Climate models have a central place in discussions of future climate change and in the attribution of late 20th Century changes mainly to anthropogenic changes in atmospheric composition. Yet most climate scientists do not work directly on these models and many are naturally uneasy at treating the models as black boxes. The scales involved in climate change range from molecular effects to planetary waves and uncertainties in many aspects of the basic physics from clouds to small-scale ocean turbulence imply that these models can only ever approximate the real world. Given those approximations, what can be taken from model results? How can robustness be judged? What confidence can be placed in projections? I will try and address those questions using model results and experiments for a wide variety of cases from present-day and paleo-climate changes.