Accurate location of the salt-sediment interface is one of the first problems to be solved during exploration around salt domes. However, determination of salt boundaries from seismic data has been notoriously missing even though forward modeling has shown that the salt flank reflection can be one of the highest amplitudes on a seismic record. The sources of difficulty are that conventional data acquisition and processing are usually designed to attenuate any signal that arrives at high incidence angles, including reflections originating from steeply dipping structures such as salt flanks. For example, if regional stacking velocities are used during processing, the salt flank reflection can be so attenuated as to be unobservable.Inverse modeling by ray methods is known to be a suitable technique for reconstruction of steeply dipping structures, provided key marker reflections are identified. Salt flank reflections have been recovered in two field-recorded examples through application of unusually high stacking velocities; the reflections characteristically have very large migration distances and cross other reflections at high angles. The inverse model computed for one example produced a salt flank with dip rates up to 70 degrees, and the computed structure compared to drilling data is accurate to within 1 percent. Results from the second example are also consistent with known geology but are more striking: dip rates of 90 degrees are achieved and the model shows a slight overhang of salt. These results show there is nothing in principle to prevent inversion of steeply dipping structures from surface-recorded seismic data, including such extreme cases as overhanging salt domes.