Traditional bioacoustic analyses and machine-learning methods indicate weak vocal dimorphism in four Cerrado antbird species

Abstract Sexual dimorphism in birds is often associated with plumage. Yet, the strength and prevalence of vocal sexual dimorphism remain poorly quantified for many Neotropical passerines. In particular, Cerrado antbirds show pronounced plumage dimorphism, but whether this is accompanied by consistent vocal divergence is unclear. This gap limits our understanding of how different communication modalities evolve under sexual and social selection. Here, we evaluated sexual differences in vocal behavior in four Cerrado antbird species ( Thamnophilus doliatus, T. pelzelni, Herpsilochmus atricapillus and H. longirostris ). We employed a comparative framework combining traditional bioacoustic measurements, mel-frequency cepstral coefficients, and deep-learning–based audio embeddings, and formally tested sex differences using supervised classification approaches. Across species and feature sets, classification performance was consistently low and only marginally better than random expectations, indicating weak and unreliable separation between male and female vocalizations. Results were largely concordant across methods, despite differences in feature abstraction. Together, our findings indicate that vocal sexual dimorphism in these species is minimal, contrasting sharply with their pronounced plumage dimorphism. This suggests that strong sexual differentiation may be expressed predominantly in one signaling modality, with limited evolutionary pressure for parallel divergence in vocal traits. By integrating interpretable and machine-learning–based acoustic analyses, our study provides a robust framework for assessing vocal dimorphism and highlights the importance of multi-modal perspectives in studies of avian communication and sexual selection.