Herbivores can have both direct (consumptive) and indirect (nutrient-mediated) effects on primary producer biomass and nutrient stoichiometry. Ecological stoichiometry theory predicts that herbivores of contrasting body stoichiometry will differentially remineralize nutrients, resulting in feedbacks on producer stoichiometry. We experimentally separated direct and indirect effects of aquatic vertebrate grazers on periphyton by manipulating grazer abundance and identity in mesocosms, and using grazer exclusion cages to expose periphyton to recycled nutrients in the absence of direct grazing. In experiment 1, we used a catfish with high body phosphorus (low body N:P), Ancistrus triradiatus, to assess consumptive versus nutrient-mediated effects of grazer density on periphyton. In experiment 2, we compared the nutrient-mediated effects of grazing by Ancistrus triradiatus and Rana palmipes, a tadpole with low body phosphorus and high body N:P. In experiment 1, we found that increasing catfish density led to lower biomass and particulate nutrients in periphyton through direct consumptive effects, but that nutrient-mediated indirect effects enhanced periphyton biomass when grazers were experimentally separated from direct contact with periphyton. As predicted by stoichiometry theory, nutrient recycling by this P-rich grazer tended to increase algal C:P and N:P (although effects were not statistically significant), while their consumptive effects reduced algal C:P and N:P. In experiment 2, grazer identity had strong effects on dissolved water nutrient concentrations, N recycling (measured with a 15N tracer), and periphyton stoichiometry. In accordance with stoichiometry theory, catfish increased N concentrations and recycling rates leading to higher periphyton N:P, while tadpoles had greater effects on P availability leading to lower periphyton N:P. Our experiments elucidate the importance of both the density and identity of grazers in controlling periphyton biomass and stoichiometry through consumptive and nutrient-mediated effects, and support the power of ecological stoichiometry theory to predict feedbacks on producer stroichiometry arising from consumer stoichiometry through nutrient recycling.