The toxicity of a series of organophosphorus (OP) and carbamate insecticides was measured against the midge Chironomus riparius in aquatic systems with and without sediment. Five molecular descriptors (molecular volume, Henry's law constant, n-octanol/water partition coefficient (KO,), molecular connectivity, and linear solvation energy) were used in regression analysis as potential predictors of insecticidal activity. The regressions were conducted for each descriptor against toxicity values for the series of chemicals.Molecular volume and Henry's law constant showed no relationship with toxicity. However, log KO, was moderately successful in describing the effect of sediment on toxicity (rZ = 0.508). Prediction of toxicity was substantially improved when a linear solvation energy (LSE) or molecular connectivity (MC) model was used in regressions. In multiple regressions conducted on carbamates arid OPs separately, use of MC or LSE parameters explained up to 95.8% of the variability in toxicity. Based on the results of regression analyses, sorptive interactions between these insecticides and sediment apparently dominate the processes affecting the toxicity of these compounds when sediment is present. In the absence of sediment, the regressions suggest that the molecular structure of the insecticides is more important than solubility or partitioning for determining toxicity.