As the asthenospheric mantle rises at oceanic spreading centres, it undergoes partial melting, producing oceanic crust and depleted mantle, both of which have lower intrinsic density than the asthenospheric mantle from which they were derived. With a warmer asthenosphere in the Archean, these effects are enhanced, leading to the possibility that subduction was no longer feasible. I investigate the density of the oceanic crust and underlying mantle for a mantle with temperatures 200 °C higher than today, using models of the chemistry of melting and the mineralogy of the ensuing rocks. For the melting model used, crustal thicknesses are 21 km and the depth to which the mantle is partially melted is 114 km, compared with 7 and 54 km for a comparable model of modern Earth. Two thermal-evolution models for Archean oceanic lithosphere are examined. One assumes twice the heat flow into the base of the plates, which severely restricts the depths to which the plates can cool with age. A second assumes the plates can cool to the depth to which the asthenosphere undergoes partial melting, resulting in heat flow into the base of the plates only 1.3 times as large as today. With the first model, oceanic plates do not become denser than an equivalent column of asthenosphere. With the second, they do after ∼50 Ma of cooling. In both cases, however, the cooling is sufficient to provide a significant driving force for the initiation of subduction because the sole requirement for a subduction-initiation drive is that the cooled lithosphere be denser than the column of differentiated asthenosphere that would replace it. This, combined with the low flexural rigidity of Archean plates, makes the initiation of subduction probably slightly easier than it is today. The relatively low density of oceanic plates results in lower slab pull, but this effect is counterbalanced both by the likelihood that some of the low-density crust may have been delaminated, and by the effect of passage of the thicker crust through the eclogite transition. Given our present knowledge of Archean thermal conditions, there does not appear to be a compelling theoretical argument against efficient subduction processes at that time.