Short answer
Field experiments say yes, within limits. Clay models coloured like a red poison frog are attacked by birds less often than brown ones; butterflies that match the local distasteful species survive better than those that do not; and replicas of harmless snakes that resemble coral snakes are attacked far more often where real coral snakes are absent. Each experiment tests one system in one place, and how warning colours first evolve remains a harder question than whether they work once common.
A warning colour, or aposematic signal, is a conspicuous appearance carried by an animal that is costly to attack — toxic, distasteful, or armed. The idea is that predators learn, or are born knowing, to avoid the pattern. It makes a clear prediction: a conspicuous defended form should be attacked less than an inconspicuous one in the same place. On its own, conspicuousness is not evidence. Bright colour can equally be a mating signal, a by-product, or, as the jewel-beetle experiments showed for iridescence, a form of concealment. Establishing aposematism needs both a demonstrated defence and a demonstrated reduction in attacks associated with the signal.
The strawberry poison frog, Oophaga pumilio, is bright red at the Costa Rican site studied and carries skin alkaloids. To test whether the colour itself deters predators, researchers set out 800 frog models made of soft modelling clay along transects in the forest, some red and some brown, and after 48 hours scored the beak and tooth marks left in them. Birds made most of the attacks, and attacked brown models at almost twice the rate of red ones, a statistically significant difference. Because the models differed in colour and carried no toxins, the difference reflects the predators' response to appearance. The models did not move.
In 1879 Fritz Müller proposed that two defended species gain by looking alike, because the predators' learning is paid for by both. The idea was supported for more than a century by comparison, theory and laboratory work before anyone tested it in the field. The test used Heliconius cydno, a distasteful butterfly that occurs in more than one colour form. The study compared the survival of its forms where each did or did not match the locally abundant distasteful species. Forms that matched the local co-model survived better than those that did not. Since different co-models dominate in different places, the same result explains why such a species stays polymorphic: selection favours a different pattern in each locality.
A Batesian mimic is harmless and resembles a defended species. Its protection should therefore depend on predators meeting the real thing. This was tested with plasticine replicas of scarlet kingsnakes, harmless snakes that carry the red, yellow and black rings of venomous coral snakes. Replicas were set out inside the coral snakes' range in the south-eastern United States and north of it in central North Carolina, with striped and plain brown replicas as controls, and comparable experiments were run in Arizona. Ringed replicas were attacked far more often where coral snakes are absent. The result confirms the central prediction of Batesian mimicry — that protection depends on the defended model being present — more directly than any amount of resemblance could.
Resemblance alone cannot distinguish Batesian from Müllerian mimicry. That depends on whether the mimic is itself defended, which has to be measured.
All three results concern an established signal in a community of experienced predators. They do not explain how a conspicuous pattern spreads when it is rare, since the first conspicuous individuals are both easy to find and unfamiliar; that problem is the subject of continuing theoretical and experimental work reviewed elsewhere. They use models or releases at a handful of sites, and predators differ between regions. And they treat the signal as humans see it. Birds, the predators in each case, have ultraviolet-sensitive vision, so a model coloured to match a frog or snake to a human eye may not match it to a bird's; the frog study chose its colours by eye and checked only that its modelling clay reflected no ultraviolet.
| Hypothesis | Study | System | Method | Result | Limit |
|---|---|---|---|---|---|
| Aposematism | Saporito and others, 2007 | Strawberry poison frog, Costa Rica | Red and brown clay models set out in the forest; attack marks counted | Brown models attacked by birds more often than red | Motionless models; one site |
| Müllerian mimicry | Kapan, 2001 | Heliconius cydno and its co-model species | Survival of matching and non-matching morphs compared | Morphs matching the local co-model survived better | One system |
| Batesian mimicry | Pfennig and others, 2001 | Kingsnakes resembling coral snakes, United States | Ringed, striped and plain plasticine replicas inside and outside the model's range | Ringed replicas attacked far more often where coral snakes are absent | Replicas, not live snakes |
| Not warning colour after all | Kjernsmo and others, 2020 | Jewel-beetle wing cases | Iridescent against plain-coloured baited models | Iridescent models detected less: consistent with camouflage | One species |
Why: Conspicuousness is read as a warning without evidence of a defence.
Fix: Look for a measured toxin, sting or taste-rejection trial. Many bright animals are harmless, and some are bright for mate choice or are mimics.
Why: Batesian protection is treated as a property of the pattern, not of the local predator community.
Fix: Check whether the defended model lives in the same area; the snake-replica experiment found ringed replicas attacked far more often outside the model's range.
Why: The Batesian/Müllerian distinction is being made from appearance.
Fix: Find out whether each is defended. Two defended lookalikes are Müllerian co-mimics, and the dynamics differ.
Each statement is labelled by kind — established fact, a standard’s requirement, observed market data, a convention, or Colourwise’s own interpretation or analysis — with the strength of the evidence behind it.
FactModerate evidence· Costa Rica
In a field experiment with clay models of the poison frog Oophaga pumilio in Costa Rica, brown models were attacked by birds more often than red models, providing experimental evidence that the frog's coloration deters predators.
Caveat: Motionless clay models at one site; attack rates on models are a proxy for predation on frogs.
Source: Experimental evidence for aposematism in the dendrobatid poison frog Oophaga pumilio
FactStrong evidence
Morphs of the polymorphic butterfly Heliconius cydno survived better where they matched the locally abundant distasteful co-model species, demonstrating Müllerian mimicry in the field.
Caveat: One three-species system, known to Colourwise from the paper's abstract.
Source: Three-butterfly system provides a field test of Müllerian mimicry
FactModerate evidence· South-eastern and south-western United States
Plasticine replicas carrying the ringed pattern of harmless kingsnakes that resemble venomous coral snakes were attacked far more often where coral snakes are absent than where they occur.
Caveat: Described from the university's news release about the paper; the paper itself could not be read.
Source: Frequency-dependent Batesian mimicry; Study Of Poisonous Snakes Boosts Old Batesian Principle Of Mimicry
FactStrong evidence
Müllerian mimicry had been supported by comparative evidence, theory and laboratory simulation for more than a century before its first experimental field test was published in 2001.
Source: Three-butterfly system provides a field test of Müllerian mimicry
Colourwise interpretationModerate evidence
A bright colour should not be labelled a warning colour on appearance alone: the label requires evidence both that the animal is defended and that the colour is associated with fewer attacks.
Based on: Colourwise's reading of experiments in which conspicuous-looking colour turned out to reduce detection (jewel beetles) or to deter attack only where the defended model lives (snake replicas).
Source: Iridescence as camouflage; Experimental evidence for aposematism in the dendrobatid poison frog Oophaga pumilio
Reviewed 6 October 2026. Colourwise summarises its sources in its own words and does not reproduce standards text or proprietary colour data. Spotted an error? Tell us.