Spatial variability in the microbial community composition of river biofilms was investigated in a small river using two spatial scales: one monitored the upstream–downstream pesticide contamination gradient, referred to as the ‘between-section variability’, and the other monitored a 100-m longitudinal transect (eight sampling sites per section) within each sampling section, referred to as the ‘within-section variability’. Periphyton samples were collected in spring and winter on artificial substrates placed in the main channel of the river. Denaturing gradient gel electrophoresis (DGGE) was used to assess the prokaryotic and eukaryotic community richness and diversity, and HPLC pigment analysis to assess the global taxonomic composition of the photoautotrophic community. In order to try to reduce the biological variability due to differences in flow velocity and in light conditions within each sampling section, and consequently to take into account only the changes due to water chemistry, nine plates (three per sampling section) subjected to similar physical conditions were chosen, and the results for these plates were compared with those obtained for all 24 plates. As shown by DGGE and by HPLC analyses, using these three substrate plates exposed to similar environmental conditions did indeed reduce the within-section variability and maximize the between-section variability. This sampling strategy also improved the evaluation of the impact of pollutants on the periphytic communities, measured using short-term sensitivity testing.