A liquid reaction medium containing dissolved air and oxyleghaemoglobin, but no energy-yielding substrate, was supplied to bacteroids confined in a stirred flow reaction chamber. The relative oxygenation of the leghaemoglobin in the chamber was determined automatically by spectrophotometry of the effluent solution, and the concentrations of free, dissolved O2 ([O2]) and rates of O2 consumption were calculated. Dissolved CO2 and NH3 from N2 fixation were determined in fractions of the effluent solution. Bacteroids utilized endogenous reserves of poly-beta-hydroxybutyrate (PHB), which were depleted by 9.2% during a typical 5 h-long experiment. Stepwise increases in flow rate (increasing supply of O2) initially produced a drop in O2 demand and resulted in a rise in [O2] and a decline in N2 fixation. Subsequently, O2 demand rose (presumably because of increased mobilization of substrate from PHB) and [O2] declined to a low level. N2 fixation was fully restored, or even enhanced, within 15-20 min of establishment of a new, steady [O2]. This pattern of regulation by O2 supply was completely eliminated by adding low concentrations (20-50 microM) of oxidizable substrate (succinate, malate, ethanol) to the reaction medium. During endogenous activity, rates of CO2 evolution were proportional to, but less than, rates of O2 consumption up to 5.4 nmol O2 min-1 mg-1, above which CO2 evolution exceeded O2 consumption. These and other features of endogenous activity are discussed in relation to sustaining N2 fixation by nodules in vivo.