Species sensitivity distributions (SSDs) are an important predictive tool in risk assessment. Usually, literature data are used to build SSDs that are mostly based on planktonic species. But, to get adequate protective thresholds for environmental communities, one could argue that SSD should be built on ecotoxicological data obtained from species found in the ecosystem that should be protected. This is particularly true when benthic algae are of concern. Due to the lack of literature data, building SSD on benthic microalgae is difficult. This paper aims in comparing SSDs, and thus protective thresholds (hazardous concentration that affects 5 % of the species of a community, HC5), built on ecotoxicological data obtained (1) from literature and (2) with specific bioassays on benthic diatoms from a lake. Thresholds were derived for protection against four herbicides separately and for a mixture of them. Sensitivity data obtained from literature were statistically lower than the specific data for all herbicides: Species tested in the literature were usually more sensitive (mainly chlorophytes), leading to more protective lower HC5. The HC5 thresholds (literature or specific) derived for protection against the mixture were also compared to the observed sensitivity of an assemblage of benthic diatom species exposed to an increasing range of herbicide mixture concentrations. We observed that one species within the assemblage (Fragilaria rumpens) was affected at a concentration below both the literature and the specific HC5 thresholds.