Single-species acute toxicity data and (micro)mesocosm data were collated for 16 insecticides. These data were used to investigate the importance of test-species selection in constructing species sensitivity distributions (SSDs) and the ability of estimated hazardous concentrations (HCs) to protect freshwater aquatic ecosystems. A log-normal model was fitted to a minimum of six data points, and the resulting distribution was used to estimate lower (95% confidence), median (50% confidence), and upper (5% confidence) 5% HC (HC5) values. Species sensitivity distributions for specific taxonomic groups (vertebrates, arthropods, nonarthropod invertebrates), habitats (saltwater, freshwater, lentic, lotic), and geographical regions (Palaearctic, Nearctic, temperate, tropical) were compared. The taxonomic composition of the species assemblage used to construct the SSD does have a significant influence on the assessment of hazard, but the habitat and geographical distribution of the species do not. Moreover, SSDs constructed using species recommended in test guidelines did not differ significantly from those constructed using nonrecommended species. Hazardous concentrations estimated using laboratory-derived acute toxicity data for freshwater arthropods (i.e., the most sensitive taxonomic group) were compared to the response of freshwater ecosystems exposed to insecticides. The sensitivity distributions of freshwater arthropods were similar for both field and laboratory exposure, and the lower HC5 (95% protection with 95% confidence) estimate was protective of adverse ecological effects in freshwater ecosystems. The corresponding median HC5 (95% protection level with 50% confidence) was generally protective of single applications of insecticide but not of continuous or multiple applications. In the latter cases, a safety factor of at least five should be applied to the median HC5.