Laboratory cultures of offspring from single families and a mass mating of Mytilus eduIis were reared to the spat stage at different temperatures. Some cultures were reared further to the juvenile stage at the same, or at an altered temperature. Using electrophoresis, frequencies of genotypes at the Odh, Pgm, Gpi, Hex, EsD, Pgk and Dia loci were recorded at the spat stage and, again, where possible, in juveniles. Significant deviations from expected genotype frequencies were observed at most loci in at least some cultures, but in no cases were these deviations related to the temperature at which the cultures had been reared. We speculate that these differences may have been due to variations in the density of mussel spat in the cultures. Very extensive and consistent Pgm genotype-specific mortalities were evident in one family, and, in another family, consistent deviations were seen at the Hex locus. Weaker, but nevertheless significant, deviations from expected genotype frequencies occurred at most other loci and it is concluded that, in most cases, the loci were acting as markers for a locus, or loci, on the same chromosome which had a strong effect on fitness. It is thought unlikely that evidence for similar levels of selection would be found in natural populations because, in wild populations, particular marker-locus genotypes would be associated with a range of genetic backgrounds rather than a single chromosome. the relative importance of natural selection compared with stochastic effects such as inbreeding, null alleles or migration. The development of bivalve larval-rearing techniques (Loosanoff and Davis 1963, Bayne 1965)has enabled offspring from single families and restricted matings to be reared under controlled conditions for the study of the genetics of M. edulis. In such studies, expected genotypes at a locus, and their proportions amongst offspring, can be found by genotyping the progenitors, and selection, whether direct or indirect, must be the major factor causing any significant deviation from expected proportions.Earlier studies involving laboratory-reared bivalves have produced evidence for direct or indirect selection at a number of loci , Adamkewicz et al. 1984, Gaffney and Scott 1984, Hvilsom and Theisen 1984, Mallet et al. 1985. Previous papers in this series (Beaumont et al. , 1989 have demonstrated indirect selection, unrelated to salinity, at the Pgm and Gpi loci, and direct salinity-related selection at the Lap locus. However, our data on fitness of specific genotypes at the Lap locus suggest that selection differentials observed in laboratory-reared individuals are very different from those which may apply in the wild. In this paper, we present data on genotype-specific mortalities at various loci in spat and juvenile mussels which were reared at different temperatures.