Physical processes not well resolved by climate models continue to limit confidence in detailed predictions of climate change. The representation of cloud and convection-related processes dominates the model spread in global climate sensitivity, and affects the simulation of important aspects of the present-day climate especially in the tropics. Uncertainty in aerosol radiative effects complicates the interpretation of climate changes in the observational and paleoclimate records, in particular limiting our ability to infer climate sensitivity. Dynamical uncertainties, notably those involving teleconnections and troposphere-stratosphere interaction, also affect simulation of regional climate change especially at high latitudes. In response, targeted field programs, new satellite capabilities, and new computational approaches are promoting progress on these problems. Advances include recognition of the likely importance of non-greenhouse gas forcings in driving recent trends in the general circulation, compensating interactions and emergent phenomena in aerosol-cloud-dynamical systems, and the climatic importance of cumulus entrainment. Continued progress will require, among other things, more integrative analysis of key processes across scales, recognizing the complexity at the local level but also the constraints and possible buffering operating at larger (system) scales.