We examine the problem of partitioning between shortening and extrusion in the India-Asia collision since 45 Ma. We compute the amount of shortening expected from the kinematics of India's motion with respect to Eurasia, using the reconstruction at collision time to put bounds on the possible amounts of surface loss within Greater India and within Eurasia. We then compute the amounts of surface loss corresponding to the thickened crust of Tibet and of the Himalayas, assuming conservation of continental crust. The spatial distribution of the topography reveals a large systematic deficit of crustal thickening distributed rather uniformly west of the eastern syntaxis but an excess of shortening east of it. This distribution indicates an important eastward crustal mass transfer. However, the excess mass east of the eastern syntaxis does not account for more than one third to one half of the deficit west of the eastern syntaxis. The deficit may be accounted either by loss of lower crust into the mantle, for example through massive eclogitization, or by lateral extrusion of nonthickened crust. A mass budget of the crust of the Himalayas indicates that lower crust has not been conserved there, but the deficit is so large that local loss in the mantle is unlikely to be the unique cause of the deficit. Alternatively, following Zhao and Morgan , lower crust may have been transferred below the Tibetan crust. We conclude that a combination of possible transfer of lower crust to the mantle by eclogitization and lateral extrusion has to account for a minimum of one third and a maximum of one half of the total amount of shortening between India and Asia since 45 Ma. This conclusion leaves open the possibility that the partitioning between extrusion and loss of lower crust into the mantle on the one hand and shortening on the other hand has significantly changed during the 45 m.y. history of the collision.