Relative to a fixed point in the upper plate U, the transient subduction hinge H can converge, diverge, or be stationary. Similarly, the lower plate L can converge, diverge or be stationary. The subduction rate V_S is given by the velocity of the hinge H minus the velocity of the lower plate L (V_S = V_H − V_L). The subduction rate 1) increases when H diverges, and 2) decreases when H converges.

Combining the different movements, at least 14 kinematic settings can be distinguished along the subduction zones. Variable settings can coexist even along a single subduction zone, as shown for the 5 different cases occurring along the Apennines subduction zone. Apart from few exceptions, the subduction hinge converges toward the upper plate more frequently along E- or NE-directed subduction zone, whereas it mainly diverges from the upper plate along W-directed subduction zones accompanying backarc extension.

Before collision, orogen growth occurs mostly at the expenses of the upper plate shortening along E–NE-directed subduction zones, whereas the accretionary prism of W-directed subduction zones increases at the expenses of the shallow layers of the lower plate. The convergence/shortening ratio is > 1 along E- or NE-directed subduction zones, whereas it is < 1 along accretionary prisms of W-directed subduction zones.

Backarc spreading forms in two settings: along the W-directed subduction zones it is determined by the hinge divergence relative to the upper plate, minus the volume of the accretionary prism, or, in case of scarce or no accretion, minus the volume of the asthenospheric intrusion at the subduction hinge. Since the volume of the accretionary prism is proportional to the depth of the decollement plane, the backarc rifting is inversely proportional to the depth of the decollement. On the other hand, along E- or NE-directed subduction zones, few backarc basins form (e.g., Aegean, Andaman) and can be explained by the velocity gradient within the hangingwall lithosphere, separated into two plates.

When referring to the mantle, the kinematics of subduction zones can be computed either in the deep or in the shallow hotspot reference frames. The subduction hinge is mostly stationary being the slab anchored to the mantle along W-directed subduction zones, whereas it moves W- or SW-ward along E- or NE-directed subduction zones. Surprisingly, along E- or NE-directed subduction zones, the slab moves “out” of the mantle, i.e., the slab slips relative to the mantle opposite to the subduction direction. Kinematically, this subduction occurs because the upper plate overrides the lower plate, pushing it down into the mantle. As an example, the Hellenic slab moves out relative to the mantle, i.e., SW-ward, opposite to its subduction direction, both in the deep and shallow hotspot reference frames. In the shallow hotspot reference frame, upper and lower plates move “westward” relative to the mantle along all subduction zones.

This kinematic observation casts serious doubts on the slab negative buoyancy as the primary driving mechanism of subduction and plate motions.

W-directed subduction zones rather provide about 2–3 times larger volumes of lithosphere re-entering into the mantle, and the slab is pushed down. This opposite behavior is consistent with the down-dip extension seismicity along E–NE-directed subduction zones, and the frequent down-dip compression along the W-directed subduction zones.

Subduction kinematics shows that plate velocity is not dictated by the rate of subduction. Along the W-directed subduction zones, the rate of subduction is rather controlled i) by the hinge migration due to the slab interaction with the “easterly” trending horizontal mantle wind along the global tectonic mainstream, ii) by the far field plate velocities, and, iii) by the value of negative buoyancy of the slab relative to the country mantle.

Alternatively, E–NE–NNE-directed subduction zones have rates of sinking chiefly determined i) by the far field velocity of plates, and ii) by the value of negative buoyancy of the slab relative to the country mantle. Along this type of subduction, the subduction hinge generally advances E–NE-ward toward the upper plate decreasing the subduction rate, but it moves W–SW-ward relative to the mantle.

All this indicates that subduction zones have different origin as a function of their geographic polarity, and the subduction process is more a passive feature rather than being the driving mechanism of plate motions. A rotational component combined with mantle density and viscosity anisotropies seems more plausible for generating the global tuning in the asymmetry of subduction zones.