Yield penalty and yield stability of aerobic rice have to be considered before promoting this water-saving technology in the tropics. The objectives of this study were 1) to compare crop performance between aerobic and flooded rice continuously over several seasons, and 2) to identify yield attributes responsible for the yield gap between aerobic and flooded rice. Field experiments were conducted at the International Rice Research Institute (IRRI) farm in dry and wet seasons. Grain yield and its attributes were compared between aerobic and flooded rice continuously for eight seasons from 2001 to 2004 using the best available aerobic rice varieties in the tropics. The yield difference between aerobic and flooded rice ranged from 8 to 69% depending on the number of seasons that aerobic rice has been continuously grown, dry and wet seasons, and varieties. When the first season aerobic rice was compared with flooded rice, the yield difference was 8-21%. The yield difference between aerobic and flooded rice was attributed more to difference in biomass production than to harvest index. Among the yield components, sink size (spikelets per m2) contributed more to the yield gap between aerobic and flooded rice than grain filling percentage and 1000-grain weight. Yield decline was observed when aerobic rice was continuously grown and the decline was greater in the dry season than in the wet season. The yield decline of aerobic rice was attributed more to changes in biomass production than in harvest index. Our data suggest that new aerobic rice varieties with minimum yield gap compared with flooded rice and crop management strategies that can reverse the yield decline of continuous aerobic rice have to be developed before aerobic rice technology can be adopted in large areas in the tropics.