IT is implicit in the generally accepted theory of the heat balance on clear nights1-3 that the cooling of the air near the ground is a direct consequence of heat loss to the surface by convection and conduction; ultimately the latter. However, Ramdas and Atmanathan4 have shown that, in India, under very stable atmospheric conditions, the lowest air temperature is not reached at the surface but at some height above it. This observation is inconsistent with the classical theory. The phenomenon has been observed by workers in other places, notably Brawand and Kohnke5 in Indiana, and Jenny6 in Lausanne ; but apparently its theoretical implications have not been appreciated. Ramanathan and Ramdas7, and more recently Rider and Robinson8, have suggested that under stable conditions certain air layers near the ground may be in radiative rather than conductive equilibrium with their surroundings, the eddy conductivity approaching the molecular value. Sutton3 and others have stated that the part of the atmosphere mainly concerned in nocturnal radiation exchange is a relatively thin layer near the ground. Now if this layer is considered to be at the same initial temperature as the surface it must be a net radiative loser of heat, and will therefore tend to cool to a temperature below that of the surface; a process more or less retarded by heat gained by convection and conduction from the ground and from the upper air layers. On the basis of this theory, the temperature reached by the surface continues to play an important part in determining that reached by the air near it, but for reasons quite different from those given in the classical theory.