Batesian Mimicry: Why Imperfect Copies Survive

A hoverfly hovering over a flower has yellow and black stripes, a narrow waist, and a wasp-like shape. It has no sting. A bird that has been stung by a wasp once will often leave it alone anyway.

That is Batesian mimicry. A palatable species evolves to resemble a defended or noxious species, and predators that have learned to avoid the model generalize that avoidance to the mimic. Henry Walter Bates described the phenomenon in 1862 after years of collecting in the Amazon, and it has been a textbook example of natural selection ever since.

The textbook rule is conditional

The standard story says selection should push mimics to resemble their models as closely as possible. But many Batesian mimics are not close copies at all. A 2013 review by David Kikuchi and David Pfennig in The Quarterly Review of Biology listed 11 competing hypotheses for why imperfect mimicry persists, and found that the one with the most empirical support is not about constraints or trade-offs. It is about relaxed selection. When the cost of attacking a model is high relative to the benefit of eating a mimic, selection on the mimic to be perfect is weaker. Imperfect mimicry can be just as fit as perfect mimicry.

Why does a harmless hoverfly look like a wasp?

Because predators learn. A bird that eats a wasp gets stung. The experience is unpleasant enough that the bird associates the wasp’s yellow-and-black pattern with pain. The next time it sees that pattern, it hesitates. A 2012 review in Science explained the mechanism plainly: the key selective mechanism in Batesian mimicry is the predator’s ability to learn, generalize, and remember. The predator learns by trial and error to avoid the harmful species and to generalize the bad experience to the mimic.

The mimic doesn’t need to be perfect for that generalization to work. It needs to fall within the predator’s perceptual range for the model.

Okay, but why doesn’t it look exactly like a wasp?

Because selection on the mimic is relaxed when the model is dangerous. Thomas Pike and Oliver Burman at the University of Lincoln tested this experimentally in a 2023 paper in Behavioral Ecology. They asked human participants to discriminate between artificial models and mimics. When participants misclassified a model as palatable, they received either a mild electric shock (high cost) or a vibratory buzz (low cost). When the cost of attacking a model was high, mimetic phenotypes could deviate more from the model while still maintaining the same level of fitness. When the cost was low, the mimics had to be more accurate.

The relaxed selection hypothesis predicts exactly that. As the cost of attacking the model goes up, the benefit of being more accurate goes down.

So why would a predator avoid something that doesn’t look quite right?

Because predators are not perfect discriminators. A 2025 paper in Nature used 3D-printed insects to map how birds and invertebrates respond to different degrees of mimicry. The researchers found that birds can learn to discriminate on the basis of subtle differences, but their ability is weaker for pattern and shape than for colour and size. Some invertebrate predators, especially crab spiders and mantises, were even less discriminating.

That matters because the predator community is not a single bird. It is a mix of species with different perceptual abilities. A mimic that looks slightly wrong to a bird might look close enough to a mantis. The apparent imperfection is partly in the eye of the beholder.

Why doesn’t selection just push the mimic closer and closer?

Because there is a cost to being more accurate, and in some conditions the cost is not worth paying. Kikuchi and Pfennig’s 2013 review grouped the competing explanations into four categories. The one with the most support was relaxed selection: imperfect mimicry is as adaptive as perfect mimicry. But there are also trade-offs. A 2010 paper in The American Naturalist argued that modifying a prey phenotype to mimic a model imposes costs, such as deforming limbs or compromising thermoregulatory or locomotor capacities. If the bird already avoids you, paying those costs buys you nothing.

A 2026 study on Adelpha butterflies in the Neotropics added another layer. The researchers found that increasing the absolute density of the prey community, not just the ratio of models to mimics, reduced predation on all species, including imperfect mimics. More prey in the community means more opportunities for predators to learn, and more learned avoidance to go around.

So what actually decides how good a mimic needs to be?

The cost of attacking the model. The perceptual limits of the predator. The density of the community. The trade-offs the mimic pays for accuracy. And the frequency of the mimic relative to the model. A 2026 paper by Yi Sun and Po-Ju Ke in the Journal of Animal Ecology modelled these dynamics and found that intermediate similarity promotes stable community dynamics and increases mimic abundance in single Batesian mimicry systems. The system doesn’t settle on perfect mimicry. It settles on the level of similarity that keeps the predator guessing without paying too much to get there.

The hoverfly with imperfect stripes is not a failed wasp. It is a successful hoverfly. Selection got it close enough to benefit from the predator’s learned avoidance, and the cost of getting closer was not worth the marginal gain. The next time you see one hovering over a flower, look at the stripes. They are not exact. They do not need to be.