A long-standing question in evolutionary biology is the relative contribution of large and small effect mutations to the adaptive process. We have investigated this question in proteins by estimating the rate of adaptive evolution between all pairs of amino acids separated by one mutational step using a McDonald-Kreitman type approach and genome-wide data from several Drosophila species. We find that the rate of adaptive evolution is highest amongst amino acids that are more similar. This is partly due to the fact that the proportion of mutations that are adaptive is higher amongst more similar amino acids. We also find that the rate of neutral evolution between amino acids is higher amongst more similar amino acids. Overall our results suggest that both the adaptive and non-adaptive evolution of proteins is dominated by substitutions between similar amino acids.