It is postulated that ball lightning is initiated by a lightning stroke that forms a large sphere of heated material. The following models are considered: (1) cooling spheres of air;(2) cooling spheres of air containing small amounts of sodium vapor; (3) cooling spheres of mixtures that are by weight (a) % carbon vapor and % air or (b) s• copper vapor and •i air. For each model calculations are made of the temperature profiles, output radiation, and average mass density as a function of time, taking account of energy transfer due to conduction, radial convection, and emission and absorption of radiation. Models I and 2 are deficient in explaining the properties of ball lightning in that the spheres are lighter than the surrounding air and thus rise. With model 3, altkough the average mass density •can approximate that of air so that the ball does not rise, there is insufficient emission of visible radiation. It is possible, however, that chemical processes, which are not considered in the present calculations, occur at the boundary of the sphere of suspended particles and the surrounding air and give relatively constant emission of light. dius would approximately equal the light output of a 4-watt incandescent bulb), an RF field strength of at least 0.1 volt/cm would be required for a time of several seconds. The value of 0.1 volt/cm was obtained by assuming that all the energy of the incident RF wave is converted into visible radiation. To initiate a ball, fields of the order of 10 • volts/cm would be necessary. The RF fields observed by Kapitza [1968] during thunderstorms are many orders of magnitude less than the sustaining field, and we do not see any basis for believing that the requisite fields exist, particularly after the completion of a lightning stroke. In a more recent theory, Uman and Helstrom [1966], following an initial suggestion of Finkelstein and Rubinstein [1964], propose that ball lightning is powered by an external dc electric field, and quantitative calculations are made to show that in the absence of convection such a ball can be formed by a direct current. It is difficult, however, to see how any theory of ball lightning requiring an external power source, either from RF or from dc electric fields, can explain the existence of ball lightning inside structures.In the present paper we consider models for ball lightning in which no external energy is supplied to the ball. That is, the energy expended by the ball is contained within the ball. We investigate quantitatively the rates of decay of temperature and light output of cooling 6887