Detectability of a poison frog and its Batesian mimic depends on body posture and viewing angle.

Aposematic signals warn predators that prey should be avoided due to dangerous secondary defences. However, as warning signals do not always produce avoidance, warning colors may evolve as a trade-off balancing detectability against signal saliency. For Batesian mimics, which display salient signals but lack secondary defenses, the costs of predator encounters are greater, potentially increasing the benefit of crypsis. This raises the question of whether imperfect mimicry may reduce detectability while retaining mimetic efficacy. We tested this hypothesis with the poisonous frog Ameerega bilinguis and undefended Batesian mimic Allobates zaparo, using computational visual modeling and screen-based detection trials with human participants. We found that both species incorporate camouflage into their warning colors, but to different degrees depending on viewing angle and behavior. Contrary to expectation, we found differences in detectability between model and mimic that do not adhere to the hypothesized cryptic mimetic phenotype. To aerial observers, we found the mimic to be more detectable than the model. To terrestrial observers, likely owing to the model's bright ventral color, we found the model more detectable in viewing angles that highlight the ventral coloration, whereas the mimic was more detectable in viewing angles that highlight the dorsal coloration. Consequently, we suggest that in addition to being the result of perceptual or developmental constraints, imperfect mimicry may also evolve as an adaptive strategy which balances camouflage with different signaling functions. Our findings complement the emerging view that aposematic signals may evolve in response to a multitude of selection pressures beyond aversion alone.