The first Bell Aerospace BGM-3 Marine Gravity Meter System available for academic use was installed on R/V Robert D. Conrad in February, 1984. The BGM-3 system consists of a forced feedback accelerometer mounted on a gyrostabilized platform. Its sensor (requiring no cross-coupling correction) is a significant improvement over existing beam and spring-type sea gravimeters such as the GSS-2.A gravity survey over the Wallops Island test range together with the results of subsequent cruises allow evaluation of the precision, accuracy, and capabilities of the new system. Over the test range, the BGM-3 data were compared directly to data obtained by a GSS-2 meter onboard R/V Conrad. The rms discrepancy between free-air gravity anomaly values at intersecting ship tracks of R/V Conrad was + or -0.38 mGal for BGM-3 compared to + or -1.60 mGal for the GSS-2. Moreover, BGM-3's platform recovered from abrupt changes in ship's heading more rapidly than did the platform of GSS-2.The principal factor limiting the accuracy of sea gravity data is navigation. Over the test range, where navigation was by Loran C and transit satellite, a two-step filtering of the ship's velocity and position was required to obtain an optimal Eotvos correction. A spectral analysis of 1 minute values of the Eotvos correction and the reduced free-air gravity anomaly determined the filter characteristics. To minimize the coherence between the Eotvos and free-air anomaly, it was necessary to prefilter the ship's position and velocity. Using this procedure, reduced free-air gravity anomalies with wavelengths as small as a few kilometers can be resolved.