Many mountain ranges in regions of plate convergence are flanked by foreland sedimentary basins that are usually thought to result from flexing of the elastic lithosphere under the passive weight of sediments and thrust sheets emplaced in the mountain belt. However, in some places the depth of the foreland basin bears little relation to the weight of basin sediments and neighboring mountains, implying that this model is too simple and/or that additional forces must affect subsidence. Here I investigate this problem using a numerical model that incorporates coupling between an upthrusting mountain range and a subsiding basin. I show that only part of the subsidence reflected in the shape of a foreland basin is a flexural isostatic response to crustal thickening and sediment loading. The remaining subsidence results from dragging of the basin margin downward by slip on a range-front fault and related flexural deformation of the surrounding lithosphere. The maximum basin depth is shown to increase with increasing fault displacement and can become decoupled from the elevation of the mountain range, especially as erosion becomes increasingly important. Deeply eroded model mountain belts may have foreland basins several times deeper than flexural models predict on the basis of surface loads. These results challenge the prevailing view that fault-bounded foreland basins result solely from the passive weight of mountains and sediments acting on the lithosphere and they highlight that slip on major range-front reverse faults, when repeated over many earthquake cycles, may be recorded in their geometry and stratigraphic records.