After I had finished a colloquium on chaotic population dynamics and the evolution of aging, a sharp grad student raised her hand and asked, "At the end of the day, what matters is this: are there more genes for aging or less genes for aging in the next generation?"

Here's what I wish I'd had the presence of mind to respond to her:

Yes, we've learned to think that that's the only thing that matters: the incremental change in gene frequency. But is that a logical necessity, or is that our model of how evolution works? If it is the model we have used all our professional lives, is it still possible for us to conceive that the biosphere might work differently from our model?

Remember where the physics community was 100 years ago. Suppose you asked a physicist about this situation: There's a train moving along the track at 50 mph. Inside the train car. a man throws a ball forward at 50 mph. How fast is the ball moving?

The "obvious" answer is 100 mph. This seems to be a self-evident fact from logic, requiring no experiment to validate it. If you're already moving at velocity V1 and you add a velocity V2 then the resultant velocity is V1 + V2.

It cannot be otherwise, or so it seemed. But 99 years ago, Einstein was broad-minded enough to regard this as an experimental question, not a logical fact. And he proposed that when you "add" V1 to V2, the result is not exactly V1 + V2, but .

When Fisher (and Wright and Haldane) invented the science of population genetics, they made the natural assumption that gene frequency could be approximated by a continuous variable, and that computing the differential change in gene frequency would be a good way to understand evolution. Certainly, that was the hypothesis to test first. In the intervening decades, we have built a mathematical edifice on the foundation of this postulate.

But how much of that edifice has consequences that can be validated quantitatively in field studies? And how many times have we had to elaborate, to create special cases, or extend the theory in ad hoc ways to agree with observations? How sure are we that the foundation of this edifice is solid?

I propose that when species interactions and ecological dependencies are taken into account, maximization of individual reproductive value will result quite generally in chaotic population dynamics. The fact that there are stable ecosystems in nature is not a logical or mathematical necessity, but a result of adaptation at the population level.

This hypothesis leaves intact some of the differential behavior of gene frequencies that we hold so dear; but individual reproductive value is not optimized globally. First, whole ecosystems are adapted for demographic homeostasis. Only within this context does individual fitness trend upward.

With the hypothesis that population dynamics are inherently chaotic, we can begin to come to terms with four great paradoxes that have provoked controversy and evaded consensus these many years:

• The evolution of sex
• The evolution of aging
• Reproductive restraint in predator-prey systems
• The persistence of genetic diversity
But I'm asking that we give up on individual reproductive value as a quantity that is globally optimized.  The imperative for demographic homeostasis is a primary agent of natural selection, rivaling reproductive potential in its import.

Perhaps 95% of evolutionary biologists think that individual selection is all the selection that there is.  5% have argued bravely that group selection can sometimes overcome the more direct effect of individual selection.  In this landscape of opinion, I am staking out a position on the far left wing, claiming that evolution not just of populations but of entire ecosystems is the rule rather than the exception.