Mimicry needs three parties: a model, a mimic that resembles it, and a receiver whose behaviour changes because of the resemblance. Without a receiver being fooled it is camouflage, convergence or coincidence — a distinction that sounds pedantic and is what stops the word covering every likeness in nature.
The definition is the useful part of this subject, because popular natural history applies "mimicry" to any resemblance and thereby drains it. A hoverfly banded yellow and black is a mimic: birds that have been stung by wasps avoid it, so the resemblance works through a third party whose behaviour it changes. A stick insect that looks like a stick is not mimicking — no animal is treating it as a stick and acting accordingly; it is simply not being seen, which is camouflage. Once the three-party requirement is in place the categories fall out naturally. Batesian mimics are harmless and copy something dangerous, which only works while they stay rare enough that the lesson holds. Müllerian co-mimics are all genuinely defended and converge on one shared warning pattern, so every predator’s education protects all of them. Aggressive mimics look harmless in order to get close to prey. And the most extraordinary cases cross kingdoms entirely: orchids that reproduce the sex pheromone of a particular wasp accurately enough that males prefer the flower to the female.
Developed record · 64% complete · reviewed 2026-08-10
What this page covers
A relationship between a model, a mimic and a receiver. Covers documented systems across insects, spiders, birds, fish and flowering plants.
Often confused with: Camouflage, where the resemblance is to the background and nothing is deceived into acting; Convergent evolution, where two lineages resemble each other for the same reason rather than through deception
Quick facts
Requires
A model, a mimic, and a receiver whose behaviour changes
Batesian
Harmless mimic, defended model — works only while the mimic is rare
Müllerian
Several defended species sharing one warning signal
A resemblance is only mimicry if something is deceived by it
Established
Specialists would state this without hedging. Multiple independent lines of evidence agree.
Mimicry describes a three-party system: a model, a mimic that resembles it, and a receiver whose behaviour is altered by the resemblance. Batesian mimics are undefended and resemble defended models; Müllerian co-mimics are all defended and converge on a shared warning signal; aggressive mimics resemble something harmless in order to approach prey or hosts. Resemblance without an affected receiver is convergence, crypsis or coincidence.
Who this applies to
documented mimicry systems across animals and plants
Studied in
Animalia, Angiospermae
You may have heard
“That insect is mimicking a leaf”
Looking like a leaf in order not to be seen is camouflage, not mimicry — nothing is being fooled into treating it as a leaf and acting accordingly. The distinction sounds pedantic and it is what stops the word covering every resemblance in nature, at which point it stops meaning anything.
Why we rate it this way, and what the caveats are
EstablishedHigh confidence
The categories are supported by predator-choice experiments, comparative morphology and, in several systems, the genetics of the resemblance itself.
How far it can be extended
The three-party structure is definitional; experimental evidence for receiver responses is concentrated in butterflies and a few other groups.
Caveats
Experimental evidence is concentrated in a small number of well-studied systems.
Receiver psychology is inferred from experiments with predators that may not be the historical selective agents.
The categories overlap; many real cases sit between them.
Still unanswered
How often is imperfect mimicry good enough because the receiver is not looking closely?
Definitions, the evidence for each category, and the boundary cases where the categories fail.
A resemblance on its own is not evidence of anything. Two species can look alike because they face the same physical problem, because one is hiding against a background, or by chance. What makes a resemblance mimicry is that some third organism responds to the mimic as though it were the model, and does something differently as a result.
That third organism is usually a predator, but it can be a prey animal, a pollinator, a host, or a rival. It also means mimicry can be tested: show the receiver both, and see whether its behaviour separates them.
What is and is not mimicry
Case
Receiver deceived?
What it is
Hoverfly banded like a wasp
Yes — birds avoid it
Batesian mimicry
Stick insect resembling a twig
No — it is simply not detected
Camouflage
Two unrelated cacti with the same shape
No receiver involved
Convergent evolution
Orchid smelling of a female wasp
Yes — males attempt to mate with it
Sexual mimicry
Anglerfish lure resembling a small animal
Yes — prey approaches
Aggressive mimicry
Words used here
Model
The organism being resembled.
Receiver
The organism whose behaviour is changed by the resemblance. Without one there is no mimicry.
Crypsis
Being hard to detect at all. Camouflage is crypsis, not mimicry.
Batesian mimicry is a harmless species resembling a defended one — the hoverfly and the wasp, the milk snake and the coral snake. It depends on predators having learned to avoid the model, which means it carries a built-in constraint: if mimics become common relative to models, predators encounter too many harmless ones and the lesson breaks down. Batesian mimicry is self-limiting in a way the other kinds are not.
Müllerian mimicry is different in a way that is easy to miss. Here every species involved is genuinely defended, and they converge on one shared warning pattern. Nobody is cheating; they are pooling the cost of educating predators, because a predator that learns the pattern from one species avoids all of them. The Heliconius butterflies of South America are the standard case, with distinct species in the same forest converging on identical wing patterns.
Aggressive mimicry inverts the purpose. The mimic resembles something harmless or attractive in order to approach prey — an anglerfish’s lure, a firefly female that copies another species’ flash pattern to attract and eat its males, a cuckoo egg matching a host’s clutch.
A harmless mimic loses its protection outside the range of the dangerous animal it copies
Established
Specialists would state this without hedging. Multiple independent lines of evidence agree.
Plasticine replicas of the harmless scarlet kingsnake in mimetic ringed colouring were attacked significantly less than plain brown replicas at sites within the range of the venomous coral snake, and lost that advantage entirely at sites beyond the model’s range. The resemblance is identical in both places; what differs is whether local predators have encountered the model.
Who this applies to
one mimic–model pair across the model’s range boundary in the south-eastern United States
Studied in
Lampropeltis elapsoides, Micrurus fulvius
You may have heard
“The kingsnake copies the coral snake so predators leave it alone”
Only where the predators have met a coral snake. The mimicry is not a property of the snake — it is a property of a relationship with a third party, and taking that party away leaves a conspicuously striped animal with nothing to hide behind.
Why we rate it this way, and what the caveats are
EstablishedHigh confidence
A field experiment with the ideal control built into the geography: the same mimic, the same predators’ ecological role, and the model present or absent.
How far it can be extended
Frequency- and experience-dependence is predicted for Batesian mimicry generally and has been shown in several other systems.
Caveats
Plasticine replicas lack movement and smell, and may attract predators a real snake would not.
Attack marks identify a predator class rather than a species.
One mimic–model pair; the strength of the effect elsewhere is not established by this alone.
Still unanswered
How quickly does mimic protection decay as a model becomes locally rare?
Mallet and Joron, 1999 · Annual Review of Ecology and Systematics
The frequency-dependent logic that predicts this result, and its contrast with Müllerian systems.
That constraint has a testable consequence, and the test is unusually clean. If Batesian mimicry depends on predators having learned the model, then the same mimic should be protected where the model lives and unprotected where it does not — with the mimic itself unchanged.
How we know
Putting fake snakes where the real venomous one does not live
A harmless kingsnake is ringed like a venomous coral snake. Does the resemblance protect it — and does it need the coral snake to be there?
Plasticine replicas were placed in the field in two colourings: the mimic’s ringed pattern and a plain brown control. They were deployed at sites inside the coral snake’s range and at sites beyond it, and predator attacks were counted from the tooth and beak marks left in the soft material. The geography supplies the control that laboratory work cannot: the mimic is identical in both places, and only the predators’ experience differs.
What happened
Where coral snakes were present, ringed replicas were attacked significantly less often than plain ones. Beyond the model’s range that advantage vanished, and ringed replicas were attacked as often or more often than plain ones.
What it shows
That Batesian mimicry is a property of a relationship rather than of an animal. Take away the predators’ experience of the model and a conspicuously striped snake is left with nothing to hide behind. It is the cleanest available demonstration that mimicry requires a receiver who has learned something.
What it does not show
Plasticine replicas do not move or smell, and may attract or repel predators for reasons unrelated to pattern. Attack marks identify a predator class at best, so which animals are doing the learning is largely inferred. One mimic–model pair in one region, and the outside-range sites differ from the inside-range ones in more ways than the snake.
The controls — what makes this evidence rather than a story
Plain brown replicas of identical size and material at every site, giving a within-site baseline attack rate.
Sites inside and outside the model’s range, which manipulates the receiver’s experience without touching the mimic.
Replicas placed in matched microhabitats and left for equal periods.
Attack marks scored blind to treatment where possible, since interpreting a dent is a judgement.
Which is what happens. Beyond the coral snake’s range, the ringed pattern stops protecting anything and becomes what it always physically was: conspicuous colouring on an edible animal. Mimicry is not a property of the mimic. It is a property of a relationship, and removing one party ends it.
Many hoverflies are, to a human eye, poor wasp mimics. One current explanation is that they do not need to be good: a bird that hesitates for a fraction of a second has already given the fly its escape.
Words used here
Batesian mimicry
A harmless species resembling a defended one, and benefiting from its reputation.
Müllerian mimicry
Several defended species converging on the same warning signal, sharing the cost of teaching predators.
Aposematism
Advertising being defended, usually with bright colour. The signal Batesian mimics copy.
Visual mimicry dominates the literature because we are visual animals, and it is not obviously the commonest kind. A receiver that navigates by scent can be deceived by scent, and one that responds to sound can be deceived by sound.
Chemical mimicry is widespread among insects that live inside other insects’ societies. Some beetles and caterpillars acquire or manufacture the cuticular hydrocarbon profile of a particular ant colony, and are carried inside and fed as though they were brood. The ants have no visual check to fail; the chemical signature is the identity.
Acoustic mimicry includes birds that reproduce the alarm calls of other species to clear a feeding area, and moths that answer a bat’s sonar with clicks resembling those of species the bat has learned are unpalatable.
Some orchids get pollinated by chemically impersonating a female insect
Established
Specialists would state this without hedging. Multiple independent lines of evidence agree.
Sexually deceptive orchids emit blends of volatile compounds closely matching the sex pheromones of specific insect species, principally solitary bees and wasps. Males are attracted and attempt copulation with the flower, transferring pollinia. Around a third of orchid species offer no reward at all, using either this or generalised food deception; deceptive species receive fewer visits but achieve higher outcrossing rates.
Who this applies to
sexually deceptive orchid genera such as Ophrys and Chiloglottis
Studied in
Orchidaceae, Hymenoptera
You may have heard
“Flowers evolved to please pollinators”
A third of the largest plant family pays nothing, and some of them actively exploit their visitors. Pollination is a transaction where it suits both parties and a swindle where the plant can get away with it.
Why we rate it this way, and what the caveats are
EstablishedHigh confidence
The pheromone match has been characterised chemically, and synthetic blends elicit the same male behaviour as the flowers.
How far it can be extended
The chemical match is specific to particular orchid–insect pairs; the strategy has arisen independently several times.
Caveats
Detailed chemical matching is demonstrated for a modest number of species.
Whether deceived males suffer measurable fitness costs is unresolved.
Reproductive success is measured over periods short relative to orchid lifespans.
Still unanswered
How do deceptive orchids track changes in their pollinator’s pheromone over evolutionary time?
Places sexual deception within the general framework of mimicry.
Sexually deceptive orchids are the strongest cross-kingdom mimicry known. The flower emits a blend of compounds matching the sex pheromone of a specific solitary bee or wasp, closely enough that males arrive and attempt to copulate with it, picking up pollinia in the process. In several species the flower is preferred to actual females.
Shape and texture contribute, but chemistry does the recruiting — synthetic blends attract males in the absence of any flower at all. Each orchid species typically deceives one insect species, which is why the pollination is so precise and why these orchids are so vulnerable to losing their pollinator.
How we know
The orchid that smells like a female bee
Why do male bees try to mate with an early spider orchid — is it what the flower looks like, or what it smells like?
The scent of the orchid and the sex pheromone of female Andrena bees were both analysed by gas chromatography and compared compound by compound. Which compounds the male bees could actually detect was established electrophysiologically, by recording from their antennae. Then the decisive step: the active compounds were synthesised, applied to dummies with no orchid present, and offered to males in the field.
What happened
The orchid produces the same hydrocarbon blend as a receptive female bee, in nearly the same proportions. Males attempted copulation with the flower, and the synthetic blend alone produced the same behaviour with no flower there at all.
What it shows
The deception is chemical. The orchid is reproducing a specific sex pheromone closely enough to fool males of one bee species, and pollen is transferred during the attempt.
What it does not show
This is one orchid species and one bee species, and generalising it to orchids is the standard error — sexual deception occurs in a minority of lineages, each targeting its own pollinator. The bioassay also measures male approach and attempted copulation, not how much pollen actually gets transferred or how many seeds result.
The controls — what makes this evidence rather than a story
Dummies carrying the synthetic blend removed the flower entirely, so any visual resemblance could not be the explanation.
Antennal recordings identified which compounds the bee can perceive, rather than assuming that chemical similarity means behavioural relevance.
The orchid blend was compared against the real female pheromone quantitatively, not merely qualitatively — the proportions matter.
It is worth sitting with what this means for the usual framing of pollination. Around a third of orchid species offer no reward whatever. Flowers are not straightforwardly cooperative partners paying for a service — they are organisms getting pollen moved by whatever means works, and sometimes the means is fraud.
Words used here
Pollinium
A packet of pollen transferred as a single mass, glued to the visitor. Orchids use them.
Why it matters: Many mimics are obviously poor copies to a human observer and still work. Whether that is because predators judge quickly, because selection is weak, or because we are the wrong receiver is unresolved.
How do sexually deceptive orchids track their pollinator’s pheromone over time?
Why it matters: The insect’s signal is itself under selection. How the flower keeps up is a question about coevolution with an unusually precise measurement attached.
How often is a resemblance mimicry rather than convergence?
Why it matters: The distinction requires evidence about a receiver, and for most claimed cases in the popular literature that evidence has never been collected.