Such interpretations seem to call for the evolution of properties that anticipate future needs. But selection lacks foresight, and no one has described a plausible way to provide it. In principle, group selection might produce results that seem to escape this limitation. For example, increased mutation rates may indeed allow populations to adapt more quickly to changed conditions, even though they harm most individuals. The evolutionary problem is that such group benefits are usually weaker than individual costs, in a well-defined sense that makes group selection effective only under very restrictive conditions⁶. So, in general, we need explanations that are based on individual fitness differences⁷.

From this perspective, the obvious function of Hsp90 is to prevent abnormalities of the kinds that appear when it is compromised. Up to a few per cent of adults heterozygous for a mutation that inactivates Hsp90 display significant morphological abnormality, so clearly there is selection to maintain its function. Likewise, increased mutation under stress might plausibly arise from trade-offs affecting individual fitness: stressed cells may simply be unable to maintain normal DNA repair without sacrificing other vital functions.

In the natural world, only living things (and their artefacts) have 'purposes', and natural selection is the ultimate source of all such 'purposeful' design⁸. When speaking of the function or purpose of some feature of an organism, we are therefore referring to the selective advantages that brought the feature into being and that maintain it in the face of recurrent damaging mutations. It is especially important, in any discussion of evolutionary processes, to observe the distinction between function or purpose on the one hand, and effect or consequence on the other. This is not a semantic quibble. Cosmic rays affect evolution by causing mutations, but we would not claim that they exist for that purpose. Similarly, developmental buffering and variable mutation rates may influence the course of evolution, but this does not mean that they evolved to that end.