They reproduce, they evolve and they mutate, all the while competing with other species for scarce resources. And if they fail to adapt, they die. It's a bug's life.

But not just any bug. These bugs are not actually "living" creatures at all--they're computer programs that function like real organisms. They live in a virtual world created in a computer's memory by an elaborate master program--in effect mimicking life.

Researchers released the digital organisms in their computer environment to watch them "evolve" and record the results in a controlled experiment, the most elaborate use to date of the "artificial life" technique.

"We simulated a world," said Christoph Adami, a senior research fellow in computation and neural systems at the California Institute of Technology. "These creatures have no idea that there's any other world but this. The world is unreal but the creatures that live in it are not."

While scientists have created software that imitates evolution for the past decade, the new work, reported today in the scientific journal Nature, takes artificial life out of the realm of scientific curiosity and puts it to work as a research tool.

As a first step, the team created a digital organism composed of a string of relatively simple computer instructions. The organism was a "bug," said Michigan State University microbiologist Richard Lenski, the team leader, and the string was the cyber-equivalent of "DNA" because it carried instructions for all the organism's functions.

The digital organism's program contained a copy command, so that the organism would "reproduce" itself, much like a computer virus, but unlike the virus, the digital organisms only functioned in the virtual environment the computer created for them.

Moreover, because the copy function had a built-in imperfection, a copy of the organism would have an error, or "mutation," in it about every 1,000 times.

Sometimes the mutation would allow the organism to perform a mathematical function that the environment "rewarded" by allowing the organism to reproduce faster, crowding out its rivals and, perhaps, driving them to "extinction." This was the computer equivalent of "natural" selection.

The team let the first bug evolve into 200 new bugs, or "species," manipulating the environment so some species remained "simple"--able only to replicate themselves--while others adapted and grew more "complex."

In the second stage of the experiment, the team watched the organisms evolve as mutations were added, then measured their "fitness"--their ability to reproduce during successive cycles.

The simpler organisms "fell apart at a faster and faster rate" as they mutated, Lenski said, but the more complex ones were "buffered" and far less affected.

The results suggest that "there seem to be governing principles that exist in life," whether they are demonstrated in biochemistry or in cyberspace, Adami said.

The team discovered that the more complex digital organisms could absorb random mutations with fewer ill effects than simpler ones, and that mutations did not necessarily have cumulative effects on an organism--that "the whole could often be quite different from the sum of its parts," Lenski said.

Both findings supported evidence gathered by biologists studying bacteria, fungi and fruit flies, and suggested "research moves you might make in real life," Lenski added. Whether scientists will embrace "artificial life" as a research tool, however, is far from clear.

"Field biologists are dealing with the idiosyncrasies of the species they're looking at, and they tend to be wary of all-embracing theories," said Inman Harvey, a neural science and robotics specialist at Britain's University of Sussex. "This approach may be taking generality and abstraction rather further than most biologists are willing to go."

Lenski said he understood the work was controversial, but said he is "happy" regardless of how the coming debate turns out: "If digital organisms behave like real organisms, that suggests that computer genetic code is like human genetic code, and that's real interesting," he said.

Interest in creating artificial life grew from the development of computer modeling--using computers to anticipate the likely consequences of events ranging from a 1-degree rise in ocean temperature to nuclear war.

An important advance came in the early 1990s when zoologist Thomas Ray created a program supple enough to allow smaller digital organisms to replicate themselves and mutate to fill the virtual environment without crashing it.

This went beyond simulation, for the master program was designed simply to accommodate the activities of the digital organisms, which "evolved" on their own. The master program did not react to the organisms or impose an outcome.

Adami and research colleague Charles Ofria took the concept several steps further. Funded by the National Science Foundation and Microsoft Corp., the researchers equipped their master program so it could be used to perform actual experiments in evolution, and teamed with Lenski, who had done extensive evolutionary studies on bacteria.

There were huge potential advantages to a computerized approach. While Lenski could produce mutations in the lab and watch their effects on microbial evolution, the process was laborious, time-consuming, hard to control and subject to error.

By contrast, with the computer "you can replicate your experiments beyond your wildest dreams," without worrying about time or complexity, Lenski said. "Historically in genetics, we have looked at one mutation at a time," he added, but with the computer "we introduced literally millions into each species."

The team plans next to create "isolated environments" within the soup bowl, he said, and run the evolutional clock for an entire ecosystem, adding or subtracting species, and seeing how species react to "perturbations"--the computer equivalent of an earthquake.