The composition dependence of the surface tension of highly nonideal organic-organic and aqueous-organic nonelectrolyte solutions is described, based on the assumption that the surface layer can be treated as a separate phase located between vapor and bulk liquid phases. The Wilson, NRTL, and UNIFAC methods are used for activity coefficients of surface and bulk phases and three techniques for calculation of molar surface areas, based on Paquette areas, Rasmussen areas, and a Langmuir-type approach are tested. Comparisons of the calculated surface tensions with experimental data yield mean absolute errors, in the best case, of less than 2.5% for the systems studied, all of which exhibit highly nonideal behavior. The surface tension predictions are found to be extremely sensitive to the values of the molar surface areas used in the computation. A Langmuir-type adsorption model is formulated to determine the surface mole fractions from a knowledge of the mixture surface tension as a function of bulk composition. A novel procedure is developed to obtain the partial molar surface area of the larger organic component as a function of composition in binary aqueous-organic systems, assuming that the two components are very dissimilar in size, and that deviations in the partial molar surface area of the smaller component (water) from its pure component molar surface area contribute negligibly to the total molar surface area of the mixture. This removes the approximation of equality of partial and pure component molar surface area for the larger organic component. Use of the Langmuir-type approach with partial molar surface areas improves surface tension predictions of highly nonideal aqueous-organic mixtures by 20% over use of pure component molar surface areas. It is an important first step in the development of a thermodynamically consistent theory of surfaces for liquid mixtures based on an accurate determination of the composition dependence of partial molar surface areas for all components. Copyright 1999 Academic Press.