In this paper, the interannual variability simulated by the coupled ocean-atmosphere general circulation modelof the institute of Atmospheric Physics (IAP CGCM) in 40 year integrations is analyzed, and compared with that bythe corresponding IAP AGCM which uses the climatic sea surface temperature as the boundary condition in 25 yearintegrations.The mean climatic states of January and July simulated by IAP CGCM are in good agreement with that by IAPAGCM, i.e., no serious 'climate drift' occurs in the CGCM simulation. A comparison of the results from AGCM andCGCM indicates that the standard deviation of the monthly averaged sea level pressure simulated by IAP CGCM ismuch greater than that by IAP AGCM in tropical region. In addition, both Southern Oscillation (SO) and NorthAtlantic Oscillation (NAO) can be found in the CGCM simulation for January, but these two oscillations do not existin the AGCM simulation.The interannual variability of climate may be classified into two typest one is the variation of the annual mean,another is the variation of the annual amplitude. The ocean-atmosphere interaction mainly increases the first type ofvariability. By means of the rotated EOF, the most important patterns corresponding to the two types of interannualvariability are found to have different spatial and temporal characteristics.