The regional geoid of the soutwest Pacific is matched reasonably well by results from a model of the upper mantle density structure (including slabs) associated with subduction zones of the region. Estimates of the geoids are obtained from Geos 3 Seasat radar altimeter data. These data are very well suited to the task of detecting intermediate wavelength (600-4000 km) geopotential variations, which contain the dominant signature of subducting slabs. Gravimetric profiles across trench/island arc complexes resolve primarily short wavelengths. The model represents subducting slabs as thin surfaces of anomalous mass per unit area. These surfaces are positioned using published seismicity results which detail the configuration of the Benioff zones. Crustal effects are ignored. Effects due to the contrast between the young thermal lithosphere of the behind arc regions (marginal basins) and the older lithosphere seaward of the trench are modeled. Results indicate that the New Hebrides slab possesses an average areal density anomaly of about 3¿105 g cm-2. This is about 3 times that which is estimated for the Tonga-Kermadec slab. Both of these inferred density anomalies are somewhat less (~10 and 65%) than anomalies predicted by a thermal model. Consistent with the results of other studies, additional modeling suggests that slabs worldwide may be an important source of large, long-wavelength gravity highs; i.e., they may contribute substantially to geopotential power of harmonic degree as low as three or four (subduction, geoid anomalies, slab, lithosphere).