A widely cited model of the evolution of functionally novel proteins (here called the model of mutation during non-functionality (MDN model)) holds that, after gene duplication, one gene copy is redundant and free to accumulate substitutions at random. By chance, some of these substitutions may suit the protein encoded by such a non-functional gene to a new function, which it can subsequently assume. Several lines of evidence contradict this hypothesis: (i) comparison of expressed duplicate genes from the tetraploid frog Xenopus laevis suggests that such genes are subject to purifying selection and are thus not free to accumulate substitutions at random; (ii) in a number of multi-gene families, there is now evidence that functionally distinct proteins have arisen not as a result of chance fixation of neutral variants but rather as a result of positive Darwinian selection; and (iii) the phenomenon of gene sharing, in which a single gene encodes a protein having two distinct functions, shows that gene duplication is not a necessary prerequisite to the evolution of a new protein function. A model for the evolution of new protein is proposed under which a period of gene sharing ordinarily precedes the evolution of functionally distinct proteins. Gene duplication then allows each daughter gene to specialize for one of the functions of the ancestral gene. However, if the ancestral gene is not bifunctional, either of the following two outcomes is expected to follow gene duplication: (i) one copy will be silenced by a mutation preventing expression; or (ii) if both copies continue to be expressed, both will be subject to purifying selection, as a high proportion of nonsynonymous mutations will have a completely or partly dominant deleterious effect.