Previously calculated conformational energy maps suggest that the alpha-helical conformation for the residue preceding a proline is disfavored relative to the extended conformation by more than 7 kcal/mol. In known protein structures this conformation is observed, however, to occur for about 9% of all prolines. In addition, introduction or removal of prolines at theoretically unfavorable positions in proteins and peptides can have modest effects on stability and structure. To investigate the discrepancy between calculation and experiment, we have determined how the conformation of the proline affects the calculated energy. We have also explored the effect of bond length and bond angle relaxation on the conformational energy map. The conformational energy of the preceding residue is found to be unaffected by the conformation of the proline, but the effect of allowing covalent bond relaxation is dramatic. If bond lengths and angles, and dihedral angles within the pyrrolidine ring, are allowed to relax, a calculated energy difference between the alpha and beta conformations of 1.1 kcal/mol is obtained, in reasonable agreement with experiment. The detailed shape of the calculated energy surface is also in excellent agreement with the observed conformational distributions in known protein structures.