Natural selection is a composite of forces acting on the gene pool of a population. Mather (1953) has pointed out two components of selection which are, in a sense, opposed to one another. These are selection for stability or immediate fitness and selection for the maintenance of variability or flexibility by which trend changes can be brought about. In naturally outbreeding populations selection for the maintenance of stability should be directed toward the maintenance of developmental homeostasis. It is this type of stabilizing selection which has been emphasized by Schmalhausen (1949), Waddington (1957), Lerner (1958), and others. But the natural environment is not constant. Short-term fluctuations and longterm time trends occur, exerting a directional selection pressure on the gene pool (Mather, loc. cit.). How, then, and within the limits of effective population size, may both stabilizing and directional selection take place in the natural environment with no reduction in population fitness and how may such changes be detected? Mean viability of the random heterozygotes is directly associated with population fitness (Band and Ives, 1963a; Dobzhansky and Spassky, 1963; Dobzhansky, 1964). It is dependent on the underlying genotypic constitution of the population and should be subject to the effects of natural selection at the genotypic level. Mean viability remains relatively constant when the same random array of random heterozygotes is tested in different laboratory temperatures (Spassky et al., 1960; Band and Ives, 1963a) and similar when random heterozygotes from different populations are tested in the same laboratory environment (Dobzhansky and Spassky, 1963). In turn, mean viability is determined by the contribution of various heterozygous classes to this fitness component (Wallace, 1962; Band and Ives, 1963a; Dobzhansky and Spassky, 1963). Experiments of Wallace (1955) and Breese and Mather (1960) also indicate that viability is not the maximum attainable. Consequently mean viability of the random heterozygotes becomes a better measure of the effectiveness of natural selection to maintain population fitness under changing environmental conditions than conventional metric traits. Variance components of viability for the random heterozygotes are also determined from the relative contributions of different heterozygous classes (Band and Ives, 1963a). If mean viability among the random heterozygotes remains relatively constant, then the maintenance of population fitness for viability may be the outcome of the joint effects of directional selection at the genotypic level and selection for the maintenance of developmental homeostasis, leading to the retention of genetic diversity in the population. Experiments on heterozygotes in different fluctuating environments (Band, 1963c) suggest that this may be the case and that directional selection pressure exerted by the environment may be minimized or impeded by selection for the maintenance of developmental homeostasis among the random heterozygotes. The retention of a coadapted gene pool may thus arise as an interaction between these two components of selection. By choice of appropriate statistical techniques, it should be possible to determine if one or both components of 1 The major portion of this investigation was done in the Department of Zoology, University of British Columbia. The research was supported by grants from the National Science Foundation and National Research Council of Canada.