Have a look at this short review of the new book Biased Embryos and Evolution by Wallace Arthur. The reviewer is Armond Leroi, a British biologist.
I mention this book for two reasons. The first is that Leroi praises the book as a good, clear introduction to the subject of evo-devo, one comprehensible to non-biologists. Personally, I've been looking for such an introduction for a while now, and I will probably pick up a copy of this book eventually.
The other reason is that Arthur apparently discusses the idea of orthogenesis. This refers to the possibility that evolution is effectively channeled into certain directions via design and developmental constraints. If this is true, then the trajectories traced out by evolving populations are not as random as is commonly thought. Such a discovery could also provide a compelling explanation for evolutionary convergence.
Orthogenesis has been very unpopular among scientists since no one has proposed a plausible mechanism by which it happens. Arthur seeks to change that. Here's Leroi summarizing Arthur's argument:
These are brave words. Orthogenesis has been a cause without mainstream sympathizers for at least 60 years. The reason for this is that no one has provided a mechanism by which it might work. Most biologists believe that the evolutionary direction of lineages is largely determined by natural selection; a minority make great play of contingency (the non-selective effects of meteor strikes and the like). So what is going on? Has Arthur discovered a new principle of evolution?
Not really, no. The fuel in his orthogenetic engine is 'mutation bias'. Mutation produces novel phenotypes, but it does not produce all novel phenotypes in equal frequency in a given population. For example, mutations that cause an animal to become smaller than normal might be more common than those that cause it to become larger. This bias is the result of the way body size is specified in development — a bias that might influence the direction that evolution takes, causing small animals to evolve more often than large ones.
Mutation bias is not enough to produce orthogenesis, however. If there is a single fitness optimum, or if the population is sufficiently large to ensure that all possible mutations are always present, then the direction of evolution will be dictated by natural selection alone. But if the landscape is rugged and population sizes small, the particular peak climbed by a population could depend on what mutations happen to be available. This is not orthogenesis of old — which posited a force independent of, or even capable of opposing, natural selection — but a reassignment of influence over evolutionary trajectories from natural selection to the kind of genetic variation available for it to work on.
If 'mutation bias' turns out to be a new term for an old idea, the same seems to be true for another unusual term: 'internal selection'. This is the idea that as one part of an organism evolves, it exerts selective pressure on other parts to change as well. Suppose a mutation increasing the length of an animal becomes fixed in a population. This might cause the subsequent fixation of another mutation that increases the animal's width, so restoring an original, harmonious, proportion. Arthur makes great play of this, but I think the interaction at the heart of this process is well known to population geneticists as 'fitness epistasis' and has often been experimentally demonstrated.
Leroi's conclusion is that Arthur's ideas are interesting, but far from proved.
I mention this here because, if Arthur is correct and mutation bias does lead to certain evolutionary pathways being favored over others, then this throws yet another monkey wrench into William Dembski's scheme for inferring design in organisms. Remember, Dembski's key ideas are that if an object is highly improbable and conforms to some independently describable pattern, then we can conclude it was designed. As has been documented elsewhere, his whole system is shot through with holes. But one of the biggest is his claim that we can carry out meaningful probability calculations regarding the formation of complex, biological structures.
Arthur's ideas would have to be taken into consideration in any such calculation. If certain sorts of phenotypes are more likely to be found in a population than others, than this will effect our measurement of how likely it is to evolve a particular complex system. And unless we can quantify very precisely which sorts of mutations are more likely to be found than others, it will be effectively impossible to carry out any such computation.
The evolution of complex systems dempends on far more variables than can possibly be captured in any elementary probability calculation. Arthur's ideas, if correct, would make that problem even more acute.