Cloud feedbacks in a warmer climate have not yet been constrained by models or observations. We present an approach that combines a general circulation model (GCM), single-column model (SCM), satellite and surface remote sensing data, and analysis product to infer regional cloud feedbacks and evaluate model simulations of them. The Atmospheric Radiation Measurement (ARM) Program Southern Great Plains (SGP) continuous forcing product, derived from a mesoscale analysis constrained by top-of-atmosphere and surface data, provides long-term advective forcing that links the models to the data. We drive an SCM with the continuous forcing for 10 cold season months in which synoptic forcing dominates the meteorology. Cloud feedbacks in midlatitude winter are primarily responses to changes in dynamical forcing. Thus we select times when observed advective forcing anomalies resemble doubled CO2 advective forcing changes in the parent GCM. For these times we construct cloud type anomaly histograms in the International Satellite Cloud Climatology Project and Active Remotely Sensed Cloud Locations data sets and simulated versions of these histograms in the SCM. Comparison of the SCM subset to GCM doubled CO2 cloud type changes tells us how relevant the selected times are to the GCM's cloud feedbacks, while comparisons of the SCM to the data tell us how well the model performs in these situations. The data suggest that in midlatitude winter, high thick clouds should increase while cirrus and low clouds decrease in upwelling regimes in a climate warming. Downwelling regime cloud feedbacks are dominated by changes in low clouds but are not as well constrained by the data.