We examined the effect of subsample size on the accuracy of information obtained from aquatic macroinvertebrate assemblage samples. Subsamples containing 100 organisms or 300 organisms were compared on the bases of processing time and the ability to discern ecological differences among samples. Independently of subsample size, assemblages differed between study streams, primarily reflecting an intermittent vs. permanent stream difference, and between seasons at most streams. It required, on average, two additional hours to process the larger subsamples. Larger subsamples gave significantly higher estimates of total richness and Ephemeroptera, Plecoptera, and Trichoptera (EPT) richness, but the relative abundances of many assemblage subsets (e.g., EPT organisms and most functional feeding groups) were similar using both subsample sizes. Larger subsamples did not typically enhance the ability to discriminate between samples from different seasons, but did more accurately distinguish among streams when differences were subtle. They also appeared to avoid Type I error in comparisons of compositionally similar reaches within a study stream.