Abstract. Understanding the impact of herbivory on plant populations is a fundamental goal of ecology.Damage to individual plants can be visually striking and affect the fates of individuals, but these impacts do not necessarily translate into population-level differences in vital rates (survival, growth, or fecundity) or population growth rates. In biological control of weeds, quantitative assessments of population-level impacts of released agents on both target invasive plants and native, nontarget plants are needed to inform evaluations of the benefits and risks of releasing agents into new regions. Here we present a 3-yr experimental demographic field study using the European root-feeding biocontrol weevil, Mogulones crucifer, first released in Canada in 1997 to control the invasive weed Cynoglossum officinale (Boraginaceae). Mogulones crucifer is an effective "search and destroy" agent in Canada, but sporadically feeds, oviposits, and develops on native nontarget Boraginaceae. We investigated the population-level impacts of this biocontrol insect on its target weed and a native nontarget plant, Hackelia micrantha (Boraginaceae), by releasing large numbers of weevils into naturally occurring patches of H. micrantha growing isolated from or interspersed with C. officinale. We followed the fates of individual plants on release and nonrelease (control) sites for two transition years, developed matrix models to project population growth rates (λ) for each plant species, and examined the contributions from differences in vital rates to changes in λ using life table response experiments (LTRE). In contrast to studies of the insect-plant interaction in its native range, as a biocontrol agent, M. crucifer increased mortality of C. officinale rosettes in the year immediately following release, depressing the weed's λ to below the population replacement level. However, λ for H. micrantha was never depressed below the replacement level, and any differences between release and nonrelease sites in the nontarget could not be explained by significant contributions from vital rates in the LTRE. This study is the first to simultaneously and experimentally examine target and nontarget population-level impacts of a weed biocontrol insect in the field, and supports the theoretical prediction that plant life history characteristics and uneven herbivore host preferences can interact to produce differences in populationlevel impacts between target and nontarget plant species.