Now, a research team has, usefully, come up with estimates of the probability of mutations being reversible. From ScienceDaily (May 11, 2011):
Physicists’ study of evolution in bacteria shows that adaptations can be undone, but rarely.
[ … ]
Jeff Gore, assistant professor of physics at MIT, says the critical question to ask is not whether evolution is reversible, but under what circumstances it could be. “It’s known that evolution can be irreversible. And we know that it’s possible to reverse evolution in some cases. So what you really want to know is: What fraction of the time is evolution reversible?” he says.
By combining a computational model with experiments on the evolution of drug resistance in bacteria, Gore and his students have, for the first time, calculated the likelihood of a particular evolutionary adaptation reversing itself.
They found that a very small percentage of evolutionary adaptations in a drug-resistance gene can be reversed, but only if the adaptations involve fewer than four discrete genetic mutations. The findings will appear in the May 13 issue of the journal Physical Review Letters. Lead authors of the paper are two MIT juniors, Longzhi Tan and Stephen Serene.
[ … ]
In the late 19th century, paleontologist Louis Dollo argued that evolution could not retrace its steps to reverse complex adaptations — a hypothesis known as Dollo’s law of irreversibility. Gore says his team’s results offer support for Dollo’s law, but with some qualifications.
“It’s not that complex adaptations can never be reversed,” he says. “It’s that complex adaptations are harder to reverse, but in a sense that you can quantify.”
*This from one 2007 paper, “Reversing opinions on Dollo’s Law”:
Dollo’s Law, the idea that the loss of complex features in evolution is irreversible, is a popular concept in evolutionary biology. Here we review how application of recent phylogenetic methods, genomics and evo-devo approaches is changing our view of Dollo’s Law and its underlying mechanisms. Phylogenetic studies have recently demonstrated cases where seemingly complex features such as digits and wings have been reacquired. Meanwhile, large genomics databases and evo-devo studies are showing how the underlying developmental pathways and genetic architecture can be retained after the loss of a character. With dwindling evidence for the law-like nature of Dollo’s Law, we anticipate a return to Dollo’s original focus on irreversibility of all kinds of changes, not exclusively losses.
So it’s not a law, just a hurdle?