1. Sex allocation in animals is predicted to be skewed from a 1:1 ratio if sons and daughters yield different marginal fitness returns per unit maternal investment. 2. We tested this prediction using the gall-forming aphid Tetraneura sorini, in which lethal fighting is common among females, whereas male-male competition is moderate. Mothers (autumnal winged females) parthenogenetically produce females and males in their abdomen and can control their sizes and numbers. The females and males do not feed after birth. After mating, females produce single eggs, from which only females (foundresses) hatch and fight intensely with one another during the galling process. 3. Larger-sized foundresses have overwhelming advantages in fighting. If mothers invest more in individual females, they can produce larger foundresses (granddaughters), which yields greater fitness returns. This situation is contrary to the original Trivers–Willard hypothesis. Thus, we predicted that more fecund mothers in this species would invest more in females. 4. The cost of producing one female was found to be 3.0-3.2 times the cost of producing one male. Population sex ratios were male-biased (54.8%–58.2% male), but population sex allocation was highly female biased (68%–72% females). Larger mothers biased their progeny sex ratios more towards females and produced larger females, thus supporting our prediction. 5. Mothers produced two types of brood: all-female broods and mixed-sex broods. Mothers with an all-female brood produced larger females than those with a mixed-sex brood, thereby offering their granddaughters (foundresses) an advantage in fighting. 6. Local mate competition has been postulated to explain female-biased population sex allocation in gall-forming aphids. However, we concluded that competition among foundresses has led to the evolution of female-biased sex allocation at the population level and dimorphism in brood sex composition.