Experiments conducted over three seasons in southern New South Wales tested the effects of concentrating anhydrous ammonia (AA) and urea fertiliser in bands occupying ~3.5% of the topsoil volume. Yield responses to applied nitrogen (N) were small or negative in a drought but larger (17 kg grain kg'1 N fertiliser) in favourable seasons. There was no consistent difference between AA and urea effects on yield, grain protein or efficiency of fertiliser-N recovery, and there were no consistent differences arising from banding depth or application time. Gaseous loss of ammonia to the atmosphere was negligible from urea granules or AA injected into the soil as gas or liquid. Soil ammonium concentration was >700 ug N g'1 in bands of ~5 cm diameter when measured 6 days after AA application but halved within 5 weeks due to nitrification. Within 1 day of banding AA or urea at sowing, pHwater in the bands rose from 6 to 8.5, leading to transient changes in microbial activity and populations. Immediately after banding, microbial biomass carbon and numbers of protozoa fell by about half, but numbers of ammonia- and nitrite-oxidisers were unchanged. Five weeks later, microbial biomass carbon and protozoa had partly recovered whereas numbers of ammonia- and nitrite-oxidisers increased 5'10-fold. After 7 months, there was a small reduction in microbial diversity in the bands, shown by analysis of fatty acid methyl esters. Seedling growth was slower where N fertiliser was applied in concentrated bands than when mixed throughout the topsoil, supporting previous research showing that roots avoid bands of highly concentrated ammonium. Banding thus provided a slow-release form of N to wheat crops, thereby reducing excessive seedling growth and the risks of haying-off.