The oxidation reaction of human oxyhemoglobin with nitrite is complex and is characterized by a lag period followed by an autocatalytic phase. On the basis of contradictory experimental results, in order to describe the time-course of the reaction, two different mechanisms have been proposed, involving either hydrogen peroxide or the superoxide anion as reaction intermediates. This paper reports a careful reinvestigation of this reaction carried out as a function of reagent concentration, buffer composition, presence of enzymatic scavengers of oxygen radicals or of other compounds which may affect the intermediate steps of the reaction. The results obtained show that: hydrogen peroxide can be definitely identified as the reaction intermediate, in agreement with the mechanism proposed by Kosaka et al. (Biochim. Biophys. Acta 702 (1982) 237-241); the reaction time-course depends in a different way on the concentrations of hemoglobin and nitrite, a finding that cannot be explained on the basis of this mechanism. A more complex reaction scheme is proposed, that provides a satisfactory description in quantitative terms for all the available experimental data.