The importance of the representation of cloud in a general circulation model is investigated by utilizing four different parameterization schemes for layer cloud in a low-resolution version of the general circulation model at the Hadley Centre for Climate Prediction and Research at the United Kingdom Meteorological Office. The performance of each version of the model in terms of cloud and radiation is assessed in relation to satellite data from the Earth Radiation Budget Experiment (ERBE). Schemes that include a prognostic cloud water variable show some improvement on those with relative humidity-dependent cloud, but all still show marked differences from the ERBE data. The sensitivity of each of the versions of the model to a doubling of atmospheric C02 is investigated. Midlevel and lower-level clouds decrease when cloud is dependent on relative humidity, and this constitutes a strong positive feedback. When interactive cloud water is included, however, this effect is almost entirely compensated for by a negative feedback from the change of phase of cloud water from ice to water. Additional negative feedbacks are found when interactive radiative properties of cloud are included and these lead to an overall negative cloud feedback. The global warming produced with the four models then ranges from 5.4° with a relative humidity scheme to 1.9°C with interactive cloud water and radiative properties. Improving the treatment of ice cloud based on observations increases the model's sensitivity slightly to 2.1°C. Using an energy balance model, it is estimated that the climate sensitivity using the relative humidity scheme along with the negative feedback from cloud radiative properties would be 2.8°C. Thus, 2.8°–2.1°C appears to be a better estimate of the range of equilibrium response to a doubling Of C02.