The rational design of proteins requires knowledge of the helix-forming propensities (s-values) of the different amino acids. There is, however, considerable controversy about the relative values for alanine and glycine. We find from experiments on mutants of barnase that the relative effect of Ala versus Gly on helix stability depends crucially on the position in the helix (whether they are at the ends (caps) or are internal) and the context (the influence of their neighbours). Glycine is greatly preferred at the N and C caps. At internal positions, Ala stabilizes the helix relative to Gly by 0.4 to 2 kcal mol-1. The variation results from a combination of burial of hydrophobic surface on folding and interference with hydrogen bonding of the protein with solvent. There is a good empirical correlation between the relative stabilizing effects of Ala and Gly with the total change in solvent-accessible hydrophobic surface area of the folded protein on mutation of Gly to Ala. It is not valid to assign to each amino acid a unique s-value that is generally applicable to all positions in all helices in all proteins.