Although there have been a number of spectacularly successful biological control introductions, most introduced natural enemies have contributed little if any to the regulation of their target weed. Understanding why some biological control agents have been successful while others have not is of central importance to improving the efficacy and reducing the risk of biological control programs. Recent theoretical and empirical work demonstrates that many invasive plants are more common in high- than in low-resource areas, tend to be better competitors under high-resource conditions and tend to be more tolerant of herbivory under high- vs. low-resource conditions. Thus, populations of invasive weeds that are growing under high-resource conditions are those most in need of biological control and are also those most likely to be most tolerant of herbivore impact. Successful biological control agents must be able to overcome this resource-based variation in plant tolerance. I will present a theoretical model for weed biological control that is based on biological control agent responses to variation in resource availability to the weed host. This model predicts that a key attribute of successful agents is that they aggregate on high resource plants. This may arise from: 1) preference for high resource plants or 2) higher performance on high resource plants, coupled with aggregation or limited dispersal.