A microcosm experimental program has been driven to test the effects of metolachlor on natural multispecific phytobenthic diatom communities issued from rivers and to identify potential descriptors of effects then usable for the diagnosis of pesticide toxic impact in situ. Concerning global quantitative approaches, the colonization of glass artificial substrates by natural communities in artificial channels didn't significantly differ between control and contaminated modalities (5 and 30 μg/l) after 2 weeks of exposure. Nevertheless, a non-parametric multivariate analysis (NP-MANOVA) processed on specific relative abundance enumeration results revealed significant differences in the final composition of multispecific diatom communities. Differences were attributed to the increase of relative abundance of tolerant species in channels exposed to Metolachlor (e.g. planothidium lanceolatum, Surirella brebissonii, Nitzschia gracilis ). A significant increase of teratologic deformities has also been demonstrated in relation with the exposure to metolachlor; notably the species Surirella angusta revealed the occurrence of frustule deformities at a high level (percentage reaching 10 %). In order to confirm these results, further ecotoxicological tests have been driven on monospecific cultures of species, one appearing sensitive (Achnanthidium minutissimum), another one classified as tolerant but able to reveal teratologic deformities (S. angusta). These short-term exposure experiments allowed us to precise ecotoxicological values (i.e. EC 10, EC 50) at a species level, and to refine some conclusions about the effects of metolachlor derived from the first experiment on diatom multispecific communities.