The discovery of superoxide dismutase was followed by a proposal that superoxide anion radical (O2.-) is a major factor in oxygen toxicity. The knowledge of superoxide chemistry, however, led some chemists to conclude that since O2.- is not very reactive in aqueous solution, the more reactive hydroxyl radical (HO.) was most likely to be the major damage causing species. Some have defended the superoxide theory by emphasizing that nonindiscriminate and selective reactivity could provide more toxicity than would high, indiscriminate reactivity. In the present study, network thermodynamic simulation was used to create a situation in which O2.- would selectively react with a substrate in a hypothetical sequence of subreactions supporting biological processes. In this situation, when the simulation of the chemical reactions was carried out using reasonable parametric values found in the literature, the selective reaction of O2.- to one molecule in the sequence caused a 95% disruption of the observable process, whereas indiscriminately targeted HO. attack caused only 0 to 35% inhibition. The major cause of the weak effect of HO. was found, in this particular model, to be a lack of sufficient availability of HO. due to both its slow generation by the Fenton reaction and a large demand for reactions with inconsequential targets. This model supports the superoxide theory of oxygen toxicity by demonstrating that a simple set of circumstances can quantitatively lead to the proposed selective superoxide toxicity. The present study also advocates the use of novel network thermodynamic simulation techniques for solving problems concerning biological oxidants and antioxidants.