Based on the grain yield data of spring wheat from the 2002-2005 regional trials with 13 cultivars (lines) on the rainfed farmlands at 18 locations of China, and by using variance analysis and GGE (genotype main effect plus genotype-by-environment interaction) biplot, this paper studied the effects of genotype, environment, and genotype x environment interaction (GEI) on the yield variation of spring wheat and the yield stability of different cultivars (lines). Comparing with genotype and GEI, the environment had much more contribution on the yield variation, accounting for 87.5%-92.0% of the total. In the GEI items, genotype x location had the greatest effect, whereas genotype x year had the smallest one. The average yield of modern spring wheat genotypes across years and locations in China was 2550 kg x hm(-2). Among the three yield components, thousand kernel weight (TKW) had the least variation under environmental change. The main environment factors affecting the yield variation were annual accumulated temperature ( > or =10 degrees C), precipitation in growth season, mean air temperature, altitude, annual precipitation, and frost-free period. The grain yield had significant positive correlations with spike number per unit area (SPU) (r = 0.675**), TKW (r = 0.456**), and kernels per spike (KPS) (r = 0.581**), and the SPU, KPS and TKW also had positive correlations (r = 0.244-0.480**) with each other, implying that these three yield components might be increased or improved synchronously.