Faunal and chemical data from mountain streams in France were analysed to examine the accuracy of a multimetric approach based on biological and ecological traits of benthic macroinvertebrate communities to assess toxic sediment pollution in streams. A toxic quality class (from high' to bad') was pre-assigned to each site using the French water quality assessment system based on toxic substances (metals, PAH, PCB) in sediment. Each trait was described in multiple categories. The affinity of macroinvertebrate taxa for the different categories of a trait was described using a fuzzy coding procedure. The relative abundance of trait categories was calculated for communities at each site. A non-parametric multiple comparison statistical procedure was used to compare relative abundances of trait categories between groups of sites assigned to different quality classes to identify the combinations of trait categories that best separated sites between adjacent toxic quality classes. Based on such sets of trait categories, we propose a statistical procedure to allocate sites to toxic quality classes from the attributes of its benthic macroinvertebrate community. This study shows that biological criteria based on macroinvertebrate traits could provide new methods for biological assessment. This is the first step towards an in situ functional tool of stream sediment contamination assessment at community level. However, we need to increase the number of sites to improve the tool's precision, extend the procedure to additional stream types and validate the design at a larger spatial scale.