During the summer of 1998, active-source seismic data were collected along a transect running 400 km NNW-SSE across the central Tibetan Plateau as the third phase of project INDEPTH (International Deep Profiling of Tibet and the Himalaya). The transect extends northward from the central Lhasa block, across the Jurassic Bangong-Nujiang Suture (BNS) at 89.5¿E, to the central Qiangtang block. A seismic velocity model for the transect to ~25 km depth produced by inversion of P wave first arrivals on ~3000 traces shows (1) a ~50-km-wide region of low velocity (at least 5% less than surrounding velocities) extending to the base of the model at the BNS; (2) sedimentary cover for the southern Qiangtang block that is ~3.5 km thick; (3) a distinct interface between sedimentary cover and Qiangtang basement or underplated Jurassic melange in the central Qiangtang block; and (4) evidence that the Bangoin granite extends to a depth of at least 15 km. The BNS has little geophysical signature, and appears unrelated to the ~5 km northward shallowing of the Moho which is associated with the BNS in central Tibet. Geophysical data along the main INDEPTH III transect show little evidence for widespread crustal fluids, in contrast to the seismic bright spots found in southern Tibet and to magnetotelluric evidence of fluid accumulations in eastern Tibet. A comparison between the global average and Tibetan velocity-depth functions offers constraints for models of plateau uplift and crustal thickening. Taken together with the weak geophysical signature of the BNS, these velocity-depth functions suggest that convergence has been accommodated largely through pure-shear thickening accompanied by removal of lower crustal material by lateral escape, likely via ductile flow. Although we cannot resolve the details, we believe lateral lower crustal flow has overprinted or destroyed evidence in the deep crust for the earlier assembly of Tibet as a series of accreted terranes.