An efficient low-input system (LIS) for fertiliser use in rice cultivation is necessary to reduce fossil energy use and pollution. Japanese people like Japonica rice, especially cv. Koshihikari. However, it has very low lodging resistance in Japanese weather condition. Our objective was to develop a LIS with the minimum sacrifice of grain yield in rice. Koshihikari was grown using conventional fertilization as a control (CON) with 4 g N $m^{-2}$., 8g $P_2$ $O_{5}$ $m^{-2}$ and 8 g $K_2$O $m^{-2}$ as a basal fertilizer dressing. It was compared with a low fertilizer treatment (LF) with only 4 g $P_2$ $O_{5}$ $m^{-2}$ as a basal dressing in the first year and no basal phosphorus fertilizer in the second year. Chopped rice straw was incorporated into the soil before the cropping season in both years. Fertilizer of 4 g N $m^{-2}$ was top-dressed at 15 days before heading in CON plots and 30 days before heading in LF plots in both years. Lodging was significantly less in LF than in CON plots, however, no rice straw effect was found in low fertilized condition. Grain yields in LF plots were reduced by 15-16% below those of CON plots. Lower yields in LF plots were associated with a reduced number of spikelets per unit area. However, big spikelet size was acquired in LF by 10 days earlier N top dressing than CON plots. A close relationship was found between spikelet numbers and N content of the plant at heading, and between grain yield or shoot dry weight and N content of the plant at maturity. Regardless of the fertilizer application methods, N use efficiency for the number of spikelets, final total dry matter and grain yield was essentially identical among fertilizer treatments. The reduced growth and yield in the LF plots resulted from low absorption of nitrogen. Conclusively, LIS can drastically reduce chemical fertilizer use and facilitate harvest operations by reducing lodging with some yield reduction..