Generalisation behaviour of predators toward warning signals displayed by harmful prey: answers from a video-game played by humans.
The persistence of several warning signals in sympatry is a puzzling evolutionary question because selection favours convergence of colour patterns among toxic species. Such convergence is shaped by predators reaction to similar but not identical stimulus, i.e. generalisation behaviour. However, studying generalisation behaviour in complex natural communities of predators is challenging, and is thus generally limited to simple variations of prey colour patterns. Here, we used humans as surrogate predators to investigate generalisation behaviours on two prey communities with different level of warning signals complexity. Humans generalisation capacities were estimated using a computer game simulating a simple (4 morphs) and a complex (10 morphs) community of defended (associated with a penalty) and palatable butterflies. Colour patterns used in the game are actually observed in natural populations of the defended butterflies H. numata, and generalisation behaviour of natural predators communities on these colour patterns have previously been investigated in the wild, allowing direct comparison with human behaviour. We investigated human predation behaviour by recording attack rates on the different defended and palatable colour patterns, as well as player survival time (i.e. score). Phenotypic similarity among the different colour patterns was precisely quantified using a custom algorithm accounting for both colour and pattern variations (CPM method). By analysing attack behaviours of 491 game players, we found that learning was more efficient in the simple prey community. Additionally, profitable prey gained protection from sharing key visual features with unprofitable prey in both communities while learning, in accordance with natural predator behaviours. Moreover, other behaviours observed in natural predators, such as colour neophobia, were detected in humans and shaped morph vulnerability during the game. Similarities between our results in humans and the reaction of natural predator communities to the same colour patterns validate our video-game as a useful proxy to study predator behaviour. This experimental set-up can thus be compared to natural systems, enabling further investigations of generalisation on mimicry evolution.