An experiment conducted from 1988 to 1997 to determine the effects of the timing of application and nutrient supply (particularly of phosphorus) is reported. The sources of applied nutrients that were compared were farmyard manure, pre-rice green-manuring with Sesbania aculeata and fertilizer application in a rice-wheat rotation on a typic ustifluvent. The application of seven tonnes farmyard manure per hectare to both the rice and the wheat crops over eight years increased organic carbon levels from 1.4 to 1.6% but had no yield effect on either crop. Phosphorus application through farmyard manure was not adequate for rice, whilst an application of 34 kg P ha−1 to the rotation gave an economic yield increase only in rice and then only in the first four years of the experiment. From the third year, green manuring was able to replace the effects of the recommended nitrogen, phosphorus and potassium fertilizer applications in increasing rice yield. Green manuring had no effect on the wheat yield but the recommended fertilizer application increased yield. Green manuring increased soil total nitrogen and available potassium levels and reduced base saturation. After adjusting rice yields for variation in transplanting date between years there was no statistical evidence of a yield trend in either crop over the period of the experiment. Farmers' practice of applying seven tonnes farmyard manure per hectare appears adequate to produce stable rice paddy yields of 4–6 t ha−1 a−1.