Recent spherical harmonic representations of the gravity field of Mars have sufficient resolution to examine the regional crustal and lithospheric structure of the Valles Marineris and are used as constraints for rift modeling. While gravity signatures of individual troughs are not evident, a broad 190--260 mgal low indicates that the central chasmata as a whole are compensated at a depth of 30--80 km, representing the thickness of crust surrounding the troughs. Furthermore, at the time large-scale relief was established, the effective thickness of the elastic lithosphere was 20 mW m-2. The calculated range of crustal thickness and heat flow can be used to test two semianalytic models of rift formation. In the first model, the distinction between single (narrow rift) and multiple (wide rift) troughs is determined by the wavelength of necking instabilities. In the second model, narrow-versus-wide morphology is controlled by the evolution of lithospheric strength during rifting. The large-scale, parallel, multiple troughs of the central Valles Marineris are morphologically similar to terrestrial wide rifts, but a lack of distinct faulting in some of the chasmata could imply that the faulted troughs may have formed as isolated narrow rifts. Our results are only marginally consistent with necking leading to a wide rift; allowable combinations of crustal thickness and heat flow lead to decoupling within the lithosphere resulting in a second principal necking wavelength that is smaller than the main trough spacing, for which evidence is equivocal. At high heat flow (>40 mW m-2), however, necking of the strong upper crust alone can yield a shorter wavelength characteristic of a single trough, allowing narrow-rift origins of faulted chasmata. In contrast, the inferred range of crustal thickness and heat flow poorly match the narrow-rift regime of the strength-evolution model but are in good agreement with its predictions for wide rifting. Furthermore, the distinction between wide rift and core complex in this model places an upper bound on heat flow of 70 mW m-2. ¿ 1998 American Geophysical Union