In assessments of ecological impact in invasion ecology, most studies compare un-invaded sites with highly invaded sites, representing the ‘worst-case scenario’, and so there is little information on how impact is modified by the population density of the invader. Here, we assess how ecological impact is modified by population density through the experimental development of density-impact curves for a model invasive fish. Using replicated mesocosms and the highly invasive Pseudorasbora parva as the model, we quantified how their population density influenced their diet composition and their impacts on invertebrate communities and ecosystem processes. The density–impact curves revealed both linear and non-linear density–impact relationships. The relationship between P. parva density and zooplankton body mass was represented by a low-threshold curve, where their impact was higher at low densities than predicted by a linear relationship. In contrast, whilst the relationship between density and zooplankton biomass and abundance was also non-linear, it was high-threshold, indicating a lower impact than a linear relationship would predict. Impacts on diversity and phytoplankton standing stock were linear and impacts on benthic invertebrate abundance and decomposition rates were represented by s-shaped curves. These relationships were underpinned by P. parva dietary analyses that revealed increasing reliance on zooplankton as density increased due to depletion of other resources. We caution against the common assumption that ecological impact increases linearly with invader density and suggest that increased understanding of the relationship between invader population density and ecological impact can avoid under-investment in the management of invaders that cause severe problems at low densities.