The brines in the Atlantis II Deep of the Red Sea occur in horizontally uniform, well-mixed layers, with the hottest and saltiest water at the bottom, separated from the successively cooler and fresher layers above by very sharp vertical temperature and salinity gradients. Data accumulated over 3 decades are used to test the widely accepted hypothesis that all heat and salt for the brine layers are supplied from below and that the layered brine structure is the result of double diffusion. Using the changes in temperature and salinity in each layer over successive time intervals, one can deduce the corresponding fluxes of heat and salt across the interfaces. It is found that the required flux of salt cannot be sustained by double diffusion alone. An alternative calculation shows that most of the salt in the successively forming upper layers must have been injected directly from the bottom of the deep through one or more vents located above the level of the lowest brine interface. For the bottom layer, however, it is not possible to obtain the observed salinity and temperature changes unless hot saline water is injected directly into that layer and some of the heat and a smaller fraction of the salt are transferred upward through the interface. This process will also maintain convection in each of the layers and keep them well mixed, as is observed. The new interpretation in terms of separate inputs at various levels in the Atlantis II Deep is also supported by recent geochemical evidence. ¿ 1998 American Geophysical Union