A case study of the Andreas et al. (1984) data on atmospheric boundary layer modification in the marginal ice zone is made. Our model is a two-dimensional, multilevel, linear model with turbulence, lateral and vertical advection, and radiation. Good agreement between observed and modeled temperature cross sections is obtained. In contrast to the hypothesis of Andreas et al., we find the air flow is stable to secondary circulations. Adiabatic lifting and, at long fetches, cloud top longware cooling, not an air-to-surface heat flux, dominate the cooling of the boundary layer. The accumulation with fetch over the ice of changes in the surface wind field is shown to have a large effect on estimates of the surface wind stress. We speculate that the Andreas et al. estimates of the drag coefficient over the compact sea ice are too high.