Abstract: A global climate model (GCM) should be able to reproduce features of the distribution of the regional to local-scale climate in which it is applied. Such features include: the climatological mean, correlation, monthly or daily variance, thresholds, extremes etc, of the distribution of climate variables of interest. Most researchers need to know how GCM simulations vary depending on climatic variables, the choice of GCM and place. These variations can be understood by studying the descriptive statistics above, and inference can be made based on these sample statistics. However, there is no standard approach to test the features above in order to determine the skill of GCMs. In this paper, we focus on correlation and regression to evaluate the performances of five coupled global climate models for simulating monthly rainfall, minimum and maximum temperatures at five stations in northeastern Zimbabwe. We use observed historic climatic data (rainfall and air temperature) as well as downscaled model predictions of the same parameters. The global climate models used were the same as those used by the Intergovernmental Panel on Climate Change (IPCC) in formulating the IPCC Special Report on Emissions Scenarios (SRES). The GCMs were evaluated by comparing observed historic climatic data with hindcast downscaled model predictions.