20th and 21st century decadal scale trends and variability in winter North American snow cover extent (NA-SCE) are investigated using coupled atmosphere-ocean general circulation model experiments participating in the upcoming Intergovernmental Panel on Climate Change Fourth Assessment Report. Significant between-model variability is found, with most models underestimating mean NA-SCE. 20th century simulations are poorly correlated with observations, and, while individual ensemble members capture the magnitude of decadal scale variability, the variability of the signal is dampened in the ensemble mean, indicating that decadal-scale NA-SCE variability is associated predominantly with internal model variability rather than external forcing. Two 21st century emission scenarios with realistic (moderate or significant) greenhouse gas emission rates produce decreasing NA-SCE trends, while one unrealistic scenario with fixed concentrations produces little or no NA-SCE trend. These results suggest that snow cover may be a sensitive indicator of climate change, and that North American snow extent will probably decrease in response to greenhouse gas emissions, although the magnitude of the response may be nonlinear.