In recent years it has become apparent that when studying plant-herbivore interactions, it is important to also consider the multitrophic context, i.e. to what extent generalist and specialist enemies act as bodyguards for the plant. I have performed a field experiment with 25 genets, varying in plant defense levels, of meadowsweet (Filipendula ulmaria) and its herbivore, the chrysomelid beetle Galerucella tenella. Each genet was treated with four different density manipulations of G. tenella (5, 15, 30, and 60 larvae per genet). The experiment was set up on an archipelago island with natural levels of a specialist hymenopteran parasitoid Asecodes parviclava and generalist natural enemies. I set up the following hypotheses: 1) larval survival will be higher on less defended plants than on more defended plants, 2) the proportion of parasitoids will be higher on more defended plants, and 3) larval survival will increase with increasing larval densities. The most important findings of the experiment were an overall very low total survival of larvae (G. tenella pupae and A. parviclava mummies), a generally high parasitism by A. parviclava on more defended plants, and that two groups of generalist predators (hunting and web-forming spiders) were more frequent on less defended plants. My study suggests that herbivore survival may not be higher on less defended plants under situations with abundant occurrence of generalist predators, as they are able to trace high-quality foraging sites.