Three generations of upward selection for the egg mass/adult weight ratio were carried out in Tribolium castaneum. The experiment involved five lines: E - selected for increased egg mass; W - selected for decreased adult weight; R - selected for egg mass/adult weight directly; L - linear index selected with economic weights of m 2: -m 1 egg mass to adult weight (m 1 and m 2 are the means for adult weight and egg mass, respectively); NL - nonlinear index selected. Adult weight (7 days after adult emergence) and egg mass (between 7 and 11 days) were measured. The NL, E, and L lines had the greatest observed responses for the ratio; the R and W lines were not effective in improving the egg mass/adult weight ratio. It was expected that the NL line would be superior to the E, L, and R lines, and that the W line would respond the least. Observed response was significant for egg mass in the NL line, and for adult weight in the E, W, and R lines. Strong selection to increase egg mass seems to represent the optimal criterion for the ratio to be improved. The usefulness of nonlinear indices as selection criteria to improve a nonlinear trait, previously found to be optimal for a trait defined as the product of two component traits, appears to hold also for the selective improvement of the ratio of two traits. Serious limitations expected for direct selection of the ratio have been confirmed in this experiment.