It was S. J. Gould’s contention that evolution is a contingent process: its path is contingent on historical events, so that if you “re-wound the tape” of evolution and started it again from an earlier point, you would not get exactly the same results.
Well, historical contingency — and macroevolution — have both been demonstrated by Richard Lenski in a long-running laboratory experiment using the bacterium Escherichia coli. The results were published June 4th in PNAS. In addition, in case irreducible complexity needed another nail in its coffin, the experiments demonstrate the accumulation of mutations over years that result in a novel change requiring more than one step.
The experiment started with one bacterium. Its daughters were divided into twelve populations. The bacteria were bred in containers with a small amount of glucose, which they could use for food, and a large amount of citrate, which they could not. The experiment ran for twenty years and more than 44,000 generations. Every 500 generations, the researchers removed a sample and froze it. These samples can be revived and allowed to multiply again.
First notable result: one population out of the twelve, after more than 31,000 generations, evolved the completely new ability to use citrate as food.
Second notable result: samples from that population, after 20,000 generations but not before 15,000 generations, are more likely to re-develop the ability to use citrate. None of the other populations have evolved that ability.
Conclusions: first, macroevolution has been observed again. This new ability makes the E. coli virtually a new species. Second, something happened between the generation 15,000 and generation 20,000 that made it easier for the bacteria to evolve the new ability. They were accumulating changes that pre-adapted them to the next step.
Read all about it on Pharyngula: Historical contingency in the evolution of E. coli and in Kasama: Evolution documented in the lab.