We measured production of reactive oxygen species by intact mitochondria from rat skeletal muscle, heart, and liver under various experimental conditions. By using different substrates and inhibitors, we determined the sites of production (which complexes in the electron transport chain produced superoxide). By measuring hydrogen peroxide production in the absence and presence of exogenous superoxide dismutase, we established the topology of superoxide production (on which side of the mitochondrial inner membrane superoxide was produced). Mitochondria did not release measurable amounts of superoxide or hydrogen peroxide when respiring on complex I or complex II substrates. Mitochondria from skeletal muscle or heart generated significant amounts of superoxide from complex I when respiring on palmitoyl carnitine. They produced superoxide at considerable rates in the presence of various inhibitors of the electron transport chain. Complex I (and perhaps the fatty acid oxidation electron transfer flavoprotein and its oxidoreductase) released superoxide on the matrix side of the inner membrane, whereas center o of complex III released superoxide on the cytoplasmic side. These results do not support the idea that mitochondria produce considerable amounts of reactive oxygen species under physiological conditions. Our upper estimate of the proportion of electron flow giving rise to hydrogen peroxide with palmitoyl carnitine as substrate (0.15%) is more than an order of magnitude lower than commonly cited values. We observed no difference in the rate of hydrogen peroxide production between rat and pigeon heart mitochondria respiring on complex I substrates. However, when complex I was fully reduced using rotenone, rat mitochondria released significantly more hydrogen peroxide than pigeon mitochondria. This difference was solely due to an elevated concentration of complex I in rat compared with pigeon heart mitochondria.