For many crops, we have poor knowledge about the relationship between pest density and damage. However, investigating pest harmfulness is particularly relevant currently in the search for alternative crop protection strategies that are unlikely to totally suppress pest populations. Here, we assessed the harmfulness of Delia radicum (L.) on broccoli (Brassica oleracea var. italica Plenck). We worked inside insect-proof cages set up in the field with additional pitfall traps to remove ground dwelling predators. Plants were manually infested with 10 levels of pest density ranging from 0 to 100 individuals per plant, following a natural infestation pattern. Surprisingly, no plants died but almost 100% of the pests introduced died over the course of the experiment. However, all broccoli development and growth traits were negatively correlated with pest density and broccoli head mass at harvest decreased linearly with pest density. The observation over time of development and growth traits showed evidence of plant compensation, suggesting that the head mass of individual plants may have reached similar values if allowed to fully mature. The relationship between pest density and damage, together with forecast models of pest population dynamics could be used to develop decision support tools assessing the relevance of preventative treatments. Keywords Crop losses; plant traits; plant injury; damage; plant compensation. Introduction Throughout the history of agriculture, the development of pest management strategies has led to dramatic decreases in crop losses to pests (Oerke 2006). However, the predominant crop protection method, that is pesticide usage, has various negative consequences, notably on human health and the beneficial fauna of agro-ecosystems (Desneux et al. 2007; Geiger et al. 2010). Alternatives are therefore urgently needed, but in contrast to pesticides, most of these alternatives (push-pull, insect netting, intercropping…) are likely to have partial efficiency. For example, conservation biological control, based on the enhancement of natural enemy populations, is thought to be an interesting method for reducing crop losses (Eilenberg et al. 2001) but is incompatible with the complete suppression of pest populations because by