Facts and theories on the evolution of vocal learning in mammals and birds are reviewed. An attempt is made to articulate principles of general heuristic importance. Different contexts in which vocal learning occurs are evaluated. It is concluded that maximal stimulation of females and benefits accruing from formation of vocal dialects have been selective pressures leading to vocal learning. Correlations between intense speciation and vocal learning, and between long-lasting pair bonds and filial song learning, need further study. In all cases where selective pressures leading to vocal learning have been identified, females are the selective agent. Use of vocal signals to attract females is much rarer in mammals, and this is correlated with a low incidence of vocal learning. The successive steps leading from genetically determined motor ontogenies to vocal learning are discussed.