The inclusion of cloud microphysical processes in general circulation models makes it possible to study the multiple interactions among clouds, the hydrological cycle, and radiation. The gaps between the temporal and spatial scales at which such cloud microphysical processes work and those at which general circulation models presently function force climate modelers to crudely parameterize and simplify the various interactions among the different water species (namely, water vapor, cloud water, cloud ice, rain, and snow) and to use adjustable parameters to which large-scale models can be highly sensitive. Accordingly, the authors have investigated the sensitivity of the climate, simulated with the Colorado State University general circulation model, to various aspects of the parameterization of cloud microphysical processes and its interactions with the cumulus convection and radiative transfer parameterizations. The results of 120-day sensitivity experiments corresponding to perpetual January conditions have been compared with those of a control simulation in order to 1 ) determine the importance of advecting cloud water, cloud ice, rain, and snow at the temporal and spatial scale resolutions presently used in the model; 2) study the importance of the formation of extended stratiform anvils at the tops of cumulus towers, 3) analyze the role of mixed-phase clouds in determining the partitioning among cloud water, cloud ice, rain, and snow and, hence, their impacts on the simulated cloud optical properties; 4) evaluate the sensitivity of the atmospheric moisture budget and precipitation rates to a change in the fall velocities of rain and snow; 5) determine the model's sensitivity to the prescribed thresholds of autoconversion of cloud water to rain and cloud ice to snow; and 6) study the impact of the collection of supercooled cloud water by snow, as well as accounting for the cloud optical properties of snow. Results are presented in terms of 30-day mean differences between the sensitivity experiments and control run. The authors find that three-dimensional advection of the water species has little influence on their geographical distributions and globally averaged amounts. The simulated climate remains unchanged when detrained condensed water at the tops of cumulus towers is used as a source of rain and snow rather than as a source of cloud water and cloud ice. In contrast, instantaneously removing cloud water and cloud ice detrained at the tops of cumulus towers in the form of precipitation yields a strong drying of the atmosphere and a significant reduction in the size of the anvils. Altering the partitioning between cloud ice and supercooled cloud water produces significant changes in the vertical distributions of the cloud optical depth and effective cloud fraction, hence producing significant variations in the top-of-the-atmosphere longwave and shortwave cloud radiative forcings. Increasing the fall speeds of rain and snow leads to a decrease in cloudiness and an increase in stratiform rainfall. Increasing the thresholds for autoconversion of cloud water to rain and cloud ice to snow yields a significant increase in middle- and high-level clouds and a reduction of the cumulus precipitation rate. The collection of supercooled cloud water by snow appeared to be an important microphysical process for mixed-phase clouds. Finally, the optical effects of snow have little impact upon the top-of-the-atmosphere radiation budget. This study illustrates the need for in-depth analysis of the spatial and temporal scale dependence of the different microphysical parameters of the cloud parameterizations used in general circulation models.