Lab ecosystems show signs of evolving.

A test of natural selection among ecosystems miniaturized to the size of flowerpots bolsters the formerly shocking concept of group selection, argues a group of New York researchers.

Group selection holds that natural selection operates on groups, such as diverse ecosystems, just as it does on individuals, to select traits that lead to higher reproductive success.

Previous tests of group selection focused on only one or a few species, explains longtime proponent David Sloan Wilson of Binghamton (N.Y.) University. Now he and his Binghamton colleagues have tracked changes at the level of ecosystems with thousands of species and millions of individuals, all of them in small containers.

The mini-ecosystems showed evidence of passing traits to "offspring" ecosystems, Wilson, William Swenson, and Roberta Elias report in a paper scheduled for a forthcoming PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES.

The practice of analyzing groups as though they were individuals has a roller-coaster history, Wilson observes. During the first half of the 20th century, no eyebrows rose if a researcher described a field ecosystem as "maturing" to an "adult" forest. Tough analysis in the 1960s, however, discredited the mechanisms proposed for explaining how evolution could work on the level of groups. To critics at the time, Wilson says, "selection at an individual level always trumped selection at a group level."

Since then, researchers say they've demonstrated group selection in the laboratory. For example, William M. Muir of Purdue University in West Lafayette, Ind., selected cages of hens that as a whole yielded more eggs than other caged groups. This approach boosted egg laying to 160 percent of the original strain's and produced milder-mannered hens.

The Binghamton team collected forest soil to inoculate containers of sterilized soil and planted them with a lab strain of mustard. To create the next generation, the researchers inoculated another set of sterilized containers with soil from chosen parents. In part of the experiment, they awarded parenthood to those ecosystems that produced either the biggest or the puniest of the mustards. A parallel experiment selected generations of a pond ecosystem for pH.

Over the course of generations, the differences in the selected traits between two lines intensified, then diminished, and then intensified again. Wilson argues that such fluctuating behavior fits the pattern expected for complex systems, such as ecosystems, that are evolving away from each other.

However, Douglas S. Stoner of the University of South Carolina in Columbia says that such variations raise questions about how long the differences might last. He's worked on natural selection among cell lines in sea squirts, and he accepts the basic premise that evolution can occur at different levels.

Purdue's Muir welcomes the work of the Binghamton group warmly. Says Muir: "I hope this will refocus the debate back on wholes rather than on parts."