We examined in the laboratory and field the foraging behavior of the stenophagous egg parasitoid Anagrus delicatus (Hymenoptera: Mymaridae) within discrete patches of its planthopper host, Prokelisia marginata (Homoptera: Delphacidae) on leaves of salt marsh cord grass, Spartina alterniflora. Females efficiently discriminated between leaves with and without hosts, leaving empty leaves within 6 min, on average. On leaves with hosts, wasps remained ca.10 times as long (mean +/- 1 se = 62 +/- 8 min); time on a patch was independent of host density. In the laboratory, newly mated female A. delicatus encountering hosts for the first time and bearing an average of 33 yolked and mature eggs, parasitized an average of 6 hosts (laying l8% of her eggs) before dispersing from patches containing an average of 81 hosts. Many more hosts were probed than were parasitized. In field experiments with mixtures of newly mated and experienced females, the proportion of egs laid by these wasps was even lower; between 4% and 26% were laid in a series of experiments that spanned 20 dates over 10 mo. In all tests, available hosts far exceeded wasp fecundity. Most, if not all hosts, were alive, available, and supported development of parasitoids if parasitized. Increasing numbers of female wasps visiting a patch led to increasing parasitism rate, to an average of 75% with 20 female wasps. Even though many hosts on a patch were probed and rejected by each wasp, rejected hosts were parasitized by subsequent wasps and yielded live parasitoid offspring; rejected hosts were not unsuitable hosts, Leaf thickness provided no refuge from parasitism by A. delicatus. Hosts laid deep within plant leaves (mean = 118 um) were no less vulnerable to attack than those laid shallow, and all hosts were well within the range of the long ovipositor of A. delicatus (mean = 394 um). Single wasps dispersed before the ratio of unparasitized to parasitized hosts was much decreased by their efforts; dispersal was not precipitated by high encounters with parasitized hosts. Handing time constraints were small (mean = 2.2 min) in relation to total patch time (mean = 62 min), and thus did not account for the low number of ovipositions. Interference among searching parasitoids was not responsible for the underutilization of hosts in the field because single wasps, searching alone in new patches not previously visited by A. delicatus, displayed the same characteristically low attack rates as wasps searching together with others. Classical arguments of parasitoid foraging, which assume maximization of oviposition rate with respect to time, are inconsistent with the substantially submaximal oviposition rates for this parasitoid. Anagrus delicatus passed up many suitable and available hosts in each patch and visited multiple patches during its lifetime; number of eggs laid increased linearly with the number of patches visited. Wasps visiting > 5 patches laid an average of 32 eggs (95% of average egg complement). Our hypothesis is that this is a foraging strategy build upon compromising the time rate of oviposition in favor of spreading ovipositions among patches and sites. We observed high mortality of host-plant leaves, and thus of host-insect patches, in the field due to leaf senescence (20-30%), which would favor spreading of parasitoid offspring among leaves.