Protein evolutionary rates have been presumed to be mostly determined by the density of functionally important amino acids in a given protein. They have been shown to correlate with variables intuitively related to functional importance of proteins, such as protein dispensability and protein-protein interactions. Surprisingly, the best correlate of the evolutionary rates has turned out to be not the functional importance of a protein, but the expression level of the protein. Drummond and Wilke suggest that the dominant role of expression levels in slowing the rate of protein evolution stems from a selection pressure against mistranslation-induced protein misfolding. We will review current evidence for and against different hypotheses on determining evolutionary rates.