Barnett et al. (1999) provide an excellent review of progress in model-based detection and attribution during the past few years. Unfortunately, the section on "Paleoclimate proxies" is flawed. We cannot agree that "OUl' best estimates of natural variability will come from CGCMs" or that "paleoproxy data have serious shortcomings that preclude their use for this purpose." On the contrary. we believe that paleoclimate data provide the on(y prospect for realistically estimating the envelope of natural climate variability and validating model simulations. Here we point out several issues that lead to erroneous conclusions in their paper.