Why the same solutions keep appearing

Eyes have evolved independently dozens of times. So has flight, at least four. So has the sleek torpedo shape of anything that chases prey through water, in fish, in reptiles, and in mammals that had already left the sea and gone back. When unrelated lineages keep arriving at the same answer, something is constraining the space of possible answers.

Convergence, and how you tell

Convergent evolution is when similar traits appear in lineages that did not inherit them from a shared ancestor. The similarity is produced by similar problems, not by common descent.

Distinguishing the two is the whole craft. If two species share a trait, either their common ancestor had it, or they arrived at it separately. The way you tell is by looking at everything in between: if the branches connecting them lack the trait, it appeared more than once. This is why building family trees from genetics rather than from appearance mattered so much. Appearance is exactly the thing convergence makes unreliable.

Why it happens so often

Three constraints narrow the space of workable answers, and between them they explain most cases.

Physics does not negotiate. Anything moving fast through water faces the same drag, and drag has a shape that minimises it. Anything using light to form an image needs an aperture, a way to focus, and a light-sensitive surface at the right distance. These are not biological facts and cannot be evolved around, so every lineage that takes on the problem is handed the same short list.

The starting materials are shared. Nearly all life uses the same handful of building blocks and the same broad toolkit of developmental machinery. Two lineages solving a problem are rummaging in similar boxes, so they reach for similar things. Light-sensitive proteins, for example, were available long before any eye existed, which is part of why eyes were so repeatable.

Most paths are not viable. Evolution cannot redesign. Every intermediate step has to be an organism that works, which rules out the majority of routes to any complex trait and leaves a small number that are reachable one viable step at a time. Different lineages searching the same narrow set of reachable paths will often find the same one.

What convergence does not mean

It is tempting to read repeated solutions as evidence that evolution is aiming at something, and it is not. Nothing is being approached. The same constraints acting on independent populations produce the same outcomes for the same reason two engineers given identical materials and an identical problem tend to converge, without either of them intending to agree.

It also does not mean the solutions are identical underneath. Vertebrate and octopus eyes look alike and are wired differently, with the octopus arrangement putting the nerves behind the light-sensitive cells rather than in front, and no blind spot as a result. Convergence produces similar function from different machinery, and the differences in machinery are usually the giveaway.

Repeatability as an experiment

This is where a simulated world earns its keep, because the obvious question about convergence cannot be asked of Earth. Rewind and run it again, and do you get the same answers?

Stephen Jay Gould argued that replaying the tape of life would produce something unrecognisable, and Simon Conway Morris argued the opposite, that convergence is so pervasive the outcome would be broadly familiar. The disagreement is unresolved because the experiment is unavailable.

It is available here in a narrow form. Every run is exactly reproducible from its seed, so the same world can be re-run identically, and different seeds produce genuinely different histories under identical rules. Comparing many seeds under one ruleset is a direct question about how much of an outcome was the rules and how much was the run, and that is precisely the comparison the campaigns on research notes are built around.