It's not very often that creationists or the evolutionary biologists they attack make claims that I can test directly. When I was growing up, I tried digging up dinosaur bones in my parents' backyard, but I only found rocks and that digging is backbreaking work. But recently William Dembski, an "intelligent design" (ID) theorist, made a claim that I could test without having to do so much as put my boots on.
The claim concerns Weasel, a simulation that Richard Dawkins coded in the 1980s for his book The Blind Watchmaker
. In the simulation there is a series of generations in which a parent gives birth to children, one of which is selected to be the parent of the next generation. The parent of the first generation is a string of random text, but through random changes and selective pressure each generation of descendants looks more like a line of Shakespeare: "Methinks it is like a weasel." If you're thinking that this is in important respects different from natural selection, you're right, and Dawkins was the first to point this out. Dawkins only aimed to provide evidence for a rather modest claim: Random changes with selection can give results exponentially faster than random changes alone.
William Dembski recently raised questions about Dawkins' program
at a web site called Uncommon Descent
. Dawkins, as paraphrased by Dembski, claims it took 43 iterations to make his program give the line of Shakespeare the first time he ran Weasel, and it took 64 iterations the second time. Dembski points at Dawkins' abbreviated list of iterations and asks why we don't see any instances in which a letter that has been selected mutates again, this time away from the target text. He then says, "It is natural to conclude that that it [Weasel] is a proximity search with locking (i.e., it locks on characters in the target sequence and never lets go)." He then points to a portion of a documentary in which Dawkins runs a program in which changes can be seen to occur, even after the target is reached:http://www.youtube.com/watch?v=5sUQIpFajsg
Though he doesn't come right out and say it, the implicature is that Dawkins has overstated his claims about his program and that there has been some dishonesty on Dawkins' part.
Unfortunately for Dembski, this is all easily testable. I pounded out my own version of Weasel in one sitting. What did I find? If I print all the children to screen, I get results much like what we see in the documentary. However, if I print only the most adapted child of each generation to screen, I get results much like the ones Dawkins gave in his book. This is entirely consistent with what he said he was doing there. Dawkins didn't say what parameters he used, but when the chance of mutation for a given character is around 1 in 20 and the population size is around 100, I reach the target in little over 64 generations. Even when the population size is decreased, it's often the case that there will be runs in which the most adapted child of any given generation is in no respect less adapted than the most adapted child of the previous generations. What's more, even when there is drift, it's not at all unlikely that this will be corrected within 10 generations. If Dawkins reached the target in 43 generations, he may well have been using a larger population size, in which case it would have been even less likely that this sort of drift would be apparent. Having seen these results, I see nothing improbable about the lines of text Dawkins gave in his book. There is no reason to suppose that Dawkins made a locking program, and there's no reason to suppose that the program he used for The Blind Watchmaker
works differently from the one we see in the documentary. If Dembski wants to say this mutation-selection behavior is counter-intuitive, I'll grant him that, but it is what happens.
Kudos to Ian Musgrave of The Panda's Thumb
for bringing this matter to my attention.