Food availability is considered to be a primary factor affecting animal populations, yet few experimental tests have been performed to evaluate its actual importance in species-rich ecosystems such as rainforests. It has been suggested that in such systems certain plant species may act as "keystone" resources for animals, but the importance of presumed keystone resources for populations has not been quantified experimentally. Using complementary seed removal and seed-addition experiments, we determined how the supply of a presumed keystone resource, seeds of Araucaria angustifolia, affects short-term demography of their main consumer group (small rodents) in a biodiversity hotspot, the Brazilian Atlantic Forest. We hypothesized that (i) the harvest of A. angustifolia seeds by human populations has negative impacts on rodents, and (ii) these seeds are a limiting resource for rodent populations. To test these hypotheses, we monitored populations of two species of numerically dominant rodents (Delomys dorsalis and Akodon montensis) within replicated control-experimental plots. Manipulations of seed supply over 2 years had little effect on population size, body condition, survival, or reproduction of the two rodents, suggesting that, in the short-term (within one generation), their populations are not food limited in Araucaria forests. Despite apparently having all the characteristics of a keystone resource, as currently defined in the literature, the seeds of A. angustifolia had limited influence on the short-term demography of their main consumer group. In situations where purported keystone resources are seasonally abundant, their actual importance may be lower than generally assumed, and these resources then may have only localized and temporary effects on consumer populations.