlation energy plays a nonnegligible role. Furthermore, full spin projection of unrestricted wave functions is mandatory. Non potential energy effects have also been considered, but their role is marginal. The results can be rationalized by the state correlation diagram approach.