The linear instability of two zonal mean flows, one computed by a general circulation model and the other corresponding to the observed winter zonal mean flow, is presented. For these calculations, we utilize a numerical primitive equation model, where the spherical geometry of the earth has been retained. By comparing the waves predicted by linear theory with the eddies that appear in the general circulation model, it is determined that significant discrepancies exist. For the wavenumber range 1 through 15, the linear theory predicts the maximum growth rate to be for wavenumbers 12-15. The wavenumbers that dominate the intermediate-scale transient eddies in the general circulation model are much longer (5–7). In addition, linear theory predicts the maximum amplitude of the geopotential perturbation for wavenumbers 5–7 to be near the earth's surface, while in the general circulation model, the maximum amplitude of this quantity for wavenumbers 5–7 is at the tropopause level. Also, the phase speed of wavenumbers 7–9 in the general circulation model is considerably faster. that it is for the corresponding waves predicted by linear theory. It is determined that these discrepancies also exist for wavenumbers 7–15 in the real atmosphere. It is concluded that these discrepancies must be due to some nonlinear process.