Collards were grown at Ithaca, New York, in two experimental habitats: pure stands and single rows that were bounded on each side by diverse, meadow vegetation. The arthropods associated with these plants were sampled on 20 dates over a 3—year period. The status of the herbivore species was measured by their rank in biomass in each sample. The two most prominent species, Phyllotreta cruciferae and Pieris rapae, maintained high status throughout the investigation, but another important species, Brevicoryne brassicae, was absent for an entire season. Pit feeders usually formed the most important herbivore guild. Nevertheless, the guild spectrum, which describes the functional structure of the fauna, varied widely in time and space. The size distributions of species and of individuals were both highly skewed toward the smaller sizes. Herbivore loads, the mean biomass of herbivores per 100 g of consumable foliage, were consistently higher in the pure stands. Moreover, herbivore loads varied significantly with season in each experimental habitat. Both the number of herbivore species and the diversity of the herbivore load were greater in the diverse habitat. Biomass was more heavily concentrated among the prominent herbivores in the pure stands; increased dominance, rather than differences in species richness, appeared to be the major cause for the lower herbivore diversity in this habitat. The diversity of predators and parasitoids was higher in the pure stands. Most of the abundant species found on collards shared a similar narrow range of hosts. As a result the species in this core group of herbivores and parasitoids were regularly associated with each other. Predators and the less abundant herbivores tended to be less specialized and served to link the collard association with the surrounding community. Plant—arthropod associations are representative of component communities, well—integrated systems that form portions of larger compound communities. This distinction facilitates the analysis of community structure. Microclimates and the effectiveness of "enemies" did not appear to differ sufficiently in the two experimental habitats to account for the observed differences in the herbivore load. The results suggest a new proposition, the resource concentration hypothesis, which states that herbivores are more likely to find and remain on hosts that are growing in dense or nearly pure stands; that the most specialized species frequently attain higher relative densities in simple environments; and that, as a result, biomass tends to become concentrated in a few species, causing a decrease in the diversity of herbivores in pure stands.