Progeny testing and selection of forage grasses by means of growing half-sib (HS) families from openpollination and polycross have been considered from theoretical and practical points of view. Special attention has been paid to the genetic variation within half-sib families, which is expected to be large as compared to the genetic variation between families. Based on observations of individual plants within plots, the environmental component of the variation is expected to be large and nonestimatable. The results of an experiment in meadow fescue (Festuca pratensis Huds.) are presented. In this experiment, randomly selected individual plants within HS families were cloned and laid out in randomized blocks. For the characters observed (earliness and raw matter yield) no significant variance component for dominance was found. The highly significant additive component estimated for earliness, as well as for yield, after each of three cuts and in total were about three times as large within as between families, as expected from the theoretical considerations. The estimated response to selection was much higher for a combination of between- and within-family selection as compared to free clone or family mean selection alone. It is suggested that a program for progeny testing and selection in a base population of perennial forage grasses should start with an experiment in which a large number of randomly selected parental clones and a fixed number of clones from each of the half-sib families derived from the mother genotypes are grown simultaneously. The selected clones within superior families could later on be further cloned, placed in a polycross field, and the new HS-families could be sown in ordinary field trials at various locations for further selection.