Skip to content
NatureHQ

Ecologyphenomenon

Convergent evolution

Vertebrate and octopus eyes are the classic case of convergence — and what makes them convincing is that they are built differently.

Unrelated lineages arriving at the same solution because there are not many solutions. Eyes, flight, echolocation and venom have each been invented several times over — and the useful part is usually where the resemblance breaks down.

The temptation with convergence is to treat it as a collection of striking coincidences, which is exactly what it is not. When unrelated lineages keep landing on the same design, the repetition is information: it says the space of workable answers is narrow, and it lets you ask what is doing the narrowing. Sometimes the constraint is physics: image-forming eyes have arisen many times, and the optical solutions — pinhole, lens, mirror, compound — are few because forming an image permits few. Sometimes it is economics: an animal living on carrion alone must find carcasses scattered unpredictably across a landscape, and searching cheaply enough means soaring flight, which is why the obligate scavengers on every continent are large soaring birds and there is no walking equivalent anywhere. Sometimes it is chemistry: a plant toxin binds at one site on one protein, and the insects that beat it — from several unrelated orders — have hit the same handful of amino acids, because there is barely another way to stop the binding without breaking the pump. The most instructive cases are those where the match is imperfect. Vertebrate and octopus eyes are the standard example, and they convince because they are built differently. The vertebrate retina faces backwards: light passes through nerve fibres to reach the photoreceptors, and where those fibres exit there is a blind spot. The octopus retina is the other way round and has none. Two lineages solved the same optical problem by different developmental routes — which is what independent invention should look like. A perfect match would be more suspicious, not less. And molecular convergence carries a warning that generalises. Any two lineages share some substitutions by chance, so a count of matches means nothing until set against how many are expected anyway — the ground on which a famous genome-wide result for echolocating bats and dolphins was challenged. The disagreement is recorded here rather than settled.

Early coverage · 44% complete · reviewed 2026-09-03

What this page covers

Convergence occurs everywhere life does. The examples here are drawn from animals, plants and fungi, and several of the strongest cases cross kingdoms.

Often confused with: Two animals looking alike, which may be shared ancestry rather than convergence; Evolution having a goal, when the repetition reflects a narrow space of workable answers; Molecular convergence being obvious from a count of matching sites, without asking how many are expected by chance

Quick facts

Repetition is information
Eyes, flight, echolocation and venom, each invented many times
The evidence is the difference
The vertebrate retina faces backwards; the octopus retina does not
Down to the amino acid
Unrelated insects hit the same sites in the same pump
Count against chance
Genome-wide convergence needs a null model — and this one is disputed

Why the repetition tells you something

A narrow space of answers, and three different things narrowing it.

Image-forming eyes, powered flight, echolocation and venom have each arisen many separate times. Repetition on that scale is evidence that the space of workable solutions is narrow — not that nature is short of ideas.

Established

Specialists would state this without hedging. Multiple independent lines of evidence agree.

Complex functional traits including image-forming eyes, powered flight, echolocation and venom systems have evolved independently in multiple unrelated lineages. Recurrence of a small number of designs is consistent with strong physical and functional constraint on the space of viable solutions.

Who this applies to
Repeated independent origins of functionally similar traits across life.
Studied in
Animalia, Plantae, Fungi
Why we rate it this way, and what the caveats are
EstablishedHigh confidence

Independent origins are established by phylogeny and, for many traits, by structural differences showing the solutions were built separately.

How far it can be extended

Documented for many independent trait origins across widely separated lineages.

Caveats

  • Similar appearance is not by itself convergence; the independence has to be established phylogenetically.
  • Constraint explains why solutions recur, not why a particular lineage took a particular route.

Still unanswered

  • How much of the constraint is physics, how much is development, and how those two contributions could be separated.

Last reviewed 2026-09-03

The evidence (2 studies)

  • Supports · primary

    Animal Eyes

    Land and Nilsson, 2012 · Oxford University Press

    Organises eye designs by optical principle and shows how few solutions physics permits.

  • Supports · primary

    The genetic causes of convergent evolution

    Stern, 2013 · Nature Reviews Genetics

    The general review of repeated evolution and how often it recurs.

Cases from this corpus, and what is doing the constraining
Solution reached repeatedlyConstraintWhere it appears here
Camera-type eyeOpticsVertebrates and cephalopods
Obligate scavengingSearch economicsTwo unrelated vulture lineages
EcholocationPhysics of ranging by soundBats and toothed whales
Catching animals for nitrogenNutrient shortageCarnivorous plants and predatory fungi
CounterilluminationThe geometry of being seen from belowFish and squid
Beating a plant toxinOne binding site on one proteinSeveral insect orders

The rows are not doing the same work. Optics is a hard physical limit on what can form an image. Search economics is an accounting constraint about whether a diet pays. A binding site is a chemical constraint on which changes can help. Grouping them as "convergence" is useful only if the differences stay visible — otherwise the word becomes a label for anything that happened twice.

The instances, on their own pages

The eye that was built twice

And the difference is the evidence.

Vertebrate and octopus eyes match optically and differ in construction: the vertebrate retina faces backwards, with the nerves in front of the photoreceptors and a blind spot where they exit. The octopus retina does not.

Established

Specialists would state this without hedging. Multiple independent lines of evidence agree.

Camera-type eyes in vertebrates and coleoid cephalopods are functionally convergent, sharing lens optics and a retinal sensor layer, while differing in developmental origin and retinal orientation. The vertebrate retina is inverted, with axons crossing anterior to the photoreceptors and exiting through the retina to form an optic disc.

Who this applies to
Camera-type eyes in vertebrates and in coleoid cephalopods.
Studied in
Vertebrata, Cephalopoda
Why we rate it this way, and what the caveats are
EstablishedHigh confidence

Basic comparative anatomy, established for well over a century.

How far it can be extended

The anatomical contrast holds across both groups.

Caveats

  • The inverted vertebrate retina is not simply a flaw; supporting cells appear to channel light through the overlying layers, and the arrangement has functional advantages of its own.
  • The two eyes also share more than optics, including some developmental genes, which complicates any account of complete independence.

Still unanswered

  • How much of the shared genetic toolkit for eye development was present in the common ancestor, and what that means for calling the eyes independent.

Last reviewed 2026-09-03

The evidence (1 study)

  • Supports · primary

    Animal Eyes

    Land and Nilsson, 2012 · Oxford University Press

    The comparative treatment of both eye types and their optics.

Diagram

Same optics, opposite construction

The order of the layers, which is where the two eyes part company.

Same optics, opposite constructionVertebrateLight arrivesNerve fibresPhotoreceptorsSupport layerFibres exit through the retina — a blind spotOctopusLight arrivesPhotoreceptorsNerve fibresSupport layerFibres leave behind the retina — no blind spotThe difference is the evidence: a perfect match would be more suspicious.
The same explanation in words

Two columns compare the layer order of a vertebrate and an octopus retina. In the vertebrate column, light arrives first at the nerve fibres, then reaches the photoreceptors, with the support layer behind — and a note records that the fibres exit through the retina, producing a blind spot. In the octopus column, light arrives directly at the photoreceptors, with the nerve fibres behind them and the support layer behind those — and a note records that the fibres leave behind the retina, so there is no blind spot. A closing line observes that the difference is the evidence of independent origin: a perfect match would be more suspicious.

It is worth being careful about what the inverted retina is and is not. It is often presented as a design flaw — light passing through nerve fibres and blood vessels before reaching the photoreceptors, with a blind spot where the fibres exit. The blind spot is real. But supporting cells appear to channel light through the overlying layers, and the arrangement puts the photoreceptors against a layer that services them heavily. The honest statement is that the two retinas are built differently, both work, and the difference is what shows they were arrived at separately.

There is a complication worth stating rather than hiding. Both eyes use some of the same developmental genes, which were present in a common ancestor that had no camera eye. So "independent" here means the organ was assembled separately from a shared toolkit — not that nothing was shared. Convergence is rarely as clean as the textbook version, and saying so costs nothing.

The same amino acids, in unrelated animals

When the target is narrow enough, so is the answer.

Unrelated insects feeding on milkweeds repeatedly acquired the same substitutions in the same protein — the pump the plant’s toxin attacks. They did not inherit the solution; there is barely another one.

Established

Specialists would state this without hedging. Multiple independent lines of evidence agree.

Independent insect lineages feeding on cardiac-glycoside-producing plants show repeated substitutions at a small set of sites in the sodium–potassium ATPase, the molecular target of those compounds, with several identical changes arising independently.

Who this applies to
Insects feeding on plants producing cardiac glycosides; the tightness of the constraint reflects one specific molecular target.
Studied in
Insecta
Why we rate it this way, and what the caveats are
EstablishedHigh confidence

Sequence evidence across many lineages, with the mechanism of the toxin and its binding site independently known.

How far it can be extended

Documented across several unrelated insect orders on the same plants.

Caveats

  • The functional effect is directly established for some substitutions and inferred for others.
  • Resistance is one part of feeding on a toxic plant; sequestering the toxin is a separate problem.

Still unanswered

  • What the substitutions cost the pump, and how insects offset that cost.

Last reviewed 2026-09-03

The evidence (2 studies)

When a trait evolves repeatedly, the same genes turn up far more often than chance would allow. The target is narrow: few genes can be altered to produce the trait without breaking something else.

Well supported

Good evidence backs this, though some details remain open.

Independent origins of similar traits implicate the same genes at frequencies exceeding chance expectation. Contributing factors include the limited number of loci whose modification produces the phenotype without unacceptable pleiotropic cost, and the availability of standing variation.

Who this applies to
Documented across many trait origins, with coverage weighted towards well-studied organisms.
Studied in
Animalia, Plantae
Why we rate it this way, and what the caveats are
Well supportedModerate confidence

A robust pattern across documented cases, though the documented cases are biased towards traits and organisms where genetic work is easiest.

How far it can be extended

The pattern is drawn from many independent cases across taxa.

Caveats

  • Publication and study effort concentrate on traits that are easy to map, which inflates apparent repeatability.
  • How far this extends to traits built from many genes of small effect is less clear.

Still unanswered

  • Whether the restricted routes reflect the genetics or the limited set of traits anybody has mapped.

Last reviewed 2026-09-03

The evidence (2 studies)

The milkweed case is the tightest constraint in the set. The plant makes compounds that jam the sodium–potassium pump, a protein every animal cell needs. An insect that wants to eat the plant must stop the toxin binding without breaking the pump, and there are very few changes that do both. Unrelated insects, from different orders, have made the same ones.

A count is not evidence until you know what chance gives you

A famous genomic result, and the objection it drew.

A widely reported result found genome-wide convergence between echolocating bats and dolphins. A re-analysis argued the count was never compared against how much convergence any two lineages show by chance.

Contested

Researchers actively disagree, and the disagreement is substantive.

Reported genome-wide convergent substitution between echolocating bats and cetaceans was subsequently re-examined against explicit null expectations, which indicated the observed count was not clearly in excess of that between comparable non-echolocating lineage pairs. The interpretation remains disputed.

Who this applies to
A specific genomic comparison and its re-analysis; not a statement about whether echolocation involves any convergent genetics.Do not extend this beyond the taxa listed — the popular version over-reaches.
Studied in
Chiroptera, Cetacea
Why we rate it this way, and what the caveats are
ContestedLow confidence

Two competent analyses of the same data reach different conclusions about whether the signal exceeds chance. NatureHQ does not adjudicate, and records the disagreement.

How far it can be extended

The dispute concerns the analysis of one dataset, and the methodological lesson generalises further than the result does.

Caveats

  • This is a dispute about analysis, not about whether bats and dolphins both echolocate — which is not in question.
  • Both analyses depend on the same alignments, so alignment quality affects each equally.

Where researchers disagree

  • The original analysis reported convergence across many loci; the re-analysis found no clear excess over comparable non-echolocating lineage pairs once chance convergence was accounted for.
  • Showing that a genome-wide excess is not established does not rule out convergence at particular hearing genes, where independent evidence exists.

Still unanswered

  • What the right null expectation is for convergence tests generally, which is a live methodological question beyond this case.

Last reviewed 2026-09-03

The evidence (2 studies)

How we know

Counting matches against chance

Echolocating bats and dolphins were reported to share convergent substitutions across the genome. Is that more convergence than any two lineages would show anyway?

The same genomic comparison was re-analysed after establishing an explicit expectation: given branch lengths and the substitution process, how many convergent amino-acid changes should occur between two lineages that share no trait of interest? Observed counts for the echolocating pair were then compared against that expectation and against control pairs of non-echolocating lineages.

What happened

Convergent substitutions between the echolocating lineages were not clearly in excess of those between comparable non-echolocating pairs once the null expectation was applied.

What it shows

That a count is not evidence until you know what chance gives you. There are twenty amino acids and a great many sites, so any two lineages will independently land on the same residue somewhere; the question is always whether there is more of it than that. The lesson is general and applies to any search for coincidence in a large dataset.

What it does not show

It does not show that echolocation involves no convergent genetics — several hearing genes have independent evidence — and it does not touch the fact that both groups echolocate. It is a dispute about one analysis, and it depends on the same alignments, so alignment quality affects both sides equally.

The controls — what makes this evidence rather than a story
  • Non-echolocating lineage pairs, which is the comparison the original analysis lacked.
  • An expectation derived from branch lengths and the substitution model, rather than an assumption that chance convergence is negligible.
  • The same alignments as the original, so the difference is the analysis and not the data.

From Determining the null model for detecting adaptive convergence from genomic data: a case study using echolocating mammals

The reasoning behind the objection is general and worth carrying away. Any two lineages, echolocating or not, will independently arrive at the same amino acid at some sites simply because there are twenty of them and a great many sites. So the number of convergent sites means nothing on its own; what matters is whether there are more than would occur between two comparable lineages that share no such trait. Supply that comparison and the reported excess was no longer clear.

None of this touches whether bats and dolphins echolocate, which is not in dispute, or whether particular hearing genes show convergence, for which there is separate evidence. It concerns one analysis of one dataset. NatureHQ carries both papers and does not adjudicate between them — and the underlying lesson is the same one applied to behavioural claims everywhere else on the site: compare the observation against what would have happened anyway.

The research behind this page

5 studies, newest first. Each one has a page explaining what it found and what it could not show.

2015Molecular Biology and Evolution

Determining the null model for detecting adaptive convergence from genomic data: a case study using echolocating mammals

Convergent substitutions between echolocating lineages were not clearly in excess of the number expected between comparable non-echolocating lineage pairs, once an appropriate null expectation was applied.

2013Nature Reviews Genetics

The genetic causes of convergent evolution

The same genes are implicated in independent origins of similar traits far more often than chance would suggest.

2013Nature

Genome-wide signatures of convergent evolution in echolocating mammals

Convergent amino-acid substitutions between echolocating bats and cetaceans were reported at many loci across the genome, including but not limited to genes with known hearing functions.

2012Oxford University Press

Animal Eyes

Image-forming eyes have arisen independently many times, and a small number of optical solutions recur because physics permits few.

2012Science

Parallel molecular evolution in an herbivore community

Insects from unrelated orders feeding on cardiac-glycoside plants repeatedly acquired substitutions at the same small set of sites in the same gene, with several identical changes arising independently.

This page is a stop on a longer route

A guided journey reads several subjects in a deliberate order, with an argument for why one follows another. You can join in the middle.

Where to go from here

Each of these follows from something on this page — a relationship in the evidence, a claim people ask about, or the next mechanism along.

How complete this page is, and what it is still missing

NatureHQ publishes its own gaps. This page is at 44% completeness against what we would call a finished subject, and was last reviewed on 2026-09-03. It carries 5 claims and answers 0 mapped search questions.

  • no research from the last few years is attached — check for newer work
  • more experiments could be explained in plain English
  • no popular claim about this subject has been checked yet
  • Plant convergence — succulence, C4 photosynthesis, carnivory — is mentioned only in passing.
  • The distinction between convergence and parallelism is treated loosely, as it often is in the literature.
  • Convergence in behaviour rather than structure is not covered here.