The idea of escape of light atmospheric gas is older than the kinetic theory, having been introduced by J. L. Waterston in 1846 (Chamberlain 1963). The thermal escape concept was revived by G. J. Stoney around the turn of the century and . given a definitive form by Jeans (1925), whose name is usually associated with it. Spitzer (1952) was the first to realistically incorporate atmospheric structure into the problem. Observation in 1955 of Lyman-IX scattering from the geocorona, the atomic-hydrogen cloud around the Earth, and many more recent satellite and space-probe measurements have led to a number of interpretive and theoretical papers, which Chamberlain ( 1963) and Tinsley (\974) have reviewed. Composition and structure of other atmospheres are discussed by H unten ( 1971 b) and Ingersoll & Leovy ( l97\). Escape from a planet is intimately linked to the nature of the exosphere, the region in which the mean free path is so long that collisions can be neglected for many purposes. Most atoms in the exosphere are in free ballistic orbits; a few of the fastest ones may be on escape orbits. The bottom of the exosphere is the critical level, or exobase, defined as where the mean free path (in the horizontal direction) is equal to the scale height. The conventional approximation of the critical level is that the atmosphere is fully collisional below it and collisionless above; Chamberlain (1963) discusses the validity of this approximation. The planetary corona can be defined as the hydrogen component of the exosphere. Global averages are used throughout most of the present review. The exosphere is actually asymmetric to a considerable degree, as are the escape processes. Neverthe less, the long mean free path permits a great deal of lateral flow, which helps to justify the approximation of spherical symmetry (McAfee 1967; Vidal-Madjar & Bertaux 1972; Tinsley 1974; Tinsley, Hodges & Strobel 1975). Atmospheric escape is an interesting phenomenon, particularly for its effect on