This quick article by Jonathan Losos and colleagues is entirely unsurprising, but good to read:
Rapid Temporal Reversal in Predator-Driven Natural Selection
Because of its potentially epochal scope, evolutionary biology is often caricatured as a strictly descriptive science, but recent years have shown that evolution can be studied on short time scales and that evolutionary biology can be both experimental and predictive. Here, we report just such an example by demonstrating the occurrence of a predicted reversal in the direction of natural selection on limb length in Anolis sagrei, a common Bahamian lizard often found on the ground in the absence of terrestrial predators.
Previous research showed that, when a larger and entirely terrestrial predatory lizard, Leiocephalus carinatus, invades, A. sagrei becomes more arboreal and that the extent of this habitat shift broadens through time (3). Hence, we predicted that the direction of selection operating on limb length in A. sagrei would change through time in the presence of L. carinatus (4): Initially A. sagrei occurs mostly on the ground, so individuals with relatively longer legs, being faster (5), would be better able to elude the predators and thereby be favored. As A. sagrei becomes more arboreal, however, we predicted that selection would favor the reverse because shorter limbs are better suited for movement on the narrow and irregular surfaces A. sagrei would use to avoid the terrestrial predator (5).
This is quite a scenario. Long-legged lizards can run faster on the ground, so initially the short-legged lizards get eaten. But in the long term, the ground is not viable as anole habitat in the presence of predation, so over time individuals that exploit trees do better. And short legs enable them to climb better.
You may not have guessed it from the fairly anaesthetic title, but this is an experiment where they basically let the velociraptors loose:
To test this hypothesis, we introduced L. carinatus to six small Bahamian islands that naturally contained A. sagrei, randomly choosing six others to serve as controls (L. carinatus occurs on nearby larger islands and is known to colonize smaller islands); the number of L. carinatus introduced (all adults) was proportional to the number of A. sagrei resident on the island. Before introduction of L. carinatus, A. sagrei individuals on each island were measured and individually marked. Islands were exhaustively censused after 6 and 12 months to determine survival (6).
Can you imagine the look on the anoles' faces? Well, I guess lizard victims are less sympathetic. Maybe if they were birds and Bolivian tree lizards came to eat their eggs?
The bottom line of the paper is that this kind of rapid reversal can occur when behaviors are plastic -- in this case, the adoption of higher arboreality takes a while to kick in, but then selects for features not adaptive on the ground.
I wouldn't really call this case a reversal of selection, though. When the bad lizards show up, surely the subset of the anole population that was already choosing more arboreal substrates had an immediate advantage. The brief increase in leg length represents disruptive selection -- the anoles that do the worst are the short-legged terrestrial ones, and long-legged terrestrial anoles do well only so far as the short-legged ones are taking all the heat.
Still you don't run across the dynamics of disruptive selection every day.
Losos JB, Schoener TW, Langerhans RB, Spiller DA. 2006. Rapid temporal reversal in predator-driven natural selection. Science 314:1111. DOI link