Abstract
Animals warn off potential predators through conspicuous, distinctive colors and patterns, called visual warning signals. These visual patterns have well understood effects on predators: they deliver unlearnt wariness and enhanced memorability. But it is not clear what specific image characteristics of such patterns cause these behavioural effects. Here we tackle the question of if, and why, specific spatial frequencies are pervasive in warning signals. We hypothesized that the dominant frequencies involved in animal warning signals may specifically stimulate predator visual systems at the typical distance at which they decide whether to attack or not.
We first developed a database of hyperspectral images of Lepidoptera (butterfly and moth) wing patterns, from both species that utlise warning signal patterns and species that do not. We also built a generic mathematical model of the bird visual system, including model ‘neurons’ sensitive to different spatial orientations and spatial scales.
We found that warning signals trigger a higher model activity, and do so maximally for a subpopulation of the model units sensitive to specific spatial frequencies. We found that at 8-12cm (the typical decision distance for some birds), the maximum model sensitivity corresponds to the peak of the bird contrast sensitivity function, namely 1 cpd.
When viewing distance is taken into account, our findings show that patterns dominated by 1cpd spatial frequencies maximally stimulate a generic model of the early bird visual system. This suggests that such patterns might be specifically deterrent to birds predators Thus, some characteristics of animal warning signals may have evolved to deter predators at the distance where a decision is made.