It is generally taken as an established fact that mitochondrial respiration is associated with the generation of small amounts of ROS (reactive oxygen species). There are many arguments supporting this side activity. A major argument is the particular physico-chemical configuration of dioxygen, which prevents the transfer of a pair of electrons. Instead, oxygen is reduced by the successive transfer of single electrons, necessarily leading to intermediates with odd electrons. The high rate of turnover of oxygen in the respiratory chain in combination with the existence of single-electron carriers supports the concept of mitochondria as the major cellular ROS generator. Experimental evidence on the ability of mitochondria to generate ROS was, however, based essentially on in vitro experiments with isolated mitochondria. A variety of structural and functional alterations associated with the removal of mitochondria from the cell, as well as the routinely applied ROS detection methods, may lead to artefactual deviation of odd electrons to dioxygen. We therefore checked these correlations in view of ROS formation, including the often reported effect of the membrane potential on the establishment of a redox couple with oxygen out of sequence. For this purpose we developed novel methods to prove the authenticity of mitochondria for ROS generation in the living cell. Based on our experiments, we can exclude spontaneous release of ROS from mitochondria. However, we describe conditions under which mitochondria can be transformed to mild ROS generators. The site of single-electron deviation to dioxygen was found to be ubiquinol interacting with the Rieske iron-sulphur protein and low-potential cytochrome b of the bc (1) complex.