Analyses of 21 15N stable isotope tracer experiments, designed to examine food web dynamics in streams around the world, indicated that the isotopic composition of food resources assimilated by primary consumers (mostly invertebrates) poorly reflected the presumed food sources. Modeling indicated that consumers assimilated only 33–50% of the N available in sampled food sources such as decomposing leaves, epilithon, and fine particulate detritus over feeding periods of weeks or more. Thus, common methods of sampling food sources consumed by animals in streams do not sufficiently reflect the pool of N they assimilate. Isotope tracer studies, combined with modeling and food separation techniques, can improve estimation of N pools in food sources that are assimilated by consumers. Food web studies that use putative food samples composed of actively cycling (more readily assimilable) and refractory (less assimilable) N fractions may draw erroneous conclusions about diets, N turnover, and trophic linkages of consumers. By extension, food web studies using stoichiometric or natural abundance approaches that rely on an accurate description of food-source composition could result in errors when an actively cycling pool that is only a fraction of the N pool in sampled food resources is not accounted for.