Senescence

Why has evolution left bodies to fall apart with age when longer-lived individuals would leave more offspring?

A body builds itself from a single cell, replaces most of its own substance repeatedly, and heals wounds that would ruin any machine. Then, after a few decades, it stops keeping up. The puzzle is not that things wear out; it is that a system demonstrably capable of maintenance stops maintaining. If living longer means more offspring, selection should have fixed this long ago.

Two popular explanations are wrong and should go first. "Wear and tear" is not an answer: molecules are damaged constantly and constantly repaired or replaced, so the question is why repair is not simply kept up — and some organisms do keep it up, hydra showing no detectable increase in mortality with age, with certain rockfish and tortoises close behind. Nothing thermodynamic forbids it. "Ageing makes room for the young" is worse, because it asks selection to favour a trait that harms every individual carrying it, and because it fails on its own terms, as we are about to see.

Now the move that does the work. Ask what fraction of a wild population ever gets old. Suppose a small mammal has a 50% chance of surviving each year — predation, cold, disease, accident, nothing to do with ageing. Then the share of a cohort still alive after ten years is 0.5 raised to the tenth, which is about one in a thousand. Nine hundred and ninety-nine of a thousand animals never encounter their own old age at all.

Sit with what that does to selection. A gene whose harmful effect appears at age ten is expressed in one animal in a thousand; one that harms at age one is expressed in half of them. Selection against the first is roughly a thousandth as strong — not absent, but so weak that mutation supplies new damaging variants faster than selection removes them. Medawar's insight, formalised by Hamilton in 1966, is that the force of selection declines with age, and it declines for a reason having nothing to do with the biology of ageing: everything else that kills you.

That alone predicts a rubbish-heap of late-acting damage — mutation accumulation. But Williams added a second, sharper mechanism in 1957. Given that early effects are weighted so much more heavily, an allele that improves something early and ruins something late will be favoured, even if the late cost is severe. Ageing then is not merely neglect; it is partly the invoice for things that paid off when the animal was young. Kirkwood's third framing puts it in currency: repair costs resources that could go into reproduction, and when you are likely to be eaten anyway, the optimal investment in a body is deliberately less than "keep it forever."

Notice that the question's premise now looks different. "Longer-lived individuals would leave more offspring" is only true if they live long enough to use the extra time, and in the wild almost nobody does.

Can this be tested rather than merely told? In two directions. Reduce extrinsic mortality and senescence should slow: animals with escape hatches — bats, birds, turtles — outlive ground-dwelling mammals of similar size by a wide margin, and Austad's opossums on a predator-free island aged more slowly than mainland ones, though that study's effect size has been argued over. And force selection to act late: Rose bred Drosophila only from eggs laid by old flies, and within a few dozen generations had lines that lived substantially longer. Lifespan moved because the selection gradient moved.

Where is it genuinely unsettled? In the weighting. Mutation accumulation and antagonistic pleiotropy both predict senescence and are hard to tell apart in practice, and which dominates is still argued. And the comparative demography is messier than the theory's tidy version: a large survey across the tree of life found species whose mortality is flat with age and some whose mortality falls with age — typically organisms whose fertility keeps rising with size, so the selection gradient does not decline in the first place.

Where this sits next to its neighbours: the collection's piece on post-reproductive life is the mirror image of this one, and the two must be read together. This piece explains why the force of selection falls away after reproduction; that one explains the exception, where helping grandchildren keeps the gradient from reaching zero at menopause. It is precisely the case that tests the rule.

Three refinements. First, senescence is not programmed: nothing here describes a mechanism for ageing, only an absence of pressure to prevent it plus an active pressure for trade-offs that cost late. Second, the argument is about a population's statistics, not any individual — an animal that would have lived thirty years is not thereby fitter, because the environment almost never lets that potential be expressed. Third, "extrinsic mortality" means everything that kills independent of age, and lowering it is exactly what human societies have done, which is why we now meet a phase of life selection never had much to say about.

• What the negligibly-senescing organisms actually do differently, and whether their germ line and body are separable at all.

Every term the collection defines is gathered in the glossary.