Coupled General Circulation models (AOGCMs) are widely used as an important tool of projecting global climate change. However, their resolution is too coarse to provide the regional scale information required for regional impact assessments. Therefore, downscaling methods for extracting regional scale information from output of AOGCMs have been developed. Regional climate models nested in AOGCMs, statistical downscaling, and dynamicalstatistical downscaling are usually used for downscaling. In this review paper, focus is placed on statistical downscaling techniques. These methods can be used to predict regional scale climate from AOGCM output using statistical relationship between the large-scale climate and the regional-scale climate, which offers the advantages of being computationally inexpensive. The principle and assumptions of three categories of statistical downscaling are introduced. Important issues in using statistical downscaling to create future climate change scenario is also discussed. At the same time, dynamical downscaling is briefly compared with statistical downscaling in terms of their advantages and disadvantages. Finally, prospects of developing new downscaling techniques by combining statistical and dynamical downscaling techniques are pointed out.