Subsidies of matter between ecosystems can strongly affect trophic interactions in food webs, yet most experiments apply a binary (i.e., subsidy vs. no subsidy) approach. This emphasis on extremes implicitly assumes linear relationships and fails to capture potential nonlinear effects in ecosystem responses along gradients of subsidy amount seen in real systems. To examine the effects of such gradients, we manipulated leaf litter biomass across seven subsidy levels in 1000-L replicated experimental ponds. As our subsidy recipient, we used a simple food web that included southern leopard frog tadpoles (Rana sphenocephala) and spotted salamander larvae (Ambystoma maculatum) as the top vertebrates. We measured temperature, dissolved N and P; dissolved organic carbon (DOC); dissolved oxygen (DO); phytoplankton and periphyton biomass; and amphibian biomass export. Dissolved N, phytoplankton and periphyton biomass were linearly related to subsidy amount, but DOC, periphyton ash-free dry mass, and DO were nonlinearly related to subsidies. Frog biomass export increased as a function of subsidy amount, while salamander biomass export increased up to moderate subsidy levels before decreasing to very low levels at the highest subsidies, suggesting not only a nonlinear response, but a non-monotonic one. Our data suggest that under high subsidies, despite high nutrient levels, the basal resource supply may have switched from an algal to a detrital one. These data show that a dose-dependent approach is fundamental for understanding how subsidy amount affects ecosystems.