Abstract Three quite different coals were burned batchwise in electrically heated beds of sand fluidized with O2 and N2 at either 800 or 900°C. The coal was injected into the bed either as a single piece (1 or 2 cm in size) or as 1.0 g batches of smaller particles (either 1.4–1.7 or 2.0–2.4 mm in diameter). The concentrations of NOx (i.e., [NO] + [NO2]), N2O, CO, and CO2 in the off-gases were measured as functions of time whilst the coal was burning. It proved relatively easy to separate the emission of any one of these gases into that from the initial, devolatilization stage of combustion and that from the subsequent burning of the char. The first stage of coal combustion, when the volatile matter burns, turns out to be very important. For a low rank coal, 70%–90% of all the N2O produced appears while the coal is undergoing devolatilization. The fraction drops to 40% for a coal with a low volatile content. The figures for NOx emissions range from 55% of the NOx arising from the volatiles in a low rank coal to 25% for a coal of high rank. In general, the effects of bed temperature and coal size were much less important than a coal's volatile content. Interestingly, changing the bed temperature altered the ratio [N 2 O] [NO x ] in the off-gases and not the total quantity of oxides of nitrogen emitted. Lower concentrations of O2 resulted in slightly less N2O and NOx being produced. The rates of production of NOx and N2O during combustion of the volatiles were found to be proportional to one another. This seems to derive from a lack of mixing of the volatiles. The fact that large (> 1 cm) coal particles float on top of a fluidized bed during devolatilization is important and, e.g., can result in large particles producing less NOx and N2O during devolatilization than tiny particles. The observations made during devolatilization conform to effectively all the fuel-nitrogen in the volatile matter being converted to HCN. Nitric oxide is then produced most probably heterogeneously on the sand from CN radicals, which alternatively can yield NCO radicals. N2O is generated during devolatilization by the reaction NCO + NO → N2O + CO occurring, probably in the gas phase. As for the burning of char in a fluidized bed, the particle size (> 1 mm) here is large enough for there to be control by mass transfer of oxygen to a particle. Oxidation is thus confined to the exterior of the char. It appears that CO, from burning char, is important, together with reaction between NO and solid carbon, in converting NO to N2. This is why the yield of NO from char-N is less than that from volatile-N.