Abstract: Rice yield components including productive tillers per hill, average yield per panicle and yield per hill under different fertilization systems were researched into in a long-term nutrient recycling experiment at the Changshu National Agroecological Experimental Station (31°33′ N, 123°38′ E). The indices for yield-related traits were collected in a single rice growing season and regression analysis used to determine the quantitative relationship between nutrient content in the rice plants at booting stage and the yield components and their stability. Results of the study show that nitrogen fertilizer favors productive tiller number increase, but dramatically inhibits per panicle average yield. Average productive tillers per hill under N treatment are 4.2, 0.9 and 0.7 higher than under CK, NP and NPK respectively. The application of nitrogen along with phosphorus eliminates the conflict between productive tillers and per panicle average yield and therefore boosts per hill yield of rice. Productive tillers and per hill yield of rice increase with the addition of pig manure, but per panicle average yield does not significantly change. Only the stability of per hill productive tillers is significantly affected by chemical fertilizer treatments. The other two yield components are not affected by chemical fertilizer or pig manure treatments. Average coefficient of variation of productive tillers under CK, N, NP and NPK are 19.5%, 17.9%, 18.3% and 22.9% respectively. Thus potassium fertilization increases productive tillers CV. There is a good fit of the linear regression model for yield-related trait indices and N and K contents of plants at booting stage. The average CV of productive tillers is directly related with K content of rice plant.