Strontium and lead isotopic data from oceanic rocks seem to require not just two or three but many ancient mantle sources, and this would remove the geochemical rationale for mantle layering, although it would not preclude layering. Seamount data require that at least some small heterogeneities are present in the uppermost mantle. There is no compelling independent evidence that rocks with chrondritic neodymium isotopes are derived from a primitive mantle source. It follows that isotopic correlations do not necessarily constrain the proportion of depleted mantle or the mean silicate (or ''bulk earth'') values of Sr and Pb isotopic compositions. Several recent geophysical arguments concerning the mantle viscosity profile, aseismic extensions of Benioff zones, and the angles and shapes of Benioff zones indicate that mantle flow associated with the plates penetrates well into the lower mantle. This would probably preclude chemical layering at or above the transition zone. These arguments suggest a mantle with heterogeneities of many sizes and compositions at all depths and with any stratification confined to roughly the bottom half. Such stratification might result from a pressure-induced viscosity increase with depth or from gravitational settling and would help to explain the longevity of heterogeneities, including primordial He and Xe isotopic signatures, and the longevity and low horizontal velocities of some hot spots. Possible sources of mantle heterogeneities include recycled oceanic and continental crust, detached continental lithosphere, and surviving primitive mantle.