Description: The spatial distribution of the subtropical salinity maximum is identified using historical and recent data from the eastern North Atlantic. In the regions with high frequency of occurrence of the salinity maximum, the relative contributions of advection, eddy diffusion and double diffusion to the salt balance below the maximum salinity layer are determined. McDougall's (1984, Journal of Physical Oceanography, 14, 1577–1589) salt balance equation for neutral surfaces is used in this analysis. The data base consists of two meridional CTD sections along 33° and 27°W between 10° and 35°N, mean temperature-salinity profiles in 5° × 5° squares presented by Emery and Dewar (1982), and mean velocity profiles in 3° × 3° squares evaluated by Stramma (1984, Journal of Marine Research, 42, 537–558). The tropical salinity maximum tongue is found to be quite persistent in its salinity value and its geographic distribution, but less clearly in its vertical or isopycnal position. Double diffusion due to salt-fingering appears to be an important process for the salt balance below the salinity maximum layer. An approximate estimate of the double-diffusive salt flux is obtained. Near the subtropical source region, the double-diffusive salt flux is balanced primarily by isopycnal advection; further to the south it is also balanced by isopycnal eddy diffusion. Maximum double-diffusive fluxes correspond in magnitude to the mean salt flux caused by the excess in evaporation at the surface in the central subtropics. The resulting isopycnal and diapycnal eddy-mixing coefficients derived by a linear inversion technique have the reasonable values of Ki = (11 ± 5) × 102 m2 s−1 and Kd = (4 ± 2) × 10−5 m2 s−1. Considering the intermittency of the double-diffusive process, limiting values for the mean eddy-mixing coefficients are determined by neglecting the contribution of the double-diffusive salt fluxes. This leads to Ki = (5 ± 2) × 102 m2 s−1 and Kd = (5 ± 1) × 10−5 m2 s−1 for the isopycnal and diapycnal mixing coefficients, respectively.