[One basic criterion] for distinguishing between random and nonrandom mutation is coherence. Darwinian evolution cannot pursue a future goal. So envisioning Darwinian evolution as akin to climbing a solitary staircase – even one with missing steps – risks a subtle, yet fatal misconception. It is all to easy to think of the top of the stairs as the target, and to focus exclusively on the path leading to it, ignoring all other possibilities. If a Darwinist visualizes steps leading to some biological feature, the temptation is to conclude the route would be easily traveled by unaided nature. However, as [Jerry] Coyne and [Allen] Orr emphasized, we need to ask whether a process is not just theoretically possible, but also biologically reasonable.
Because random mutation and natural selection have no goal, Darwinian evolution faces the huge problem of incoherence: Like a drunkard’s walk, the next evolutionary step a population of organisms takes is very likely to be unconnected to the last step. The upshot is that even if a gradual route toward a complex structure exists – even one with no missing steps – if the route is lengthy enough, the likelihood of reaching it by random mutation is terrible.
Michael Behe, The Edge of Evolution: The Search for the Limits of Darwinism (Free Press: 2007), pp. 112-113