The balance of forces determining the successful control of ragwort Senecio jacobaea by introduced insects was investigated in a field experiment by manipulating the time of disturbance, the level of interspecific plant competition, and the level of herbivory by the cinnabar moth Tyria jacobaeae and the ragwort flea beetle Longitarsus jacobaeae. We used a factorial design containing 0.25—m2plots arranged as 4 Blocks × 2 Disturbance Time (plots were tilled in Fall 1986 or Spring 1987) × 3 Plant Competition levels (vegetation other than ragwort was Removed, Clipped, or Unaltered) × 2 Cinnabar Moth levels (Exposed, Protected) × 2 Flea Beetle levels (Exposed, protected). The response of ragwort was measured as colonization, survivorship, and reproduction of the first ragwort generation, establishment of juveniles in the second generation, and changes in ragwort biomass from 1987 through 1990. We also made annual measurements from 1987 through 1990 of the allocation of space (the limiting resource in the Unaltered competition treatment) among the categories ragwort, other species, litter, and open space. Natural enemy responses were characterized by relating variation in the concentration of enemies and the concentration of ragwort among patches. We found that abundant buried seed and localized disturbances combined to activate incipient ragwort outbreaks, and that interspecific plant competition and herbivory by the ragwort flea beetle combined to inhibit the increase and spread of incipient outbreaks. Time of disturbance had little effect on the outcome of biological control. Under conditions in the Removed and Clipped treatments (where there was sufficient open space for germination and establishment), reduction in seed production in the first generation caused by cinnabar moth larvae led to a reduction in plant numbers in the second generation, but caused only a weak effect on ragwort cover and no detectable effect on ragwort biomass over the longer term from 1986 through 1990. At the spatial scale examined, inhibition by the ragwort flea beetle and plant competition took the extreme form of elimination of all ragwort individuals except the pool of seed buried in the soil. Our findings lead us to (1) reject the view that successful biological control leads to a stable pest—enemy equilibrium on a local spatial scale, (2) strongly endorse "search and destroy" and weakly endorse "complementary enemies" strategies suggested by Murdoch et al. (1985) as ways to improve control, and (3) emphasize resource limitation in the pest at low density as a key feature distinguishing biological control of weeds from biological control of insects.