An explicit treatment of electron correlation is required to predict accurate energetics, barrier heights, and saddle point geometries for chemical reactions. Several theoretical methods for treating electron correlation (multireference configuration interaction, perturbation theory, and coupled cluster methods) have been thoroughly evaluated for the F(P-2) + H-2(X-1 Sigma(g)(+)) and O(P-3) + H-2(X-1 Sigma(g)(+)) abstraction reactions as well as for the H'(2S) + HCl(X-1 Sigma(+)) exchange reaction using correlation consistent basis sets. The basis set dependence of the reaction energy defects, barrier heights, and saddle point geometries have been determined for each theoretical method. Addition of diffuse functions to the basis set (aug-cc-pVnZ) was found to substantially increase the convergence rate. Calculations with the largest basis set (aug-cc-pV5Z) allowed an unambiguous comparison of the relative performance of each correlation method. For each reaction, the R-UCCSD(T) results closely parallel the most accurate MRCI results and are in good agreement with experiment. In contrast, unrestricted perturbation theory methods predict barriers that are too large by 2.7-4.4 kcal/mol (MP2), 3.5-4.2 kcal/mol (MP3), and 1.3-1.7 kcal/mol (MP4).