Based on nitrogen (N) balance in soil system, a new N fertilization recommendation method, i.e. “N return index (NRI)” method, was introduced. The NRI concept was brought forth by the Nutrient Return Doctrines, and means how much N must be returned to soil in order to maintain soil N balance when one kilogram N is taken off by crop harvesting. On the assumption that other than N fertilization, the N inflow via atmosphere deposition, irrigation and seeding (N biofixation and N in straw incorporation exclusive) and the N outflow via erosion, runoff, leaching, volatilization, and nitrification-denitrification, are constant, an equation for calculation of NRI (NRI=1−⊿E/F_N) was deduced, where ⊿E stands for increase in N output from soil to environment induced by N application; F_N for N input with fertilizer applied, and ⊿E/F_N is net N loss. With the target yield (Y_t) of the fields most farmers are familiar with set as the only variable to distinguish fields different in soil fertility, an equation for NRI-based recommendation of N fertilization was worked out, i.e. RNR= RNRI×N₁₀₀×Y_t/100”, where RNR stands for recycled N rate, RNRI for the regional NRI, which is a relatively stable constant in a specific farming area, N₁₀₀ for N uptake by one hundred kilogram grains, which is in a significant linear relationship with crop yield. By taking into account the actual production practices, the equation was finally amended as "RNR=RNRI×N₁₀₀×Y_t/100−W_N", where W_N stands for N input via biofixation and straw incorporation. To verify effectiveness of NRI, two regions in Lixiahe Plain and Taihu Lake Basin in Jiangsu Province were selected as for the study. Based on the soil N budgeting data of the study areas during the rice growing season, RNRI was worked out to be 1.52, based on traits of rice cultivars, N₁₀₀ was 8.67×10 ^-5Y_t+1.161 (r=0.828 0**) and based on N biofization and straw incorporation, WN was N 52.3 kg hm ^-2, then the equation of “RNR=1.32×10 ^-6Y_t² 1.76×10 ^-2Y_t−52.3” for recommendation of N fertilization for rice in the study areas was figured out. Based on the data of the fertilization experiment on 45 cultivars of japonica rice in recent years, comparison was performed between N fertilization recommendations using different methods, the fertilizer effective function (FEF), NRI and theoretical N rate (TNR) in effectiveness. Results show that the optimal economic N application rate (OENR) recommended by FEF, the N-recycling N application rate (RNR) by NRI and the theoretic N application rate (TNR) by TNR was N 246.8±42.5 kg hm ^-2, 216.9±27.3 kg hm ^-2 and 176.9±22.3 kg hm ^-2 respectively. Compared with OENR, RNR was 2.4% less in rice grain yield and 500 yuan hm ^-2 in economic benefits, but saved N 29.9 kg hm ^-2 and 10.3 kg hm ^-2, respectively, from N application and soil-N net loss and moreover kept soil-N in a rough balance. The new method can be used to make recommendation for N fertilization with target yield being the only variable and without the need to go through field experiment, however, further study is still required to see whether RNRI will remain relatively stable for long, and whether long-term use of the method would lead to soil-N surplus or not.