Works by Josh Mitteldorf

In the summer of 1997, I happened on a new piece of information that took on a deep explanatory significance. The subject was: How has natural selection managed to evolve adaptations whose benefits seem to accrue only over many generations, and thus are dispersed over whole gene pools, not limited to the individual in which the adaptation first appeared? The new fact was actually a few years old, and came to me from a news article in Science, which mentioned the discovery of the eyeless gene in drosophila.

Of the careers I have passed through on the way to evolutionary biology, the most recent is computer programming. I was active in the late 1980's when serial-control programming languages evolved into object languages. Before that time, computer programs were frequently born of a single brilliant mind, and grew organically as features were added. The structure of programs from the early era is now referred to as "spaghetti code", because chains of execution thread tortuously through the body of code, guided by statements that say GoTo line xxx. After the revolution, computer languages had features to encourage the writing of "structured programs", with whole "objects" - their operations and control variables - segregated and isolated from each other.

For a one-time, quick-and-dirty programming project, the old languages were more efficient. But the new style was easier to debug; and for programs that must be adapted to new hardware or communications standards each year, modification in the new languages was safer: modifying code within one unit entailed much less risk of damaging the integrity of the other parts of the program.

Mammals and fruitflies share a gene named eyeless in flies and Pax6 in mice that dictates the inclusion of eyes in a developing embryo. The remarkable thing about the gene is the level on which it operates: it is capable of turning on and off the full mechanism for development of an eye, including optical structures, light detection and nerve hookups. Insertion of the gene into the area of a chromosome that controls leg or wing development causes an eye to grow on the leg or wing of a developing embryo. Insertion of multiple copies causes multiple eyes to appear.

Here's the revelation that struck me that morning in July: If chromosomes may be analogized as a genetic computer program for development and metabolism, then their organization may be more like new-style, object-oriented languages than the old-style spaghetti code. If the eyeless gene proves to be typical of a class of control structures, the implication is that genes are organized into functional units that can be called like subroutines by a small number of command genes.

How did chromosomes come to be organized in a top-down structure? Returning to my favorite topic, the new fact was both an enigma and a clue, another mystery and a hint about how all such mysteries might be resolved.

The body of evolutionary theory developed in this century implies that natural selection ought to be nearsighted. The success of a gene in penetrating a population ought to depend largely on its consequences for the viability and reproductive success of its individual bearer. The gene's long-term effect for progeny in an entire community or species ought to be a far less potent influence on natural selection. This is because mutations appear first in a single individual. The first test it must pass: can it spread to dominate a local population deme? Only a gene that succeeds at this level can ever be tested for its long-range effect on the population.

Nevertheless, there appear to be adaptations common among higher organisms that have clear long-term benefits for the population but serve no purpose for the individual - they may even be maladaptive for the individual. How did they evolve? How did they get by the "first test" screening? Broadly observed examples of such apparently long-range adaptations are sex and death: The sharing of genes in sexual reproduction offers no obvious advantage to the individual or his progeny, yet sex is the dominant mode of reproduction in higher organisms. Senescence and death of the individual have negative adaptive value for the individual, yet are near-universal features of multi-celled organisms. There are other examples as well that are less ubiquitous and less central to the life cycle, but are compelling nevertheless:

• Why are populations of a single species as genetically diverse as they often are?
• Why do most carnivores refrain from preying on their own species?
• Are cooperative and altruistic behaviors more broadly targeted than can be explained by the selfish interest of the individual (or his genes)?
• Are mutation rates optimized at a non-zero level that promotes evolutionary change?
• Are mutation rates higher in certain areas of the genome where they are more likely to result in new structures of positive adaptive value?
• Do mammal populations under stress exercise birth control to avoid the tragedy of the commons?

It has been the fashion among evolutionary theorists of the last thirty years to adopt exclusively the second approach. The line of reasoning about the "first test" was debated intensively in the late 1960's, and the argument against the cogency of long-range selection is now widely regarded as insuperable.

When I read about the eyeless phenomenon, I thought of it as another example of a ubiquitous adaptation with short-term cost and positive long-range value. The organization of chromosomes into structures that correspond to physiological functions in the fully-developed soma is a discipline that probably carries short-term costs, just as it takes the computer programmer longer to write tightly-structured, object-oriented programs. The payoff is in the long run, when evolution can try different variants of an adaptive system without compromising the integrity of the system. For example, the eyeless gene may facilitate genetic variation in which eyes are placed higher or lower in the head, without having to re-evolve the entire mechanism of the eye for each trial. The point is that each individual gets to try just one eye placement; there can be no adaptive advantage for him in a system which makes it easy to try other placements. That advantage accrues to the entire evolving population, and only over many generations.

So the eyeless phenomenon is another of the mysterious adaptations that has managed to emerge despite its apparent lack of short-range advantage. But here is how it may also be a key to approaching all such mysteries:

If this is an example of a general phenomenon, then it may be that the process of evolution is itself highly evolved. Biological systems may be replete with adaptations that promote the efficient operation of evolution. If so, an important class of such adaptations would soften the impact of the "first test", permitting some marginal adaptations to spread through a population when they are neutral or even slightly detrimental to the individual, so that their long-range effect may be tested.

If the mechanism of evolution has indeed been optimized by its own process of adaptation and selection, there is then a candidate for the theorist's "first approach" to the problem of long-range selection. Evolution has found ways to blunt the force of individual selection, enforcing a high level of diversity and a slower, broader pattern of natural selection so that adaptations with a long-range value have a greater chance of prevailing over short-term expedients. One of the beneficiaries of this regime is the very process that created it, the evolution of evolution.

The mystery of long-range adaptation is then reduced to a bootstrapping problem: How did the process of evolution of evolution get started? For an answer to this question we may draw clues from two sources:

First is the work of Sewall Wright on a conceptual framework he espoused all his life, and referred to as the shifting balance theory. Wright theorized that genetic drift within small population demes could blunt the determinism that might otherwise drive evolution into adaptive blind alleys. The same mechanism could serve for the closely related problem of blunting the force of individual selection long enough for group selection and long-range selection to become a factor.

Second, is the insight we derive from the new science of complexity. Interactions among many simple systems may produce results on the global scale that are hypersensitive to details on the fine scale. Herein is another potential escape from the determinism of short-term individual selection.

I have since learned of a paper in Evolution by Gunter Wagner and Lee Altenberg in which these same ideas are developed formally and in more detail, but without mention of the implications concerning group selection. Their paper is entitled Complex Adaptations and the Evolution of Evolvability.  I am encouraged that there is a blossoming interest in the evolution of evolvability, but puzzled that the researchers in this field don't relate their work to the problem of group selection.