Can we explain all terrestrial magmatism with a single unifying theory? If the mantle is convectively stirred to the point of homogeneity at all scales, then certainly not. Plate tectonics introduces heterogeneity into the mantle having dimensions of slabs. As a slab sinks it may become deformed or stretched if shear stress gradients are high. The Rayleigh numbers used in simulations of mantle convection lead to chaotic convection and rapid mixing. Plumes are invoked to re-introduce heterogeneity into the upper mantle. These calculations ignore effects of plates, continents, pressure, phase changes, and layered convection. Plate tectonics, lithosphere, continental roots, pressure effects, high Prandtl number, stratification and free-slip boundary conditions organize and impede mantle flow. The mantle,convecting sluggishly and passively, is not well mixed. Convective stirring takes time and heterogeneities are constantly being introduced into the shallow mantle. Processes at ridges, however, homogenize the products of heterogeneous mantle melting.

New data confirms that under lower mantle (LM) conditions, material rich in Ca, Al and Na (i.e. eclogite) is less dense than pyrolite or any plausible LM mineralogy. The idea that slabs, particularly young ones, will be stuck in the UM is strengthened. Piclogite, particularly if carbonated, melts deeper and at lower temperature than normal mantle; it can develop low seismic velocities as it equilibrates (Presnall, 2003). About 15% of ocean floor surface area is composed of young (< 20 My) lithosphere approaching trenches and in back-arc basins; roughly 0.2 km³/year of such material is currently entering trenches. This is about the magma budget of 'midplate' magmatism. A similar area (plateaus, aseismic ridges) has thick crust. Young and thick-crust material does not subduct to great depths. It warms up and thermally equilibrates on short times scales, becoming neutrally buoyant, and can be sampled again by partial melting on characteristic time scales of 1-2 Gyr at leaky transform faults, incipient plate boundaries, extensional regions of the lithosphere and migrating ridges. Thicker slabs of older oceanic crust are more likely to sink deeper into the mantle, after contributing their sediments and fluids to the shallow mantle. The upper mantle is therefore a highly heterogeneous assemblage of enriched and depleted lithologies representing a wide range in chemical composition, scales, ages, melting point, fertility,and isotopic compositions.

The homogeneous nature of MORB usually attributed to long-term stirring and mixing of the mantle source is more likely due to homogenization near the extraction site by the sampling process (partial melting and magma mixing). Ubiquitous crustal and slab-scale heterogeneities comprised of recycled oceanic crust or lithosphere in the shallow mantle explains the statistical properties of oceanic basalts and may also be responsible for melting anomalies themselves. Shallow chemical heterogeneities cause topographic and crustal thickness anomalies without substantial thermal anomalies. Scattering of high frequency seismic waves is one way to test this hypothesis. The anisotropy and anelasticity of the asthenosphere is consistent with such a structure.The presence of CO₂ in the upper mantle, long advocated by Dean Presnall, explains the missing-CO₂ problem (and the helium paradoxes) and, together with trapped young slabs, may be responsible for the low seismic velocities, Q, and melting anomalies, at 'normal' mantle temperatures. Shallow recycling of crust- and slab-sized objects obviates the need for concentrated hot jets localized under 'hotspot' volcanoes and deep mantle reservoirs.