Climate model experiments incorporating model runs conducted using different conditions have become popular for the study of uncertainty affecting model output and projections of climate change. Recently, such experiments have been used to also study of the uncertainties in producing high-resolution projections of climate change based on methods for dynamic downscaling of global models. In this paper, we discuss an initial analysis of a subset of the ensemble being produced by the North American Regional Climate Change Assessment Program. Using an approach based on a functional analysis of variance, we examine the differences between two different dynamic downscaling methods and demonstrate that there are significant differences between the two models and their projections of summer temperature and precipitation over North America. Copyright © 2010 John Wiley & Sons, Ltd.