The effect of different rates (0–140 kg/ha) of nitrogen fertilizers on soil cyanobacteria and rice crop performance were studied in a rice-cropping system on an alkaline Fluvent soil at Valencia, Spain, during three consecutive crop seasons (1990–92). The results showed that the rice fields of Valencia favour the development of N2-fixing cyanobacteria. Nitrogen fixation varied during the cultivation cycle, reaching its highest values at the maximum tillering stage, 5–6 weeks after sowing, and showed a positive correlation with the abundance of cyanobacteria and a negative correlation with the amount of N fertilizers used. Grain yield increased with increasing amounts of N fertilizers up to 70 kg N/ha. N rates appeared to affect grain yield by causing variations in the number of panicles/m2. Leaf chlorophyll readings at the end of the tillering stage were positively correlated with the number of panicles/m2, suggesting that it could be a useful parameter for predicting productivity. There was a significant increase in the N uptake of the rice but a decrease in the apparent N recovery and Nuse efficiency of applied fertilizer N, with the application of increasing rates of N fertilizer. In all instances, except in plots fertilized with 140 kg N/ha, the amount of N removed by plants was significantly higher than that applied as N fertilizer. The differences were positively correlated with the values for N fixation, suggesting a significant contribution by N fixation to rice production. These results show that a rational use of biological N fixation, in combination with inorganic N fertilization, would permit the input of N fertilizers to be reduced by c. 50% without any significant loss of productivity and with an ecological benefit for the whole ecosystem.