Side-chain 2H and backbone 15N relaxation data have been collected at multiple temperatures in the samples of the SH3 domain from alpha-spectrin. Combined analyses of the data allowed for determination of the temperature-dependent correlation times tauf characterizing fast methyl motion. Molecular dynamics simulations confirmed that tauf are dominated by methyl rotation; the corresponding activation energies approximate methyl rotation barriers. For 33 methyl groups in the alpha-spectrin SH3 domain the average barrier height was thus determined to be 2.8 +/- 0.9 kcal/mol. This value is deemed representative of the "fluid" hydrophobic protein core where some barriers are increased and others are lowered because of the contacts with surrounding atoms, but there is no local order that could produce systematically higher (lower) barriers. For comparison, the MD simulation predicts the average barrier of 3.1 kcal/mol (calculated via the potential of mean force) or 3.4-3.5 kcal/mol (rigid barriers after appropriate averaging over multiple MD snapshots). The latter result prompted us to investigate rigid methyl rotation barriers in a series of NMR structures from the Protein Databank. In most cases the barriers proved to be higher than expected, 4-6 kcal/mol. To a certain degree, this is caused by tight packing of the side chains in the NMR structures and stems from the structure calculation procedure where the coordinates are first annealed toward the temperature of 0 K and then subjected to energy minimization. In several cases the barriers >10 kcal/mol are indicative of van der Waals violations. The notable exceptions are (i) the structures solved using the GROMOS force field where tight methyl packing is avoided (3.0-3.6 kcal/mol) and (ii) the structure solved by means of the dynamic ensemble refinement method (Lindorff-Larsen, K.; Best, R. B.; DePristo, M. A.; Dobson, C. M.; Vendruscolo, M. Nature 2005, 433, 128) (3.5 kcal/mol). These results demonstrate that methyl rotation barriers, derived from the experiments that are traditionally associated with studies of protein dynamics, can be also used in the context of structural work. This is particularly interesting in view of the recent efforts to incorporate dynamics data in the process of protein structure elucidation.