Cloud amount records for the U.S.A. have been analyzed in the context of the 'warming world' analogue model described by Lough et al. (1983). Cloud amount increases over practically the entire U.S.A. in all seasons. This result considerably strengthens the more tentative conclusion of Henderson-Sellers (1986) that cloud amount increases over Europe in the same warming world scenario. These results are in contrast to the few numerical model predictions of cloud changes in warming world experiments. A possible, rather tantalizing, conclusion is that current GCM cloud prediction schemes tend to enhance temperature increases through cloud-climate feedback whereas the historical data could suggest a negative feedback. Part, possibly all, of this difference may be the result of the fundamental distinction between the two experimental scenarios: the equilibrium change modelled by GCMs as compared to the smaller transient change represented by the historical analogue. On the other hand the current 'real-world' experiment is, itself, a transient change in boundary and atmospheric conditions. At the least, surface-observed cloudiness seems to offer a useful and complementary data source with which to examine one aspect of the performance of numerical climate models.