Phenotypic and vocal differentiation among subspecies of Piculus flavigula (Aves, Picidae) suggest a major taxonomic reassessment

Widespread species with visibly variable phenotypes often conceal significant evolutionary diversity. The Yellow-throated Woodpecker Piculus flavigula is a Neotropical species distributed across the Amazon and the Atlantic Forests, and is currently classified into three allopatric subspecies: P. f. flavigula (northern Amazonia), P. f. magnus (southern Amazonia), and the Atlantic P. f. erythropis. Although these groups are diagnosed based on geography and general appearance, their morphological and vocal distinctiveness have not yet been formally evaluated. We assessed their phenotypic and acoustic variation using traditional morphometric and vocal traits, combined with automated color pattern segmentation and machine learning-based acoustic analysis. A total of 113 specimens (65 males and 48 females) were measured for wing chord, tail, culmen, and tarsus length. Plumage color patterns were analyzed using the R package recolorize, which enables segmentation from uncalibrated standardized photographs. Additionally, 117 recordings were analyzed for five acoustic parameters (bandwidth, duration, peak, and minimum and maximum frequencies) and further processed through BirdNET, a neural network-based algorithm that extracts 1,024-dimensional feature embeddings. Multivariate statistical tests (MANOVA, LDA) and supervised classification (SVM) were used to assess group separability. We found significant morphometric differences among subspecies, particularly in culmen and tail length (MANOVA p = 0.016). While P. f. erythropis showed the most distinctive phenotype, including a unique barred ventral pattern and higher peak frequencies (p < 0.001), comparisons between P. f. flavigula and P. f. magnus also revealed consistent, but subtler, differences in tail (p = 0.022), culmen (p = 0.015), and vocal parameters. These two Amazonian subspecies showed partial overlap in morphological measurements but moderate separation in acoustic space. Classification models based on BirdNET embeddings achieved 86% accuracy in distinguishing vocalizations among groups. These findings support the phenotypic and acoustic distinctiveness of P. f. erythropis, while also suggesting divergence between P. f. flavigula and P. f. magnus, providing robust evidence for a re-evaluation of taxonomic limits within the Piculus flavigula species complex.