Ancestral state reconstruction of acoustic traits in tinamous (Aves: Tinamidae): a first approximation to the evolutionary history of complex phenotypes

Reconstructing ancestral acoustic signals can reveal which aspects of animal communication are conserved, transformed, or reorganized during evolutionary diversification. However, these signals have a complex structure that is difficult to represent using conventional descriptors. Here, we used deep-learning models to characterize acoustic structure in Tinamidae, a lineage of Neotropical birds well suited for exploring the evolutionary history of vocal signals under a phylogenetic comparative framework. We processed 2,630 recordings from 46 species, extracted high-level acoustic traits, and reconstructed ancestral states using a dated phylogeny and a sample of posterior trees. We then sonified these reconstructions to interpret hypothetical acoustic structures at ancestral nodes and their evolutionary trajectories. We assessed methodological uncertainty and compared the reconstructions with ancestral states estimated from conventional acoustic traits. High-level acoustic traits showed strong phylogenetic signal and were best described by an Ornstein-Uhlenbeck model. Ancestral sonifications retained interpretable temporal and spectral organization, and spectral centroid showed strong consistency between conventional and high-level reconstructions. Overall, our results suggest that acoustic evolution in Tinamidae has unfolded within functional limits that allow variation among lineages without indefinite phenotypic drift. We emphasize that these reconstructions should be interpreted as hypotheses about ancestral acoustic structures, not as replicas of particular signals.