The dependence on horizontal resolution of the climate simulated by the National Center for Atmospheric Research Community Climate Model (CCM2) is explored. Simulations employing R15, T21, T31, T42, T63, and T106 horizontal spectral truncations are compared. Parameters associated with the diagnostic cloud scheme are modified for each resolution to provide similar global average cloud radiative forcing at each resolution. Overall, as with earlier studies, there are large differences between the low resolution R15 and T21 simulations and the medium resolution T42 simulation. Many climate statistics show a monotonic signal with increasing resolution, with the largest variation occurring from low to medium resolution. Although the monotonic signal is often from the low resolution simulations toward atmospheric analyses, in some cases it continues beyond the analyses at the highest resolution. Where convergence occurs, it is not always to the atmospheric analyses, and the highest resolution simulations are not the best by all measures. Although many climate statistics converge, the processes that maintain the climate do not, especially when considered on a regional basis. The implication is that the finer scales are required to capture the nonlinear processes that force the medium scales. Overall, it appears that, at a minimum, T42 resolution is required, but higher resolution would be better. Applications at T42 should take into consideration how model errors indicated by these resolution signals might affect any findings.