The qualitative agreement of two climate models, HADCM2 and ECHAM3, on the response of surface climate to anthropogenic climate forcing in the period 2020-2049 is studied. Special attention is paid to the role of internal climate variability as a source of intermodel disagreement. After illustrating the methods in an intermodel comparison of simulated changes in June-August mean precipitation, some global statistics are presented. Excluding surface air temperature, the four-season mean proportion of areas in which the two models agree on the sign of the climatic response is only 53-60% both for increases in CO2 alone and for increases in CO2 together with direct radiative forcing by sulphate aerosols, but somewhat larger, 59-70% for the separate aerosol effect. In areas where the response is strong (at least twice the standard error associated with internal variability) in both models, the agreement is better and the contrast between the different forcings becomes more marked. The proportion of agreement in such areas is 57-75% for the response to increases in CO2 alone, 64-84% for the response to combined CO2 and aerosol forcing, and as high as 88-94% for the separate aerosol effect. The relatively good intermodel agreement for aerosol-induced climate changes is suggested to be associated with the uneven horizontal distribution of aerosol forcing.