The mean global distributions of pressure, temperature, wind, moisture, cloudiness, precipitation, evaporation, and surface heat balance simulated for January by the two-level Mintz-Arakawa atmospheric general circulation model are compared with the corresponding observed fields. Although there are a number of shortcomings, in general the large-scale distribution of global climate is reasonably well portrayed by the model, in spite of its limited vertical resolution. The model simulates the semi-permanent cyclones and anticyclones of both the tropics and higher latitudes in approximately their correct positions, together with the associated large-scale temperature and circulation fields of the middle and lower troposphere. In comparison with models of greater resolution, these results suggest that with further selective improvements in the physical parameterizations, relatively coarse global models (of correspondingly lower computational demands) are useful tools in the study of many aspects of climate. The most prominent errors of the present model simulations are in the portrayal of processes related to the transfer of moisture. The simulated cloudiness is about half that observed (in the Northern Hemisphere), and the average precipitation rate is about twice that observed and extends over too broad a zone in the tropics. The cloudiness error is evidently due to the model's production of clouds only during precipitation, and its failure to simulate nonprecipitating cloudiness at all. The precipitation error is due to an apparent simulation of excessive convective rainfall and has noticeably affected the heat and moisture balances in the tropics. Also in the tropics, the simulated January surface air temperature is higher than the specified sea-surface temperature, and has resulted in a net downward sensible heat flux at the surface between about 20N and 20S in contrast with observation. The evaporation, which occurs almost exclusively over the tropical oceans, is simulated to be about 50% too great. These errors evidently compensate each other in the net surface heat balance everywhere except at high southern latitudes, where low amounts of simulated cloudiness permit excessive surface insulation. In the troposphere, the simulated zonally averaged January 400 mb temperature is approximately 5°C above that observed in the equatorial and tropical regions; at 800 mb the simulated temperature more closely resembles observation. The meridional gradients of geopotential height in mid-latitudes at both 400 and 800 mb are somewhat steeper than those observed, as is the meridional gradient of the 400 mb zonal mean temperature. The associated maximum zonal winds at 400 and 800 mb are about 60% stronger than the observed winds, at least in the Northern Hemisphere. At 800 mb the simulated relative humidity is approximately half that observed between about 30N and 20S, while at higher latitudes it exceeds observation by an average of about 15%. These errors way be related to the model's tendency to simulate too great a strength for the quasi-stationary oceanic cyclones of middle and higher latitudes, and too small an intensity for the individual transient waves. The associated midlatitude Ferrel cell in the mean meridional circulation is therefore both too weak and too narrow. The subtropical oceanic anticyclones are more realistically simulated, and at least in the Northern (winter) Hemisphere, and the strength of the associated Hadley circulation resembles that derived from observation. The simulation errors noted here furnish a guide for the continuing modification and improvement of the model, and new integrations over longer time periods with improved boundary conditions are in preparation. Such experiments, will permit the systematic determination of the characteristic or natural variability of the model's simulated climates, which is of critical importance in the model's use for the study of climatic change.