We have tested the sensitivity of phytoplankton to the herbicides atrazine and nicosulfuron in experiments conduced in increasingly complex systems, from single strain phytoplankton cultures (microplates) to mesocosms mimicking whole ecosystems. The endpoints used to assess sensitivity to atrazine and nicosulfuron were total biomass increase, photosynthetic efficiency, and community diversity, depending on the system considered. Nicosulfuron appeared to be very much less toxic to phytoplankton than atrazine, in accord with the planned changes in agricultural practices to reduce the effects of surface water contamination on aquatic biota. Nevertheless, nicosulfuron had significant effects in some systems (principally microcosms), whereas the single monocultures were almost insensitive to it. This points out the inaccuracy of standardized toxicity test on phytoplanktonic algae alone for predicting the effects of xenobiotics on natural communities and the need for tests in microcosms and mesocosms to obtain reliable evidence about the toxicity of a given chemical on freshwater aquatic ecosystems.