The joint distribution of temperature and precipitation is a basic metric capable of shedding light on both the functioning of the Earth's climate system and.climate-model performance. Here, with the aid of high-quality observational data sets, we investigate the joint behavior of these variables as observed, and as simulated by the suite of coupled climate models comprising the CMIP3 dataset. The data are then filtered, and the frequency dependence of the joint probability density functions and time evolution are compared. Finally, an assessment of model performance is undertaken in the light of this fundamental distributional measure.