Alkaline lavas hosting mantle xenoliths occur in the hot back arc of the Cordillera of western Canada and Alaska and provide constraints on lithosphere thickness (within 5 to 10 km) and controls on the lithosphere-asthenosphere boundary (LAB). Such information is complementary to geophysical investigations of lithosphere strength and deformation in similar continental regions. Barometry of spinel lherzolite xenoliths and alkaline lavas from the Cordillera show LAB depths from 50 to 80 km, and marked thinning of the lithosphere at the Cordillera-craton transition coincident with the Rocky Mountain Trench-Tintina Fault. Similar barometric data globally for continental intraplate lavas show a relatively common lava equilibration depth of 65 to 80 km, that mostly coincide with LAB depths observed by seismology, and with the spinel- to garnet peridotite transition. The LAB is inferred as accumulated partial melt, modulated in depth by mantle solidus depression caused by varying H2O storage capacity across the garnet-spinel peridotite phase change. This mechanism does not explain cases where LAB depths are < 65 km, which instead require locally hotter asthenosphere. In many regions preserving a shallow LAB, foundering and wholesale replacement of mantle lithosphere is invoked. With the exception of North China, mantle xenoliths sampled from continental margins record cooling timescales that appear far longer than required for foundering and replacement in the past 50 Myr. Either cooling histories cannot be adequately modelled using closure/diffusion arguments or wholesale replacement of mantle lithosphere is a dubious process.