Parameters necessary for predicting direct and correlated responses for large and small 13-day larval weight in T. castaneum on two levels of nutrition were estimated in the base population. Larval weight in the GOOD environment was approximately twice that observed in POOR. Heritabilities (estimated from the ratio of sire component to total phenotype variance) of larval weight on the two rations were similar, 0·21 ± 0·06 and 0·19 ± 0·05 for GOOD and POOR, respectively. Heritabilities based on dam-offspring covariances were similar to these, but those obtained from full-sib covariances were much larger (0·97 ± 0·07 for GOOD and 0·69 ± 0·07 for POOR). This suggested that considerable dominance rather than maternal effects were present. The genetic correlation between growth on GOOD and growth on POOR was estimated as + 0·60 ± 0·21.
The selection experiment was replicated four times with each replication extending over eight generations. Good agreement between predicted and observed values for direct selection was observed for large selection in both environments and small selection in POOR. However, response to small selection in GOOD was significantly greater than predicted in all four replications and was associated with increased selection differentials. Realized heritabilities were approximately the same for both directions in GOOD yet asymmetrical responses occurred for all replications due to unequal selection differentials. On the other hand, realized heritabilities were asymmetrical in POOR. Those observed for small selection were more than twice the size of those calculated for large lines. However, the responses in POOR were symmetrical since the selection differentials varied inversely with the realized heritabilities.
Because of the asymmetry observed for heritabilities and selection differentials, correlated responses were poorly predicted. The average effective genetic correlation between growth in GOOD and growth in the POOR environment agreed remarkably well with the base estimate, yet asymmetry of the genetic correlation was a consistent phenomenon with values for the large lines being less than the base parameter while small lines were uniformly larger.
Asymmetries of the various genetic parameters were not anticipated from base estimates. They were not caused by sampling or chance fluctuations since all four replications were remarkably consistent. Asymmetry for any one genetic parameter (e.g. heritability) was associated with a particular environment or direction of selection while other genetic parameters reacted asymmetrically in populations exposed to a different set of environmental treatments.
For maximum performance in a single environment, these results show that selection should be practiced in that environment. With regard to mean performance in GOOD and POOR environments, selection for large size in POOR gave some 25% more gain than selection in GOOD. Selection for small size in either environment was equally effective in obtaining minimum size in both environments.