Understanding how warning colours first evolved has long vexed scientists. Writing in Science, Loeffler-Henry et al.1 present data that indicate a hitherto unknown mechanism underlying this phenomenon.

In 1867, Charles Darwin wrote to Alfred Russel Wallace, the co-discoverer of the theory of natural selection, seeking an answer to the question, “why are caterpillars sometimes so beautifully and artistically coloured?” (see go.nature.com/4tdeepk). Conspicuous colourful larvae were a problem for Darwin because the colours of immature animals could not be explained by his theory of sexual selection (“the advantage which certain individuals have over other individuals of the same sex and species solely in respect of reproduction”)2, given that juveniles don’t reproduce.

Wallace realized that to avoid predator attacks, prey with secondary defences (those that are usually used during, or just before, contact between prey and predator) might need traits that help predators to distinguish them from edible prey. Addressing Darwin’s caterpillar question, he wrote: “Now supposing that others … are protected by a disagreeable taste or odour, it would be a positive advantage to them never to be mistaken for any of the palatable catterpillars, because a slight wound such as would be caused by a peck of a bird’s bill almost always I believe kills a growing catterpillar. Any gaudy & conspicuous colour therefore, that would plainly distinguish them from the brown & green eatable catterpillars, would enable birds to recognise them easily as a kind not fit for food, & thus they would escape seizure which is as bad as being eaten” (see go.nature.com/3tir8pd).

Darwin was pleased with Wallace’s explanation, remarking, “I never heard any thing more ingenious than your suggestion & I hope that you may be able to prove it true” (see go.nature.com/3kttvgb). Wallace incorporated warning as one of the major categories in his classification of coloration3, although it was his contemporary Edward Poulton who formalized the concept and named it aposematism, defining4 it as “an appearance which warns off enemies because it denotes something unpleasant or dangerous”.

Many species have conspicuous coloration that accurately advertises a defence or other characteristics enabling them to escape predation5 — but there is an outstanding problem in explaining how this arises. How can a conspicuous form initially evolve in a cryptic (camouflaged) population? Being conspicuous increases the probability of being discovered by a predator, and because conspicuous prey will initially be rare, predators will not learn to avoid the colour signal, and thus the risk of extinction of the conspicuous form will be high.

The geneticist Ronald Fisher recognized this paradox6 back in 1930. His solution was that aposematic prey often live ‘gregariously’ in social groups, so a predator might sample an unpalatable caterpillar and then avoid all others nearby that would probably be from the same clutch of eggs. Yet subsequent analysis (by phylogenetic reconstructions of family trees) of the ancestral states of conspicuous and social caterpillars, for example, shows this could not be true. Warning coloration always evolved in solitary species, not in gregarious ones, but gregariousness evolved in species with either aposematic or cryptic larvae7.

A second solution is that conspicuousness evolves gradually. However, experiments with artificial prey and great tits show that predators do not learn to avoid more-exaggerated conspicuous signals after being trained on a less-conspicuous prey item8. A third possibility, that aposematic prey survive attacks, has more traction. When hand-reared (naive) birds were given five types of aposematic insect prey, the overwhelming majority of prey survived, either being dropped unharmed or losing only a leg9.

A fourth possibility is that predators are simply fearful of trying prey that have an unusual appearance. There is some evidence for this, although predators vary hugely in their pickiness10. Now there is a fifth, completely new solution.

Loeffler-Henry and colleagues considered the fact that some cryptic animals, when in competition with other members of their species or when under predatory threat, display a bright colour patch, but only briefly (Fig. 1). In an imaginative leap across various subdisciplines, the authors reasoned that species with conspicuous lower (ventral) surfaces but cryptic upper (dorsal) surfaces might provide a pathway through which full-blown conspicuousness could evolve.

Focusing on frogs, toads and salamanders, some of which are toxic, such as yellow-bellied toads (Bombina variegata), California newts (Taricha torosa) and poison dart frogs (Dendrobatidae family), the authors mapped a phylogenetic tree of species that had a range of anti-predator characteristics: fully cryptic species; fully conspicuous species; species with conspicuous coloration present as small patches on their ventral surfaces (PV); species with a fully conspicuous ventral surface (FV); and species that had both cryptic and conspicuous forms (polymorphic species). Using two data sets, a large one that lacked information on chemical defences, and a smaller one that included them, the authors uncovered a plethora of evolutionary relationships. What seems to be key to the origin of aposematism is that amphibians with hidden colour signals, specifically the chemically defended FV state, are probably the most important evolutionary precursor of the aposematic conspicuous defended state.

In turn, the FV defended state arises from the PV defended state or the undefended FV state, and these FV species themselves evolve from the undefended PV state. The authors report that around 90% of the conspicuous, FV conspicuous or polymorphic species that they analysed are chemically defended, and this is also true for a good proportion of PV conspicuous and cryptic species. This indicates that amphibians are honest signallers rather than species that ‘cheat’ by mimicking warning colours in the absence of defences. The authors also found evidence that aposematic species can evolve back to cryptic or polymorphic species, mirroring the surprising evolutionary flexibility seen in transitions between mimicry and crypticity in coral snakes11, an observation that questions the idea that stable evolutionary end points exist.

Remarkably, scientists already knew about mix-and-match cryptic–conspicuous forms of protective coloration in three other contexts. One of these is deimatism, in which cryptic prey briefly flash a hidden conspicuous patch to cause the predator to hesitate — such prey might or might not have chemical defences12. Another is flash coloration, whereby prey expose conspicuous patches while fleeing but hide them as soon as they come to rest, causing the predator to search for an inappropriate object13. The third is distance-dependent camouflage, in which defended prey are cryptic far off but conspicuous up close14.

However, none of these three examples was used to solve the aposematism paradox until now. With a new solution at hand, namely that aposematism can evolve without loss of crypsis, it is essential to examine how widespread the phenomenon is by investigating other groups of species with ‘dangerous’ reputations, such as sea slugs and snakes. Once again, Wallace has led the way, and we mortals simply follow on behind.