D ominance is the masking at the level o f the phenotype o f the presence o f one o f the two alleles at a diploid locus. We discuss whether it could be an automatic conse=quence o f cellular physiology, as proposed initially by Wright and elaborated by Kacser and Burns, or whether it might be a consequence o f natural selection having favoured it in the past, as argued by Fisher. With some exceptions, such as “supply-driven” processes, the physi= ological explanation is generally valid. At the same time, the extensive pleiotropy o f gene action implies that the functioning o f a locus can be influenced by many other loci. This means that selection can act on a secondary locus and affect the relationship between alleles at the locus o f primary interest. Therefore in principle dominance could have evolved in the manner sug= gested by Fisher; but when considered in detail, the difficulties with his model imply that the physiological explanation is to be preferred at present.