We describe a quantitative angular momentum (AM) model for predicting rotational transfer (RT) and vibrotational transfer (VRT) in collisions between CO2 and hot H atoms. This molecule is important in several contexts, not least as a bridge between the relative simplicity of diatomic molecules and the complexities of polyatomic RT and VRT. We show that for pure RT, an AM constraint dominates but that this changes to a dominant energetic constraint in the case of VRT. The requirement that the (001) vibrational channel be opened simultaneously with the generation of AM imposes special restrictions which effectively limit the trajectories that lead to VRT. The origin of this is a constraint-induced restriction on the effective impact parameter (b(n)(max)) for individual Delta j channels and the effect is manifest as reduced probability for populating low Delta j channels. In CO2-H* this leads to a shift in the peak of (VRT) Delta j probabilities away from zero as found experimentally for the (001) vibrational mode. We report a Monte Carlo trajectory calculation similar to that of Kreutz and Flynn [J. Chem. Phys. 93, 452 (1990)] but predict an exponential-like dependence of pure RT on Delta j. For VRT to (001) the constraint-induced restrictions on b(n)(max) are incorporated quantitatively and the vibrational channel-opening velocity is treated as a vector quantity. The results of these calculations are in good agreement with experiment. The underlying mechanism, likely to be general in VRT, is clearly revealed in plots of relative velocity versus rotational AM change.