Velocity-model determination during seismic data processing is crucial for any kind of depth imaging. We compared two approaches of grid tomography: prestack stereotomography and normal-incidence-point (NIP) wave tomography. Whereas NIP wave tomography is based on wavefield attributes obtained during the common reflection surface stack and thus on the underlying hyperbolic second-order traveltime approximation, prestack stereotomography describes traveltimes by local slopes (i.e., linearly) in the prestack data domain. To analyze the impact of the different traveltime approximations and the different input-data domains on velocity model building, we applied two implementations of these techniques to two profiles of a field marine data set from the Levante Basin, eastern Mediterranean. Because ofthe presence of a thick, tabular mobile unit of the Messinian evaporites, strong vertical and lateral velocity contrasts had been expected. The velocity models revealed the reconstruction of high-velocity contrasts by grid tomographic methods is limited because of the smooth description of the velocity distribution. The lateral resolution of velocities obtained from prestack stereotomography appears to be better than those from NIP wave tomography, which is related to the difference in the approximation of traveltimes, the determination of input data, and the description of the velocity distribution. Other differences are caused mainly by different implementations of the inversion schemes. Nevertheless, both algorithms provide suitable models for high-quality depth imaging, whereas most of the reflections are fairly flat in CIGs.