Thermal evolution models for the earth which are based on a parameterization of the convective heat transport are critically reexamined. Traditionally, it has been assumed that the temperature dependence of the mantle viscosity implies that internal temperature and convective heat loss are strongly coupled. Recent numerical work on the heat transport by variable viscosity convection demonstrates that the dependence of the heat flow on the mantle temperature may in fact be much weaker than expected. I compare thermal evolution models with strong and weak dependence of the heat loss on the temperature. With the weaker dependence, plate velocities and heat flow in the Archean were not more than 50% higher than today, while with the strong dependence, much larger differences are predicted. In the former case the Archaean mantle temperatures are somewhat higher, and the present-day ratio of radioactive heat production over heat loss (Urey ratio) is 50% or slightly less. The Urey ratios in the traditional parameterized evolution models are >70%. The predictions of both kinds of models are compared with the independent geological, geochemical, and palaeomagnetic evidence. Although this evidence is subject to some uncertainties, it favors in every case the evolution models based on a weak coupling of heat loss to the interior temperature.