A wide range of size spectra was encountered in 11 lakes of the Quebec Eastern Townships; most were bimodal with maximum biomass represented between 1 and 4 mg (wet weight) and between 64 and 256 mg and minimum biomass between 8 and 32 mg. PCA was used to show that the two size modes represented independent size fractions and that they were related to different environmental factors. The biomass of the small size fraction increased strongly with submerged macrophyte biomass and with lake trophic status. The biomass of the large size fraction was independent of the small size fraction and increased with wave exposure at the site. As the macrophyte biomass increased, more of the invertebrates were found upon the plants, and the small size fraction became more dominant. This is probably a result of the animals upon the vegetation being more exposed to fish predators than those hidden within the substrate. Linear regression models are developed to predict the biomass of both size fractions as well as the slope and intercept parameters of the normalized size distribution (NSD). The size-fraction approach (using PCA) complements the simpler NSD approach; it detects functionally independent size fractions in the community that NSD parameters cannot reveal.