The characteristics of the Jarger and hotter natural hydrothermal systems are discussed, and several different models of heat source and heat-transfer mechanism reviewed. It is shown that, with models depending on mixed conductive-convective heat transfer from a magma chamber, the observed heat flows reauire either a contact area of several hundred square kilometres for the conductive transfer link, or very small separations between a chamber of circulating magma and the convective hot water system. None of the models is entirely satisfactory, and the possibility of heat transfer by magmatic steam is reconsidered in the light of the new isotope evidence, which indicates that most of the water discharged is of surface origin. It is suggested that it would be possible for this surface water to penetrate to and become absorbed by heated rock or magma at great depths and pressures, and for it to be ejected from the rock by subsequent heating, or to escape under reduced pressure to provide a very efficient high-temperature heat-transfer medium that would mix with ground water near the surface.