The ECHAM-1 T21/LSG coupled ocean-atmosphere general circulation model (GCM) is used to simulate climatic conditions at the last interglacial maximum (Eemian, 125 kyr BP). The results reflect the expected surface temperature changes (with respect to the control run) due to the amplification (reduction) of the seasonal cycle of insolation in the Northern (Southern) Hemisphere. A number of simulated features agree with previous results from atmospheric GCM simulations (e.g., intensified summer southwest monsoons) except in the Northern Hemisphere poleward of 308N, where dynamical feedbacks in the North Atlantic and North Pacific increase zonal temperatures about 18C above what would be predicted from simple energy balance considerations. As this is the same area where most of the terrestrial geological data originate, this result suggests that previous estimates of Eemian global average temperature might have been biased by sample distribution. This conclusion is supported by the fact that the estimated global temperature increase of only 0.3 8C greater than the control run has been previously shown to be consistent with CLIMAP sea surface temperature estimates. Although the Northern Hemisphere summer monsoon is intensified, globally averaged precipitation over land is within about 1% of the present, contravening some geological inferences but not the deep-sea d13C estimates of terrestrial carbon storage changes. Winter circulation changes in the northern Arabian Sea, driven by strong cooling on land, are as large as summer circulation changes that are the usual focus of interest, suggesting that interpreting variations in the Arabian Sea sedimentary record solely in terms of the summer monsoon response could sometimes lead to errors. A small monsoonal response over northern South America suggests that interglacial paleotrends in this region were not just due to El Nino variations.