A number of recent studies have used gravity data from mid-ocean ridges to infer along-axis crustal thickness variations, and hence variations in melt supply. A problem with these studies is that it is difficult to distinguish the effects of crustal thickness variations from those of crustal and mantle density variations. We analyze gravity data from the 12-13 km offset Blake Spur Fracture Zone in 140 Ma crust in the western North Atlantic. The advantage of studying this area is that mantle density variations are likely to be much smaller than those at the ridge axis, and crustal thickness variations are well constrained by expanding spread profiles (ESPs). Thus we are able to separately quantify the contributions made to the gravity anomaly by crustal thickness and crustal density variations. After removal of the gravity effects of the seabed and the oceanic basement topography, as defined by a grid of multichannel seismic reflection profiles, the gravity data show a linear 5-mGal high along the fracture zone. After removal of the effects of crustal thickness variations defined by the ESPs, and of a basal high-velocity layer beneath the fracture zone interpreted to be partially serpentinized upper mantle, the data show a 5-mGal gravity low at the fracture zone. An alternative interpretation of the basal layer, as a thickened Moho transition zone, also leads to a residual gravity low. These results, together with other recent results from fracture zones on the Mid-Atlantic Ridge, suggest that crustal thickness variations at fracture zones may be consistently underestimated by studies based on gravity data. The discrepancy may be explained by variations in crustal density due to enhanced fracturing in the vicinity of fracture zones, and to disproportionate thinning of oceanic layer 3 near segment boundaries. ¿ American Geophysical Union 1995