Moving-base gravity gradiometers are currently under development at Bell Aerospace Co., Charles Stark Draper Laboratory, and Hughes Aircraft Co. In principle, these instruments can be mounted on stable platforms in aircraft or marine survey ships. Unlike a conventional gravimeter, the gradiometer is insensitive to vertical (heave) accelerations of the vehicle and does not require an Eotvos correction. Gradiometers thereby offer an overwhelming improvement in the speed and accuracy of gravity surveys. In addition, a gradiometer provides more information than a gravimeter because a tensor quantity is measured, rather than a scalar.How should the gradient data be used? The simplest approach is to convert the gradients into gravity anomalies by integrating along the vehicle path, and then apply conventional interpretation models. This approach is sound if the survey tracks are closely spaced because continuous two-dimensional gravity anomalies completely define the gravity field outside the earth, according to potential theory. On the other hand, one can widen the survey tracks considerably if a more sophisticated approach is adopted that exploits the extra information in the gradients.In this paper, we compare the 'simple' approach and a more elaborate ('optimal') one that uses the entire gradient tensor. The comparison is based on information theory concepts: How much information is in the gradient data? How much is lost if the gradients are converted to gravity anomalies? Statistical models are fitted to Bouguer residual gravity anomalies from a salt dome field in the Louisiana Gulf Coast, and the interpretation rms errors are evaluated for each approach based on optimal estimation theory. The results show that a gradiometric survey with 18-km spacing contains the same information as a gravimetric survey with 8-km spacing. The sensitivities of the results to the following survey parameters are determined: track spacing, gradiometer noise, vehicle speed, and aircraft altitude.