The six selection experiments reported all gave positive response. Rainbow trout is the fish species with most estimates of genetic and phenotypic parameters. Heritabilities for economically important traits are of low to moderate size but with large phenotypic variance, particularly for body size and age at maturation. The effect of inbreeding is negative for the traits studied and should, therefore, be avoided in practical breeding. Crossbreeding between strains and inbred lines has given varying results and does not, in general, offer advantages as a breeding method for rainbow trout. Estimates of genotype-environment interaction vary from negligible to considerable effect. It is, therefore, necessary to study the magnitude of genotype-environment interactions under the environment where the breeding program is going to be used. The most important traits to be improved are: growth rate, feed efficiency, age at sexual maturation (when producing large fish), disease resistance and meat quality. In rainbow trout, individual and family selection are the most promising selection methods. Since additive genetic variance seems to be present for all economically important traits, selection should always be used in breeding programs, and crossbreeding should be included only when heterosis is large. Since fecundity is high in rainbow trout the breeding program can be concentrated at one or a few breeding centers. The importance of evaluating breeding values of fish under practical farming conditions (test stations) is stressed. Multiplier stations should be developed according to needs. It is concluded that selection is a powerful tool to increase productivity in rainbow trout and that it will increase the rate of domestication.