The hypothesis that viscous resistive forces in the upper mantle may oppose the lateral displacement of subducted lithosphere was initially advanced in the early stage of the plate tectonics theory (Havemann, 1972) and then reconsidered by Uyeda and Kanamori (1979) to explain the generation and distribution of back arc basins in the world. This last model was based on the hypothesis that the position of the trench is stationary, in an absolute reference frame, whatever the velocity of the subducting plate may be. Under this assumption, the deformation of the overriding lithosphere is determined by its relative motion with respect to the trench. When the overriding plate moves landward, it undergoes tensional deformation (back arc opening), since its separation from the subducting plate is prevented by the lithostatic load, that largely overcomes the tensional strength of the overriding lithosphere (Shemenda, 1993).

However, the basic assumption of this model cannot easily be reconciled with the fact that advancing, stationary and retreating subduction boundaries are now clearly recognized (Taylor and Karner, 1983; Carlson and Melia, 1984; Royden, 1993b, 1996). For instance, Uyeda and Kanamori (1979) suggested that a fully developed anchor effect is expected at the Mariana subduction system, while the fast advancing of the Mariana trench is recognized (e.g. Carlson and Mortera-Gutierrez, 1990).

The possible influence of the slab-asthenosphere interaction on back arc dynamics has been then reconsidered by Scholz and Campos (1995), who suggested that back arc extension occurs when the overriding plate moves landward, trailing the slab (Fig. 2d). The extensional stress in the back arc zone is induced by a combination of the above force and of the hydrodynamic force which resists the motion of the slab through the viscous asthenosphere (sea anchor force). The magnitude of this force depends on the trench-normal component of the overriding plate velocity, on the slab surface and on the mantle viscosity. Steady state plate motion provided by current kinematic models implies the equilibrium of forces in a subduction system, i.e. the force which drives the overriding plate must counterbalance the sea-anchor mantle resistance to maintain the observed plate velocity. Thus, the estimate of the sea-anchor effect may give information on the force that pull the overriding lithosphere landward. The proposed dynamic context may produce tensional failure in the overriding plate when the anchor force reaches the average tensional strength of the lithosphere. Under some simplyfing assumptions, Scholz and Campos (1995) estimated the sea anchor force for 29 circum-Pacific subduction zones. In the zones where back arc opening never occurred, or ceased, the values of the above force are negative (the overriding plate approaches the trench) or positive, but lower than 2.2x1012 Nm-1. In 4 subduction zones, associated with active back arc basins (Mariana, Kermadec, Tonga and Hikurangi), the above force ranges between 7.4 and 9.7x1012 Nm-1, i.e. values slightly lower than the presumed strength of the intact lithosphere (~1013 Nm-1, corresponding to a 200 MPa deviatoric stress, averaged over a 50 Km thick oceanic lithosphere). To account for the occurrence of back arc extension in the above zones, the authors supposed that back arc opening, once initiated, requires a force (3x1012 Nm-1 = 60 MPa averaged stress) considerably lower than the one acting in the initial stage. However, this does not explain how back arc extension began. Furthermore, one should consider that protracted plate thinning could strengthen, rather than weaken, the stretched lithospheric domains, due to conductive cooling of the lithospheric mantle (Sonder and England, 1989; Ruppel, 1995).

As Scholz and Campos admit, the sea-anchor model fails to explain back arc activity observed in the Ryukyu and New Hebrides subduction zones, for which the estimated driving force is close to zero. The above authors also admit that the occurrence of back arc extension cannot be related to large-scale plate motion only, but it must also depend on local factors.

The implications of this model cannot easily be reconciled with the kinematics of the Mediterranean T-A-BA systems, since the development of these zones was mainly determined by the seaward motion of arcs, with very slow convergence rates between the overriding and subducting plates (e.g. Biju-Duval et al., 1977; Dercourt et al., 1986; Mantovani et al., 1997, 2000a).

Discussions as to unresolved problems with this model are also reported by other authors (e.g. Taylor and Karner, 1983; Uyeda, 1986; Flower et al., 2001).