The nuo-6 and isp-1 genes of C. elegans encode, respectively, subunits of complex I and III of the mitochondrial respiratory chain. Partial loss-of-function mutations in these genes decrease electron transport and greatly increase the longevity of C. elegans by a mechanism that is distinct from that induced by reducing their level of expression by RNAi. Electron transport is a major source of the superoxide anion (O(⋅) (-)), which in turn generates several types of toxic reactive oxygen species (ROS), and aging is accompanied by increased oxidative stress, which is an imbalance between the generation and detoxification of ROS. These observations have suggested that the longevity of such mitochondrial mutants might result from a reduction in ROS generation, which would be consistent with the mitochondrial oxidative stress theory of aging. It is difficult to measure ROS directly in living animals, and this has held back progress in determining their function in aging. Here we have adapted a technique of flow cytometry to directly measure ROS levels in isolated mitochondria to show that the generation of superoxide is elevated in the nuo-6 and isp-1 mitochondrial mutants, although overall ROS levels are not, and oxidative stress is low. Furthermore, we show that this elevation is necessary and sufficient to increase longevity, as it is abolished by the antioxidants NAC and vitamin C, and phenocopied by mild treatment with the prooxidant paraquat. Furthermore, the absence of effect of NAC and the additivity of the effect of paraquat on a variety of long- and short-lived mutants suggest that the pathway triggered by mitochondrial superoxide is distinct from previously studied mechanisms, including insulin signaling, dietary restriction, ubiquinone deficiency, the hypoxic response, and hormesis. These findings are not consistent with the mitochondrial oxidative stress theory of aging. Instead they show that increased superoxide generation acts as a signal in young mutant animals to trigger changes of gene expression that prevent or attenuate the effects of subsequent aging. We propose that superoxide is generated as a protective signal in response to molecular damage sustained during wild-type aging as well. This model provides a new explanation for the well-documented correlation between ROS and the aged phenotype as a gradual increase of molecular damage during aging would trigger a gradually stronger ROS response.