The time-term technique, modified to accommodate asymmetry in the shot-station configuration, has been used to derive a three-dimensional model of crustal thickness from data obtained in the 1965 east coast on-shore off-shore experiment (ECOOE). Subcrustal velocities were found to be 8.15 km/sec in the northern part of the area and 8.25 km/sec in the southern part of the area. The depth to the M discontinuity beneath the middle Atlantic states has been contoured on the basis of some 200 time terms; crustal thicknesses vary from about 30 km in some areas over the coastal plains and continental shelf to nearly 60 km in a few localities along the crest of the Appalachian Mountains. Three-dimensional gravity computations on the basis of the seismic structure and under assumptions of different straight-line velocity-density relationships show that the predicted gravity values are well correlated in size and shape with the observed Bouguer anomalies for dp/dV between 0.20 and 0.23. Straight-line velocity-density relationships with dp/dV greater than 0.23 (such as those of Birch for constant mean atomic weight) do not yield predicted gravity values compatible with the observed anomalies. This result indicates that the mean atomic weight of subcrustal rocks is less than that of rocks in the lower part of the crust.