The large air gun array now in use on R/V Conrad provides enough energy to record mantle reflections with military sonobuoys. The data are used to compute T2/X2 interval velocities in layers 2 and 3 (usually as a combined, single-layer solution) and to identify the critical reflection point from mantle. Comparison of the T2/X2 and conventional refraction solutions (when constant layer velocity is assumed) with the observed critical reflection points shows that the T2/X2 solutions are in better agreement. Nearly half the refraction solutions are in reasonable agreement with the observed critical reflection points, but the remainder yield solutions whose velocities and thicknesses are significantly less than those obtained with the T2/X2 results and whose computed critical reflection points fall short of the observed values. These findings imply that the travel time data are sometimes failing to reveal the intermediate layer (7.2--7.6 km/s) of Sutton et al. (1971), which may be more widespread than is indicated by first-arrival refractions. All of the sonobuoys were located within ancient, stable plates where the basement is relatively smooth. Most of the sonobuoy records show an offset in the travel time due to a shadow zone with an abrupt termination of layer 3 arrivals at the limiting ray, whereas the mantle events are recorded well beyond the limiting ray. This is interpreted as showing a downward decrease of positive velocity gradients within layer 3. Although velocity gradients are required to account for refracted wave amplitude variations, the constant layer velocity assumption provides accurate structure sections if mantle reflection data are available.