About 65% of all emissions of nitrous oxide, N2O, are from soils, and are caused by aerobic nitrification and anaerobic denitrification. Tropical forest soils are probably the most important single source, followed by cultivated soils. Emission rates in natural systems are related to the rate of N mineralization from organic matter, and N deposition; in agricultural systems they are related to the quantities of N used as fertilizers and, where relevant, to recent land use change. The global budget for N2O is not well balanced, and sources may still be underestimated. Direct evidence of a positive feedback of global warming on N2O emissions comes from studies of air in ice cores. One of the projected effects of future global warming is a lowering of water tables in northern peatlands; experiments suggest that this would lead to increased emissions, but that the effect on total emissions would be small. The results of many experiments with non-peatland soils indicate that the effect of temperature on soil emissions is generally positive, and that the rate of increase may be very steep when denitrification is the principal process involved. Process-level modelling suggests that the reason is increased soil respiration, which causes an increase in anaerobic volume in which denitrification can take place, in addition to the increased denitrification rate per unit anaerobic volume brought about directly by the rise in temperature. These results imply that generally a positive feedback on emissions from soils is likely. However, in some environments, a large proportion of total annual emissions can occur during freeze–thaw cycles; such cycles may become more or less frequent, depending on the climatic zone, and this may result in either a positive or negative feedback effect due to global warming. Models of global and regional trends give very conflicting predictions of the direction and the magnitude of climatic impacts on fluxes, but the prediction of a positive feedback seems to be the more soundly based.