1. To improve mechanistic understanding of plankton responses to eutrophication, a mesocosm experiment was performed in the shallow littoral zone of a south Swedish lake, in which nutrient and fish gradients were crossed in a fully factorial design. 2. Food chain theory accurately predicted total biomass development of both phyto- and zooplankton. However, separating zooplankton and algae into finer taxonomic groups revealed a variety of responses to both nutrient and fish gradients. 3. That both nutrients and fish are important for phytoplankton dynamics was seen more clearly when viewing each algal group separately, than drawing conclusions only from broad system variables such as chlorophyll a concentration or total phytoplankton biovolume. 4. In some taxa, physiological constraints (e.g. sensitivity to high pH and low concentrations of free CO2) and differences in competitive ability may be more important for the biomass development than fish predation, grazing by herbivorous zooplankton, and nutrient availability. 5. We conclude that food chain theory accurately predicted responses in system variables, such as total zooplankton or algal biomass, which are shaped by the dynamics of certain strong interactors ('keystone species'), such as large cladocerans, cyanobacteria and edible algae (<50 mum), whereas responses at finer taxonomic levels cannot be predicted from current theory.