Aposematism is a major axis of phenotypic diversification in nymphalid butterflies

Butterfly wing patterns serve a wide array of functions in visual communication. Conspicuous color can deter predators by signaling chemical defenses, while mimicry of such warning patterns is widespread. Other functions include mate attraction, often linked to sexual dimorphism, and camouflage, typically on the ventral side. Despite this range of functions, we still lack a quantitative understanding of how different visual signals interact or constrain one another in their expression and evolutionary dynamics - largely due to challenges in capturing high-dimensional color patterns. Here, we address this challenge by using artificial intelligence based image analysis to generate the first large-scale quantitative map of wing color pattern variation in Nymphalid butterflies. Visual features (embeddings) of wing color patterns revealed that chemically defended species cluster along a major axis of variation in morphospace, which was associated with differences in aposematic color patterns and strong phylogenetic structuring. Moreover, in highly aposematic butterflies, phenotypic differences between sexes and between dorsal and ventral surfaces are reduced - likely to maintain a consistent visual signal and enhance the efficacy of predator deterrence. Finally, rates of color pattern evolution peak in species with intermediate levels of aposematism, and are biased toward the male dorsal surface. Our findings emphasize aposematic coloration as a central axis of phenotypic variation in nymphalid butterflies that structures other visual signals and shapes the tempo and mode of phenotypic evolution. Overall, our approach provides a scalable and objective framework for mapping visual signals using AI-based features, with broad applicability across various biological systems.