Keystone species

How can removing one uncommon species reshape an entire ecosystem?

The intuitive accounting for an ecosystem is by weight. The species that dominate the biomass — the grass, the trees, the vast shoals — should be the ones that matter, and a scarce animal contributing a fraction of a percent of the living tissue should matter proportionately little. Remove it and you might expect a small dent.

Sometimes you get a small dent. Sometimes the community rearranges itself entirely. What distinguishes the two cases, and why should abundance be such a poor guide?

The founding experiment is worth telling properly, because it is unusually clean. In the mid-1960s Robert Paine worked on the rocky shore of the Washington coast, where the intertidal zone held a mixed assemblage — mussels, barnacles, limpets, chitons, several seaweeds — and one predatory starfish, Pisaster ochraceus. Paine did not model anything. He picked plots and physically threw the starfish off them, repeatedly, for years, and watched.

The mussels took over. Freed from their principal predator, they spread across the rock and squeezed nearly everything else out, and the number of species in the cleared plots fell by roughly half. The starfish had not been feeding the community or building its habitat. It had been eating the single best competitor for the one resource in genuinely short supply on a rock face: space.

Sit with that, because it is the whole mechanism. What made the starfish powerful was not its own mass but its position — it sat on top of a competitive hierarchy that would otherwise collapse into a monoculture. Its predation was the only thing preventing one species from winning. Ask what that implies and a general principle appears: influence in a community flows through whom you interact with, and a small animal that constrains a dominant one is leveraging that dominant one's mass, not its own.

Does the leverage always run through predation? No — and that is a good check on the idea rather than a complication of it. A beaver is a keystone by construction, converting a stream into a pond and creating habitat that nothing else in the system provides. Fig trees in some tropical forests fruit out of season, so an uncommon tree sustains a large fraction of the frugivores through the lean months. In each case, take the species away and something no other member supplies goes missing.

And what makes the effect large rather than merely real? Feedback. Mussels that gain space produce more mussels that take more space; that is a self-reinforcing loop with nothing to damp it once the predator is gone. Keystone effects are big because removing the constraint lets a loop run.

The definition worth keeping is the one the ecologists eventually converged on: a keystone species is one whose effect on the community is disproportionately large relative to its abundance. Both halves are load-bearing. A dominant species with a huge effect is not a keystone — it is simply dominant, and calling it one drains the word of content. That is the most common misuse: "keystone" gets applied to anything ecologically important, at which point it stops distinguishing anything.

There is also a serious honesty problem in how the concept is popularised, and the best-known example is the one to be careful with. The story that wolves returned to Yellowstone in 1995, suppressed elk browsing, let willow and aspen recover, brought back beaver, and thereby changed the rivers is a genuinely contested claim in the ecological literature, not an established result. Several researchers have argued the vegetation recovery is patchy, that elk numbers fell for reasons including hunting, drought, bears and bison, and that the causal chain has been asserted more confidently than the data support. Wolves plainly had effects; the tidy cascade is a hypothesis under active dispute. Repeating it as settled misrepresents the field — and, less obviously, it is a weaker case than Paine's for exactly the reason that made Paine's strong: nobody ran the controlled removal, because you cannot fence a landscape.

Which points at the real difficulty. Keystone status is only demonstrated by removal, and removals are either accidental, historical, or unethical. Most claimed keystones are inferences from correlation, and the confidence attached to them should vary accordingly.

• Why some communities reorganise into a new stable state after a keystone is lost, and refuse to return when it comes back.
• How conservation should prioritise when keystone status can rarely be demonstrated experimentally.

Every term the collection defines is gathered in the glossary.