2008 : WHAT HAVE YOU CHANGED YOUR MIND ABOUT? WHY?

In 1975, as a newly minted PhD who had just published my first paper on group selection, I was invited by Science magazine to review a book by Michael Gilpin titled Group Selection in Predator Prey Communities. Gilpin was one of the first biologists to appreciate the importance of what Stuart Kauffman would call "the sciences of complexity." In his book, he was claiming that complex interactions could make group selection a more important evolutionary force than the vast majority of biologists had concluded on the basis of more simple mathematical models.

Some background: Group selection refers to the evolution of traits that increase the fitness of whole groups, compared to other groups. These traits are often selectively disadvantageous within groups, creating a conflict between levels of selection. Group selection requires the standard ingredients of natural selection-a population of groups, that vary in their phenotypic properties in a heritable fashion, with consequences for collective survival and reproduction. Standard population genetics models give the impression that groups are unlikely to vary unless they are initiated by small numbers of individuals with minimal migration among groups during their existence. This kind of reasoning turned group selection into a pariah concept in the 1960's , taught primarily as an example of how not to think. I had become convinced that group selection could be revived for smaller, more ephemeral groups that I called "trait groups." Gilpin was suggesting that group selection could also be revived for larger, geographically isolated groups on the basis of complex interactions.

Gilpin focused on the most famous conjecture about group selection, advanced by V.C. Wynne-Edwards in 1962, that animals evolve to avoid overexploiting their resources. Wynne-Edwards had become an icon for everything that was wrong and naïve about group selection. Gilpin boldly proposed that animals could indeed evolve to "manage" their resources, based on non-linearities inherent in predator-prey interactions. As resource exploitation evolves by within-group selection, there is not a gradual increase the probability of extinction. Instead, there is a tipping point that suddenly destabilizes the predator-prey interaction, like falling off a cliff. This discontinuity increases the importance of group selection, keeping the predator-prey interaction in the zone of stability.

I didn't get it. To me, Gilpin's model required a house-of-cards of assumptions, a common criticism leveled against earlier models of group selection. I therefore wrote a tepid review of Gilpin's book. I was probably also influenced by a touch of professional jealousy, as someone who myself was trying to acquire a reputation for reviving group selection!

I didn't get the complexity revolution until I read James Gleik's Chaos: Making A New Science, which I regard as one of the best books ever written about science for a general audience. Suddenly I realized that as complex systems, higher-level biological units such as groups, communities, ecosystems, and human cultures would almost certainly vary in their phenotypic properties and that some of this phenotypic variation might be heritable. Complexity theory became a central theme in my own research.

As one experimental demonstration, William Swenson (then my graduate student) created a population of microbial ecosystems by adding 1 ml of pond water from a single, well-mixed source to test tubes containing 29 ml of sterilized growth medium. This amount of pond water includes millions of microbes, so the initial variation among the test tubes, based on sampling error, was vanishingly small. Nevertheless, within four days (which amounts to many microbial generations) the test tubes varied greatly in their composition and phenotypic properties, such as the degradation of a toxic compound that was added to each test tube. Moreover, when the test tubes were selected on the basis of their properties to create a new generation of microbial ecosystems, there was a response to selection. We could select whole ecosystems for their phenotypic properties (in our case, to degrade a toxic compound), in exactly the same way that animal and plant breeders are accustomed to selecting individual organisms!

These results are mystifying in terms of models that assume simple interactions but make perfect sense in terms of complex interactions. Most people have heard about the famous "butterfly effect" whereby an infinitesimal change in initial conditions becomes amplified over the course of time for a complex physical system such as the weather. Something similar to the butterfly effect was occurring in our experiment, amplifying infinitesimal initial differences among our test tubes into substantial variation over time. A response to selection in the experiments is proof that variation caused by complex interactions can be heritable.

Thanks in large part to complexity theory, evolutionary biologists are once again studying evolution as a multi-level process that can evolve adaptations above the level of individual organisms. I welcome this opportunity to credit Michael Gilpin for the original insight.