In the literature, there is a widespread tendency to believe, or to admit the possibility, that back arc basins are causally linked to subduction. In this work we argue that this hypothesis is poorly supported by observational evidence and by the quantifications of the effects of subduction processes. The major questions which still lack convincing answers are:

a) why does back arc extension occur in some subduction zones and not in others?

b) why in a number of basins has subduction preceded of My the beginning of back arc extension and why in other basins has back arc activity ceased much earlier than subduction?

c) why are marginal basins often associated to strongly curved arcs?

d) why do T-A-BA systems not develop along the entire length of the convergent boundary, but only occur along a fraction of it?

e) why is the occurrence of back arc extension not correlated with any of the major parameters of subduction systems?

Furthermore, it has been argued by several authors that the implications of subduction related driving mechanisms cannot easily be reconciled with the observed space-time deformation pattern of T-A-BA systems in the circum-Pacific (e.g. Uyeda and Kanamori, 1979; Taylor and Karner, 1983; Uyeda, 1986; Tamaki and Honza, 1991; Flower et al., 2001) and Mediterranean regions (Mantovani et al., 1997, 2000a, 2001a). Then, the attempts to quantify the effects of subduction, by numerical and analogue modelling, have shown that the tensional stress induced in the overriding plate by the subduction related driving mechanisms here discussed (slab pull, corner flow and sea-anchor models) is not strong enough to cause back arc extension in the lithosphere, unless it includes zones of weakness and the interplate coupling is very low.

In general, the basic concept suggested in this work is that the slab pull force, or any other subduction related force, cannot produce the detachment of the arc from the overriding plate and, consequently, slab roll back cannot occur. This implies that the only effect of these forces may be the steepening of the slab, without trench retreat. In addition, it is argued that this last effect may only be achieved when the mantle surrounding the slab has relatively low viscosity, as occurs, e.g., beneath the Mariana arc. When, instead, this condition is not fulfilled, as under the Peru-Chile subduction zone, the slab maintains a very low dip angle, due to the high resistance of viscous forces in the mantle.

In our view, the detachment of the arc from the overriding plate may only be produced by external forces, induced for instance by the oblique indentation of a strong and buoyant structure against an accretionary belt (Fig. 3). This mechanism may provide plausible and coherent answers to the questions mentioned above and has not to face conceptual difficulties, since its plausibility can be demonstrated under given tectonic conditions. The main problem is demonstrating that such conditions have actually occurred in the zones where T-A-BA systems developed. Unfortunately, this demonstration is not easy, since there is no direct evidence on the paleo-kinematics of plates and on the rheological properties of the structures involved. These features may only be tentatively inferred and reconstructed by the analysis of as many as possible geological, geophysical and, where available, geodetic observations, which, in addition, are often affected by a poorly known uncertainty. In spite of this, we think that the presently available evidence is sufficient to consider the extrusion model an important candidate as genetic mechanism of back arc extension. For the Mediterranean area, an accurate study of the post-Eocenic deformation pattern, and of its compatibility with the implications of the various driving mechanisms so far proposed (Mantovani et al., 1996, 1997, 2000a, 2001a, 2002) led us to grow an high confidence on the hypothesis that T-A-BA systems were generated by the interaction of buoyant structures, driven by plate convergence. As regards the circum Pacific regions, we noted in literature a growing attention on the idea that back arc extension is closely connected with extrusion processes. In the text, we point out that for most of the T-A-BA systems reported in the Table it is possible to find in literature an extrusion-related interpretation which may provide plausible explanations of the observed features. When this is not possible, as for the Lau-Havre, Shikoku and Parece Vela basins, we have discussed the compatibility between the observed tectonic context and the conditions required for the occurrence of the extrusion model.

Another interpretation of back arc opening, which can be classified as non-subduction related, has been recently proposed (Flower et al., 1998, 2001). This model suggests that back arc extension in the western Pacific could be an effect of the extrusion of asthenospheric mantle, induced by the closure of the Tethys ocean. In particular, the occurrence of this mechanism has been discussed for the Izu Bonin-Mariana T-A-BA system (Flower et al., 2001), where it is recognized as a major cause of slab roll back and basin opening. This idea is mainly based on petrological evidence and seismic anisotropy data, which would indicate an eastward flow of asthenospheric material from the India-Eurasia collision zone towards the western Pacific area. However, one should also consider other possible interpretations of the above evidence. For instance, the eastward roll back of slabs, predicted e.g. by the slab pull (Fig. 2) and extrusion (Fig. 3) models, would recall asthenospheric material from the surrounding zones, causing an eastward asthenospheric flow like the one inferred for the eastern Asian area. Furthermore, one should also understand if the presumed mantle flow can produce the complex time-space deformation patterns of the West Pacific arc-trench systems, with particular regard to the different evolution of the Izu Bonin and Mariana arcs. In our opinion, explaining such features as effects of the interaction between lithospheric structures, characterized by laterally heterogeneous buoyancy and mechanical strength, seems to be less problematic.