This is the question addressed in Mathieu Chouteau’s paper, just published in Animal Behaviour (vol 153:49-59).
How do bird predators consider the natural range of variation in toxin load among butterfly species in a community? Mathieu asked chicks what they thought of the chemically-defended butterflies composing the mimicry communities of Eastern Peru. He measured how fast they would learn to avoid butterfly-flavoured food crumbs, as opposed to tasty crumbs. Surprisingly, Mathieu found that only a few butterfly species with low loxin content can improve their aversiveness to chicks. Most well-defended prey elicited similar learning rates, irrespective of the wide variation in their concentration in cyanogenic or other distasteful compounds.
Then, why should butterflies invest in high toxicity ? Some can be really nasty ! If natural predators such as jacamars or flycatchers learn like chicks, the benefits to butterflies in loading themselves with toxins may lie elsewhere than in the education of predators. Perhaps in accessing competition-free hostpants, or in making better pheromones, or in elicting immediate taste-rejection ?
Another important consequence is that differences in toxicty among mimetic butterflies may not represent a major determinant of their roles as mimics vs. models. On the contrary, if all toxic species are considered equally aversive by predators, their roles in predator education should be determined primarily by their relative abundances. Fritz Muller was right again!