When host quality varies, parasitoid wasps are expected to oviposit selectively in high-quality hosts. We tested the assumption underlying host-size models that, for solitary species of wasps, quality is based on host size. Using Ephedrus californicus, a solitary endoparasitoid of the pea aphid, we evaluated the influence of aphid size (= mass), age and defensive behaviours on host selection. Experienced parasitoid females were given a choice among three classes of 5-day-old apterous nymphs: small aphids that had been starved daily for 4 h (S4) and 6 h (S6) respectively, and large aphids permitted to feed (F) normally. Wasps attacked more, and laid more eggs in, small than large aphids (S6>S4>F). This rank-order for attack did not change when females could choose among aphids of the same size that differed in age; however, wasps oviposited in all attacked aphids with equal probability. Host size did not influence parasitoid attack rates when aphids were anaesthetized so that they could not escape or defend themselves. As predicted by host-size models, wasp size increased with host size (F>S4; S6), but large wasps required longer to complete development than their smaller counterparts (S4<S6<F). For parasitoids of growing and feeding hosts (= koinobionts), host quality is not a linear function of current host size; it is influenced by the host's age and development stage which determine the host's potential to support parasitoid growth. Host choice in E. californicus reflects a trade-off between maximization of fitness gains per egg and the economics of search-time allocation. Because large aphids are more likely to escape parasitization, a wasp must balance her potential gain in fitness by ovipositinng in a high-quality (large) aphid against her potential cost in terms of lost opportunity time if the attack fails.