In an effort to evaluate the ability of atmospheric general circulation models to simulate observed monsoon variations we have analyzed many of the models participating in the Atmospheric Model Intercomparison Project, a coordinated experiment to simulate the period 1979-88. We have found that the ability of a model to correctly simulate interannual variations is regionally dependent and related to its ability to correctly respond to remote SST forcing and to portray a realistic climatology (Sperber and Palmer 1994). The use of a regional rainfall index as a means of model verification has proven to be a stringent test of a models ability to simulate interannual variations. With regard to the influence of the remote SST forcing, nearly half of the models evaluated exhibited fundamental difficulties by their failure to even realize the correct phase of the observed all-India rainfall/SST teleconnection. Using such teleconnections provides an objective means of stratifying model performance. We find that the link between Indian monsoon rainfall and SST is strongest under ENSO conditions, particularly when substantial anomalies in the tropical Pacific Ocean persist during June-September. At other times little or no consensus among the simulations exists with regard to Indian monsoon rainfall, even in the initial condition sensitivity simulations performed with the ECMWF model. Contrary to this, the simulations of Nordeste rainfall variations are more coherent owing to fact that their dominant controlling factor is tropical SSTs. Also, examination of the ECMWF initial condition sensitivity integrations indicates that the simulation Nordeste rainfall variations is much less sensitive to the specification of initial conditions.