The net strength of the sea ice and surface albedo feedbacks in simulations of climate change has been estimated for several general circulation models (GCMs) using a variety of methods. Here several methods are applied to quantify this feedback in a single CO2-doubling experiment with the United Kingdom Meteorological Office GCM. Different methods give values differing by up to a factor of 2 and are shown to be estimatiang different quantities. An unbiased comparison of models is made using consistent methods, and an attempt is made to explain the differences in model sensitivity in terms of control climate and parameterization. Comparing the standard experiment to a parallel experiment using prescribed sea ice extents allows examination of the feedback as it actually operates in the model. The importance of cloud is emphasized, both in shielding surface albedo changes and so reducing their effect, and in providing it own contribution to planetary albedo changes. Statistical methods are used to examine the relationship between surface heat balance and albedo parameterization, and planetary albedo changes, which incorporate the effects of clouds and determine the net feedback. Estimates of the magnitude of the feedback in various GCMs vary from 0.16 to 0.7 m-2 K-1, with the highest value exaggerated by the method of analysis. We conclude that care is needed in quantifying these feedbacks and that previously published numbers are not all comparable. The feedbacks in two of the GCMs are exaggerated by unrealistically extensive sea ice, but the smaller values given by other GCMs may still be unrealistic, because of uncertainties in the treatment of clouds. ¿ American Geophysical Union 1989