The interaction between developmental bias and natural selection: from centipede segments to a general hypothesis

Do limitations to the ways in which mutations can alter developmental processes help to determine the direction of phenotypic evolution? In the early days of neo-Darwinism, the answer given to this question was an emphatic 'no'. However, recent work, both theoretical and empirical, argues that the answer should at least be 'sometimes', and possibly even a straightforward 'yes'. Here, I examine the key concept of developmental bias, which encompasses both developmental constraint and developmental drive. I review the case of centipede segment number, which is a particularly clear example of developmental bias, but also a rather unusual one. I then consider how, in general terms, developmental bias and natural selection might interact, with the result that it is their interaction, rather than either process on its own, that determines evolutionary direction. Essentially, the whole argument is about the extent to which phenotypic variation is developmentally structured as opposed to amorphous or random. This issue can be traced back to the very beginning of evolutionary biology, and in particular to a difference of opinion between Darwin and Wallace, who emphasized, respectively, character correlation and character independence.