Sexual Selection
or, why peacocks should have gone extinct, and didn’t
Look at a peacock’s tail and try to explain it.
The bird drags around six pounds of iridescent feathers it cannot shed, cannot hide, and cannot fold into anything aerodynamic. The tail makes the bird slower. It announces its location to every fox within two hundred meters. It costs an enormous amount of metabolic energy to grow each spring. By any reasonable accounting the tail is a death warrant the bird carries around in case it forgets where to find one.
Darwin himself looked at peacocks and got nervous. In a letter to Asa Gray in 1860 he wrote: The sight of a feather in a peacock’s tail, whenever I gaze at it, makes me sick! Sick because his theory of natural selection said useful traits should propagate and useless ones should not, and the tail was so spectacularly useless it bordered on satirical. A peacock with a slightly smaller tail would be a peacock with a slightly better chance of escaping a fox, raising slightly more chicks, and propagating its slightly-smaller-tailed genes. Over generations the tails should shrink. They do not. They have, if anything, gotten more extravagant over evolutionary time.
Darwin worked out an answer eventually, in The Descent of Man in 1871, and called it sexual selection. The idea was that in some species the limiting factor on reproduction is not whether you can survive but whether you can mate, and that mate choice can drive trait evolution in directions that survival selection never would. Specifically: peahens prefer extravagant tails. A peacock with a more spectacular tail mates more, regardless of how short his life is. The genes for extravagant tails get propagated. Sexual selection runs in the opposite direction from survival selection, and where the two collide, the winner is whichever one matters more for getting genes into the next generation.
This was a reasonable answer but it left a deeper question hanging. Why do peahens prefer extravagant tails? What makes that preference adaptive? Darwin himself was uncomfortable with the answer and largely punted on it; the puzzle would not be properly resolved for another sixty years.
The mathematician Ronald Fisher, in 1930, gave the resolution that still anchors the modern theory. He pointed out that the trait (tail size) and the preference (the peahen’s taste for tail size) are genetically linked through the act of mating. A peahen with a strong preference for big tails will mate with a big-tailed peacock. Their daughters will inherit both genes for big-tail-preference (from mother) and genes for big tails (from father, if female). Their sons will inherit both genes too. This means the trait gene and the preference gene end up correlated in the population. They co-vary.
And once they co-vary, you have a positive feedback loop. Selection on the trait drives evolution of the preference, and selection on the preference drives evolution of the trait, because the genes are statistically tied together. The result is what Fisher called runaway selection: tail length and tail-length preference both drift toward extremes. A bigger tail attracts more mates; a stronger preference for bigger tails ensures that the next generation has even more sons with big tails and even more daughters with strong preferences. The system feeds on itself. Survival cost is the only thing slowing it down.
The runaway is the engine of the peacock’s tail. It is also the engine of a great deal more. Bowerbirds — small-bodied creatures that build elaborate stick-and-feather constructions on the forest floor, sometimes the size of a couch, decorated with carefully arranged colored objects, sometimes including a deliberate use of forced perspective — build their bowers because females have evolved a preference for elaborate bowers and the genes for building-elaborate-bowers and the genes for preferring-elaborate-bowers are linked in the population. The bower is useless to the bird. The bower is everything to the gene that codes for the bower.
The Israeli biologist Amotz Zahavi, in 1975, added a complementary mechanism known as the handicap principle. Zahavi pointed out that costly traits make particularly good honest signals because cheaters cannot afford them. A peacock who can survive while dragging six pounds of feathers around is, by demonstration, a peacock with strong genes underneath. The tail is reliable advertisement precisely because it costs so much; a low-quality peacock simply cannot pay the bill. Female preference for the tail tracks underlying quality through the proxy of "can he afford this." Zahavi’s theory and Fisher’s theory are not exclusive; both are running in real populations, often at the same time.
From the gene’s point of view, none of this is mysterious. The genes that code for big tails are getting copied into the next generation; the genes that code for small tails are not. The bird carrying the big tail is irrelevant except as a vehicle. The fact that the tail kills him slightly faster than a small tail would is not the gene’s problem — the gene has already been copied by then. Sexual selection is one of the cleanest examples of the selfish-gene logic applied: the trait is for the gene, not for the bird, and the bird is whatever the gene built to make more genes.
This is also the mechanism that, when applied across species and across substrates, ends up explaining the peacock’s feathers, the bowerbird’s bower, the elk’s antlers, the firefly’s flash, the songbird’s song, the deer’s rut, the cricket’s chirp, and on and on. Anywhere mate choice matters, sexual selection is doing some of the evolutionary work, and where the preferences and the traits are genetically linked, runaway is on the table.
It also explains, with a slight extension, a great deal of human behavior — particularly the parts of human behavior that look extravagant, costly, and seemingly useless from a survival perspective. The page on Beauty picks up where this one ends; the cosmetics industry is, among other things, a multi-billion-dollar response to sexually-selected human visual perception. But the underlying mechanism is the one Fisher described in 1930. Trait and preference, linked through mating, in positive feedback, with the only brake being the cost of being killed.
One more layer worth flagging before the experiment. Sexual selection sits inside a broader frame of strategic interactions between agents whose interests overlap and conflict at the same time — which is the territory of game theory. Specifically: when survival selection and sexual selection pull in different directions, the question of which one wins is not a fixed answer but the equilibrium of a game between strategies. Dawkins, in the Battle of the Sexes chapter of The Selfish Gene, walks through one such game — coy-versus-fast females paired with faithful-versus-philanderer males — and shows that you get a stable mixed equilibrium where some fraction of each strategy persists in the population. One of the clean predictions of that math is that some percentage of females in any given population will not have a steady mate, because the equilibrium contains both strategies and not because something has gone wrong. The page on game theory (forthcoming) develops this further. For now, just notice: sexual selection produces extreme traits, but who carries those traits and who chooses them is itself the result of a many-player game running underneath the trait dynamics.
The Experiment
Below is a small population of two hundred agents, each one carrying two genes: a trait value (think of it as tail length) and a preference value (think of it as preferred tail length in a mate). Half the population is female and half is male, by convention; the underlying math is symmetric. Each generation, females rank male candidates by how well their trait matches the female’s preference, mate selectively, and produce offspring. The offspring inherit averaged genes plus a small mutation. Survival cost on extreme traits acts as a brake.
In the arena below, females are on the left and males on the right. Each female is drawn as a small dot with a soft halo whose radius is proportional to her preference — the bigger the halo, the stronger her taste for elaborate males. Each male is drawn as a dot with a tail trailing behind him whose length is proportional to his trait. When mating happens, a faint line briefly connects the chosen pair. Watch the tails grow and the halos grow over generations as the runaway pulls them along together. The trace below shows the population means.
Things to try:
Set preference strength to 0. The males’ tails stay short; the females’ halos stay small. Without any preference, females mate at random, and there is no sexual selection — just drift held near zero by the survival cost.
Set preference strength to about 2.5 and survival cost to 0.05 (the default). Wait. The first thirty or forty generations look like nothing is happening — the trait and preference need to become statistically correlated through mating before the runaway can kick in. Then both start to grow, slowly at first, then visibly faster. Watch the tails get longer and the halos get wider together.
Crank survival cost up to 0.3. The runaway slows or stops; tails settle at a modest length; halos stay modest. The cost of carrying a long tail balances the reproductive bonus, and the population finds a lower equilibrium. This is roughly why peacocks aren’t the size of small airplanes.
Drop survival cost to nearly zero with strong preference. The tails grow visibly until the simulation hits its display ceiling and auto-resets. In real biology this rarely happens because real costs are non-negotiable, but in the math the runaway is unbounded.
While the runaway is happening, watch the faint orange lines that appear briefly each generation. Those are mating events — one female (left) chose one male (right) to be the father of her offspring. Notice that the lines tend to connect halos to long tails. That bias is what couples the two genes and drives the feedback. No coupling, no runaway.
Sexual selection is one of the few evolutionary forces that can produce traits in opposition to survival. Most of evolution is about not dying. Sexual selection is about not failing to mate, which is a different optimization with a different gradient, and the two often pull in different directions. Sally and Harry, when they pay attention, are running both at once. So is every peacock. So is every bowerbird. The peacock is the dramatic case because the costs are so visible. The mechanism is universal.