Morphometric and plumage color traits
We included phenotypic traits from adults only, excluding juveniles. We recorded geographic coordinates from sampling locations when available, or assigned approximate coordinates based on detailed localities from museum labels when explicit geographic coordinates were missing. We measured seven morphometric traits (i.e., body mass, total culmen length, beak depth and width at the nares, right wing chord, right tarsus length, and tail length) using a caliper (accuracy ± 0.1mm) from 193 males, 125 females, and nine individuals of unknown sex (Table S1). We combined beak measurements with a principal components analysis to obtain a single index that reflects beak size (PC1 explained 58.6% of variance), in which larger values of PC1 mean larger beak size. We measured plumage reflectance from museum skins from 86 males (Table S1). To reduce the effect of color degradation with specimen age, we included only specimens collected within the last 50 years (Armenta et al., 2008). We measured the spectral properties from the three body patches that differ between subspecies (i.e., rump, throat, and belly), collecting five reflectance measurements (from 300 to 700 nm) per patch. We used an Ocean Optics USB2000 spectrophotometer, and a PX-2 pulsed Xenon light source (Ocean Optics Inc., Dunedin, FL, USA), with the average reading set at 50, integrating time at 16 msec, boxcar smoothing at 10, and correction for electric dark noise activated. The tip of the fiber-optic probe was covered by a black anodized aluminum cap, angled at 45 degrees to reduce specular glare (as in Uy & Stein, 2007). To ensure consistency among measurements, we calibrated the spectrophotometer with a Spectralon® white diffuse standard and dark standard (Labsphere) after every five individuals measured. Spectral data were averaged between the five reflectance measurements per patch, we applied a Loess smoothing factor of 0.3, and corrected for any negative values. We estimated mean brightness (B2; Delhey et al., 2003) for each individual patch from the reflectance reads, using the R package PAVO2 (Maia et al., 2019). Morphometric measurements and plumage brightness values were used to compare phenotypic traits between individuals of the pure parental subspecies and fit geographic clines (see below).