Plate tectonics introduces chemical, thermal, viscosity, melting and density inhomogeneities into the mantle and stress inhomogeneity into the plates. Idealized models often assume uniform mantle, rigid homogeneous plates,non-passive mantle, and ad hoc explanations for island chains, melting anomalies and continental breakup. Plates, however, drive and break themselves and organize the underlying mantle, in common with other cooled-from-above systems. Pressure, often ignored in simulations, suppresses thermal expansion and the Rayleigh number making the deep mantle a sluggish system with gigantic features, consistent with tomography, and isolating it from the upper mantle and plate tectonics (except by conduction and gravity).Large scale chemical stratification is therefore likely. Plate tectonics, with adjectives such as rigid, homogeneous, isothermal, fixed, subsolidus, reservoir, steady-state etc. dropped, is a much more powerful concept than generally believed. Cracks, rifts, dikes, incipient plate boundaries, melting anomalies and variations in melt volume and chemistry are natural parts of the general theory of plate tectonics. The long-sought alternative theory to deep mantle plumes may just be a less restricted view of plate tectonics. It appears to be the adjectives, assumptions and other baggage that are the problem. Many of the geochemical paradoxes associated with deep plumes and primordial views of the mantle can be traced to the reservoir concept where deep seismic boundaries are assumed to delineate reservoirs. The mantle is heterogeneous, as it should be from plate tectonic considerations (recycling, inefficient melt and gas extraction, history ).This suggests that sampling theory and dispersed components may explain the diversity of basalts. The central limit theorem (CLT) predicts that large scale averagers, such as ridges, should have less variance and less extreme values than xenoliths, inclusions, seamounts or OIB, as observed. Homogeneity is achieved by partial melting, averaging and magma chamber processes, not by large scale convection. This idea is tested with Os and He isotopes, which are as different from each other and from the standard isotopes as possible. The conclusion is that both MORB and OIB are products of a heterogeneous upper mantle, sampled in different ways (volume of mantle averaged, degree of melting, magma chamber processes). The CLT plus mass balance calculations obviate the need for an undegassed reservoir or lower mantle components . High ³He/⁴He components can be ubiquitous in the shallow mantle but only expressed in OIB, off-axis seamounts and other volcanic systems sampling small mantle volumes, or at the onset of volcanism.