If a harmless animal wants to play it safe, it may mimic something dangerous. The nonvenomous scarlet kingsnake, for example, closely mimics the colors and patterning of the venomous eastern coral snake. First proposed by Henry Bates in 1862, Batesian mimicry is widely accepted -- and often used to illustrate natural selection, Charles Darwin's mechanism of evolution -- despite having never been experimentally proven.

David Pfennig finds this lack of experimental support surprising. So Pfennig, an associate professor of biology, recently performed a series of experiments to put Batesian mimicry to the test.

Along with his wife, Karin, a postdoctural student at the University of Texas, and Carolina undergraduate David Harcombe, Pfennig set out to test Batesian mimicry's critical prediction -- that harmless mimic species should enjoy protection from predators only in those areas where the dangerous species, known as the model, occur. "It's a very simple prediction," Pfennig said. "If you get out of the area where the dangerous models are present, then the protection should break down."

That's because Batesian mimicry is a predator-driven system. Imagine a hungry raccoon near Wilmington, where both scarlet kingsnakes and coral snakes are found. This raccoon is a result of generations of evolutionary selection to avoid the poisonous coral snake -- and anything that looks like it. So it probably sees the scarlet kingsnake as a threat. But here in Chapel Hill, where coral snakes are not found, raccoons would not have been under selection to avoid coral snakes. Chapel Hill raccoons would therefore have no good reason to avoid the kingsnake, which is a potential meal.

Pfennig and his team identified dozens of sites in North and South Carolina, both where coral snakes and scarlet kingsnakes occur together, and where the coral snake does not occur. At each site, the team left three snake replicas. One was designed to look like the scarlet kingsnake, with ringed bands of red, black, and yellow.

The other two replicas were created not to resemble any particular snake, but to serve as controls. One was plain brown. A third replica used the same colors and in the same proportions as the ringed replica, but this time arranged in stripes down the length of the "snake." "If we used only ringed and plain replicas," Pfennig said, "we wouldn't know whether predators just have a generalizable avoidance of bright colors. So we wanted to make striped replicas in the same colors as the ringed, just to make sure that wasn't happening."

The replicas were made from plasticine, a non-toxic modeling clay which retained bite and scratch marks from any animal attempting to make a meal of one of the fake snakes. "We knew predators would likely be eating some of the clay," Pfennig said, "so we used an inert substance that wouldn't harm them."

Pfennig and his team collected the replicas after one month and brought them back to Chapel Hill, where Karin Pfennig, who had no knowledge of each replica's original site, tallied the bites, scratches, and gouges on each. About 5 percent of the replicas were attacked.

"We wanted to examine what proportion of all the attacks were on the ringed replicas," Pfennig said. "If this is truly mimicry, the proportion of attacks on ringed replicas should be lower in areas where there are coral snakes than in areas where there are no coral snakes."

The proportion of kingsnake replicas attacked was significantly greater in areas where coral snakes do not occur--about 65 percent, compared to only eight percent attacked in areas where coral snakes occur.

The team repeated the experiment in Arizona. This time they fashioned a replica of the sonoran mountain kingsnake, which mimics Arizona's poisonous western coral snake. They used the same striped and brown control replicas. Their Arizona findings were consistent with the North Carolina numbers: about 50 percent of attacked replicas were kingsnakes placed in areas where western coral snakes do not occur. Only 13 percent were attacked in areas where coral snakes occur.

This suggests that predators avoid coral snake mimics only in areas where coral snakes live, Pfennig said. Bates appears to have been right. "You get a lot of resemblances in nature that are not due to mimicry," Pfennig said. "But in this experiment we found a very clean pattern, which suggests that there's no other plausible interpretation."

The team's results appeared in the March 15 issue of Nature.