Field but not lab paradigms support generalisation by predators of aposematic polymorphic prey: the Oophaga histrionica complex

The persistence of novel aposematic forms, and thereby the evolution of aposematic polymorphism, remain intriguing. Novel and rare forms could be disproportionally attacked by predators that already learned to avoid a pre-existing and more common aposematic form. Alternatively, novel forms could be less frequently attacked if predators are reluctant to attack unknown potential prey (neophobia) or if previous learning allows them to generalise and recognise the novel form as toxic. We used colour variation in polymorphic poison frogs ( Oophaga histrionica complex) to test whether predators familiar with one aposematic form do generalise their avoidance behaviour to other aposematic forms. To strengthen our inference, we combined a field test of attack rates to local and non-local models with a lab experiment of generalisation capabilities by newly born chicks. Field predators attacked a significantly lower proportion of 529 aposematic compared to 150 cryptic models. Predators co-occurring with the local aposematic form of O. histrionica equally avoided non-local forms, especially in areas where the species was abundant. Forty-two lab chicks learned to discriminate between an aposematic and a cryptic image, but failed to generalise to other aposematic images, even though we tried with six combinations of aposematic forms. To better mimic the situation in the field, we further tested whether chicks trained with a set of four simultaneous aposematic images would generalise better. They failed to learn the discrimination task. Our data contrast with previous field studies on other poison frogs, and support a role for generalisation, and arguably not neophobia, in predator avoidance of novel aposematic forms.