A study was undertaken to investigate the environmental impact of herbicides on natural communities of freshwater periphyton and phytoplankton in the river Ozanne and in related nearby water reservoirs, including both pristine and pesticide- (atrazine and isoproturon) contaminated stations. The microalgal toxicity of both herbicides was investigated by short-term studies, using the in vivo fluorescence pattern to perform dose-effect experiments. The taxonomic composition of the communities sampled was assessed by microscopy and by HPLC pigment analysis. The EC50 (periphyton) or EC125 (phytoplankton) values, calculated using in vivo fluorescence endpoints, increased with the herbicide concentration found in the water. In contrast, the structure of the algal communities (periphyton) inhabiting the contaminated stations seemed to be permanently affected when compared to the reference community. A 'memory effect' could be detected, both in herbicide sensitivity and in the structure of periphytic communities that persisted even when peak contaminations had disappeared. This study shows that the response of algal communities is likely to reflect past selection pressures, and suggests that the function and structure of a community could both be modified by the persistent or repeated presence of microcontaminants in natural environments. We could use short-term ecotoxicological tests on freshwater microalgae to assess the effects of past temporary contaminations by agricultural pesticides, and combining this with diversity indices could make it possible to assess the ecotoxicological risk of herbicide contamination even when a complete chemical analysis of the contamination is not feasible.