Data from a 14-yr double-crop rice (Oryza sativa L.) experiment allowed comparison of the long-term effects of N fertilizer from different sources (urea and in situ grown azolla [Azolla microphylla Kaulf.] and sesbania [Sesbania rostrata Bremek. & Oberm.]) on N balances, soil N pools (both total and available), and yields. Although data show that plant-available N was maintained over time in both wet (WS) and dry seasons (DS), yields declined significantly, indicating a decline in physiological N use efficiency. The yield declines were generally similar regardless of N source in both seasons. The WS decline averaged 150 kg ha−1 yr−1 in the three added-N treatments, while the DS decline averaged 185 kg ha−1 yr−1. After 27 crops, the cumulative positive N balance was estimated at 1244, 348, 646, and 1039 kg N ha−1 in control, urea, sesbania, and azolla treatments, respectively. There was no significant change in soil total N content in the control and urea treatments, whereas it increased to 344 to 541 kg after 27 crops in the sesbania and azolla treatments. Conservation of the soil N status and positive N balance, in spite of the high amounts of N removed through grain and straw, reflect the N contribution (13–46 kg ha−1 crop−1) from nonsymbiotic N2 fixation. In addition, sesbania and azolla were estimated to add ≈57 to 64 kg ha−1 crop−1 through symbiotic N2 fixation. These results demonstrate that in rice–rice cropping systems biological N fixation plays a vital role in replenishing the soil N pool. However, continuous application of green manure N (GM-N) did not increase crop N availability, perhaps because of the presence of a recalcitrant soil organic matter fraction. Residual effects on rice grain yield and N uptake were observed only with GM-N sources.