Thermal histories have been calculated for simple models of the earth which assume that heat is transported by convection throughout the interior. The application of independent constraints to these solutions limits the acceptable range of the ratio of present radiogenic heat production in the earth to the present surface heat flux. The models use an empirical relation between the rate of convective heat transport and the temperature difference across a convecting fluid. This is combined with an approximate proportionality between effective mantle viscosity and T-n, where T is temperature and it is argued that n is about 30 throughout the mantle. The large value of n causes T to be strongly buffered against changes in the earth's energy budget and shortens by an order of magnitude the response time of surface heat flux to changes in energy budget as compared to less temperature-dependent heat transport mechanisms. Nevertheless, response times with n=30 are still as long as 1 or 2 b.y. Assuming that the present heat flux is entirely primordial (i.e., nonradiogenic) in a convective model leads back to unrealistically high temperatures about 1.7 b.y. ago. Inclusion of exponentially decaying (i.e., radiogenic) heat sources moves the high temperatures further into the past and leads to a transition from 'hot' to 'cool' calculated thermal histories for the case when the present rate of heat production is near 50% of the present rate of heat loss. Requiring the calculated histories to satisfy minimal geological constraints limits the present heat production/heat loss ratio to between about 0.3 and 0.85. Plausible stronger constraints narrow this range to between 0.45 and 0.65. These results are compatible with estimated radiogentic heat production rates in some meteorites and terrestrial rocks, with a whole-earth K/U ratio of 1--2¿104 giving optimal agreement.