Output from four experiments with the Goddard Institute for Space Studies climate model, simulating climates ranging from ice age conditions to Mesozoic conditions, are used to test whether simple parameterizations of infrared (IR) emissions to space can be applied to annual average climate changes. The possible predictors of IR emissions examined are zonally averaged surface temperature, fractional cloud over, average cloud top height, tropospheric lapse rate, and atmospheric moisture content. First, parameterizations are derived from latitudinal variations simulated by the model for the current climate. The simplest such parameterization, with IR emissions being a linear function of surface temperature alone, predicts the simulated climate changes in IR emissions, with errors of about 35%. Inclusion of additional regression terms derived from latitudinal variations in the current climate does not improve the prediction of changes in IR emissions. However, by deriving regressions from one climate change and applying them to the other changes, we demonstrate that adding a term describing lapse rate changes reduces the error in the predicted IR flux changes to between 25 and 30%. The lapse rate changes approximately follow the change in moist adiabatic lapse rate in the topics and the change in the critical lapse rate for baroclinic adjustment in middle latitudes, unless the climate change is so large as to cause a qualitative change in the dynamical regimes. Only a very slight further reduction in the remaining error in the predicted changes in IR emissions was achieved by including an additional predictor, and there was no clear-cut best choice for the additional predictor.