Flight call variation and vocal repertoire in the Hooded Grosbeak Hesperiphona abeillei

Many species in the cardueline finches (Fringillidae) display complex intraspecific diversity in social vocalizations, providing a rich system to study the evolution and maintenance of divergence in acoustic signals. Yet, formal descriptions of vocal variation are unavailable for a majority of Fringillids globally, especially in tropical taxa. Here we describe geographic variation in vocalizations of Hooded Grosbeaks (Hesperiphona abeillei) and qualitatively compare the call structures with its congener, Evening Grosbeak (Hesperiphona vespertinus) using recordings from two community science databases that archive sound files along with geo-referenced location information. Recordings identified as social calls of Hooded Grosbeaks were classified into potential groupings through independent qualitative (location-blind expert opinion) and quantitative (cluster analysis and discriminant function analysis) methods. Results revealed a complex vocal repertoire containing more diversity in song and plasticity in call elements than that of the Evening Grosbeak. Both classification methods concurrently identified at least three distinct subgroups of flight calls, one of which was geographically differentiated (southern Mexico to Guatemala) from the other groups, which appeared to be more widespread across the Sierra Madre in Mexico. The combined structural and geographic differences among populations suggest the existence of a complex of learned call types similar to those of the Evening Grosbeak and Red Crossbill. Establishing further understanding of the geographic variation in Hooded Grosbeak calls will require additional geo-referenced recordings, as well as further field studies clarifying the behavioral significance flight calls, songs, and other vocalizations in the Hooded Grosbeak.

Variation in acoustic signals, which facilitate social and reproductive behaviors in a taxonomically diverse group of vertebrates, has emerged as a promising indicator of the fault lines in geography, behavior, and ecology that delineate populations (Angulo and Reichle, 2008; Saitoh et al., 2015; Toews and Irwin, 2008). The discovery of such vocally distinct populations can lead to new insights about a particular species’ ecology and evolution that might otherwise be obscured by morphological similarity. However, even in well-studied taxonomic groups such as the Nearctic avifauna, the presence, degree, and significance of intraspecific signal variation often remain unclear (Carpenter et al., 2022; Robinson et al., 2024).

One such group, the cardueline finches (Family: Fringilldae), have been a focal point of recent study due to intraspecific diversity in social call dialects (henceforth “call types”) (Groth, 1993; Benkman et al., 2009; Kirsch et al., 2025). For example, the Red Crossbill (Groth, 1988, 1993) and the Evening Grosbeak (Sewall et al., 2004; Kirsch et al., 2025) have been shown to exhibit discrete vocal types among otherwise morphologically similar populations. Describing and quantifying these vocal types has proved fruitful by leading to subsequent description of differences in foraging behavior, migratory patterns, breeding ranges, bill morphology, and mating preferences that hints at the potential for morphologically cryptic evolution (Benkman et al., 2009; Centanni et al., 2024; Duman and Hahn, 2024). Moreover, the apparent absence of vocal variation among intraspecific populations in ecologically similar and closely related species, such as the Pine Siskin (Spinus pinus) and the White-winged Crossbill (Loxia leucoptera) provides a contrast which further highlights the uncertain evolutionary drivers of call type formation.

While vocal variation has been studied extensively in some Nearctic and European fringillids (Groth, 1993; Martin et al., 2020; Robinson et al., 2024), many of their tropical and eastern palearctic relatives remain obscure, leaving critical gaps in our understanding of the call type phenomenon and its origins. A prime candidate for vocal study is the Hooded Grosbeak (Hesperiphona abeillei), known from less than 200 recordings across a geographically diverse range from northern Mexico to Central America (Carter, 2024). The closely related Evening Grosbeak (Hesperiphona vespertinus) presents a well-established example of quantitatively distinct flight call types which may roughly align with established subspecies groups (Grinnell, 1917; Robinson et al., 2024; Kirsch et al., 2025). Flight calls are the critical vocalization for social coordination of movements in the Fringillidae, appearing to play an especially central role in cultural recognition and potentially facilitating population differentiation in Evening Grosbeaks and the related Red Crossbill (Loxia curvirostra). Moreover, Evening Grosbeaks are one of few passerines that appears to lack a complex reproductive song with elements distinct from social calls, which is unusual given the high complexity of songs in the Fringillidae generally (Pieplow, 2019). Instead, the species may rely on sequences of stereotyped flight calls in the place of a typical song (Pieplow, 2019; Duman and Hahn, 2024), suggesting that these short vocalizations may be of outsized social and evolutionary importance within Hesperiphona. Very little research has focused specifically on the Hooded Grosbeak and flight calls are not formally described in the species, but its large range over a series of partially contiguous cordillera suggests a strong potential for the existence of distinct call types (Erickson and Valdés, 2022; Carter, 2024).

Here, we analyzed 134 recordings of Hooded Grosbeak songs, flight calls, and trills to quantitatively characterize their structural elements, identify geographic variation, and assess potential homology with Evening Grosbeak flight calls and trills. Our primary goals were to provide the first formal description of the Hooded Grosbeak’s vocal characteristics and repertoire, identify any distinct vocal/call type groups, and compare Hooded Grosbeak and Evening Grosbeak vocalizations to better understand their functions and evolutionary origins.

Recording review

To investigate vocal variation in Hooded Grosbeaks, we downloaded all (n = 134) Hooded Grosbeak recordings archived in Xeno-Canto and the Macaulay Library, the two most robust digital archives for neotropical birds (Macaulay Library, 2024; Xeno-Canto, 2024), which we believe to together represent most or all of the accessible audio data for this species. Recordings were reviewed regardless of quality. For each downloaded audio file, we recorded the location, date, and used Raven Pro 1.6 (K. Lisa Yang Center for Conservation Bioacoustics, 2024) to measure and analyze all recordings with default resolution and contrast settings and a window size of 11 kHz x 3s. We employed a two-step approach for sonogram review.

In the first step of review, we manually classified the vocal variations into three putative categories: songs, flight calls, and trills. Manual classification was employed prior to cluster analysis to flag calls potentially misidentified at the species level and to sort the vocalizations into biologically significant units for algorithmic review. In each recording, we identified each unique call or call sequence variation and recorded the location on the sonogram with a screenshot. Vocal sequences with at least three temporally distinguished elements (defined here as a single, continuous sound or “syllable”) or which were plastic (appearing as modifications of the same sequence throughout the recording) were considered to be putative songs. Stereotyped single vocal elements were considered to be putative flight calls. Rapid sequences of multiple modulated sub-elements (producing a uniformly harsh or buzzy tone) were considered trills. Sequences of three or more stereotyped elements given at regular intervals were not classified, but the individual elements were considered in the second step of review.

In the second step of manual review, two of us (CTC and MAY) attempted to identify significant and repeated variation among flight calls and trills. This process partially replicated the methods used to identify Evening Grosbeak call types in Sewall et al. (2004) to provide a baseline against which to compare the more objective hierarchical cluster analyses, which can be vulnerable to over- or under-classification depending on input parameters. For each unique flight call or trill variation within each recording file, we used Raven Pro 1.6 to measure duration (s), beginning and ending frequency (Hz), maximum frequency and time at each. Based on qualitative comparisons and measurements, each reviewer made independent attempts to identify discrete categories with less internal than external variation (i.e. potential call types) within the flight calls and trills. After sorting the calls into potential type categories, we compared our results and established common categories that we had agreed upon, leaving calls that we sorted differently unclassified. Reviewers were blind to recording metadata (i.e. location and date) during the process. Potential flight calls that were interspersed among plastic song sequences were categorized as “Type 0”, or possible song elements. Apparent flight calls that did not match at least two other recorded flight calls were categorized as “Type 4”, or unknown flight calls.

Data analysis

As a quantitative test of our qualitative flight call distinctions, we performed a Hierarchical Cluster Analysis (Ward’s Method) on all putative or potential flight call recordings. Features analyzed included duration, initial, maximum, and terminal frequency, total change in frequency, and change from maximum to terminal frequency. Analysis was performed without reference to qualitative assignments of call type to allow the clustering algorithm to operate independently using only spectrographic characteristics and derived measurements. Our goal was exploratory: to determine whether calls formed natural groupings in multivariate acoustic space, rather than to impose or test a predefined classification. Following initial clustering, we visually inspected cases in which calls appeared to be strongly discordant with nearby observations in multivariate space. For these cases, we revisited the original spectrograms in Raven Pro to evaluate whether measurement error had occurred (e.g., incorrect placement of temporal or frequency bounds). Corrections were made only when discrepancies between the recorded values and the spectrogram were clear and unambiguous. No adjustments were made based on cluster membership alone. After correcting these verified measurement errors (N = 4; <3%), we repeated the cluster analysis on the updated dataset. We evaluated the consistency of inferred structure across independent approaches, including qualitative classification, hierarchical clustering, and discriminant function analysis. All features that were included in the cluster analyses were also included in the DFA and all samples retained their call-type label we previously identified with the cluster analysis. We left variables untransformed as none varied significantly from normal distributions. All analyses were performed in JMP Pro 17 (JMP, 2018).

The Hooded Grosbeak vocalizations we examined were broadly variable in element number, plasticity, and call structure. Across 134 recordings, we identified 25 songs, 44 flight calls, and 13 trills. Many vocalizations remained unclassified, most of which were either potential disjunct song phrases (Figure 1) or vocal elements which we could not verify to be Hooded Grosbeaks.

Hooded Grosbeak songs

Many longer call sequences were classified as songs and included variable series of apparent flight calls, trills, and plastic elements not identified aside from in such sequences (Figure 2). The presence of these plastic elements was found regularly and across recordings from different regions, indicating the presence of individualized song as a widespread element of the species’ vocal repertoire.

Hooded Grosbeak flight calls

Qualitative analysis of flight calls revealed three apparently discrete call types (Figure 3). The ‘short rising’ flight call, tentatively type 1 (n = 24), is a brief, clear-toned ascending element reminiscent of an Empidonax flycatcher (Empidonax spp.) ‘whit’ call in tone quality. The ‘long rising’ flight call, tentatively type 2 (n = 10), begins in a rapid ascent like type 1, but quickly levels out to create a whistled quality, similar to a Swainson’s Thrush (Catharus ustulatus) flight call. Putative type 3 (n = 15), the ‘peaked descent’ flight call, is much higher in pitch, and consists of a rapid (nearly vertical) ascent to a peak and then a long descent, creating a quality reminiscent of the poorly known Evening Grosbeak type 5, which is hypothesized to overlap in range with Hooded Grosbeaks (Kirsch et al., 2025; Sewall et al., 2004). Some recordings of this type were accompanied by a distinctive, peaked polyphonic element above the highest point in the call. Flight calls which could not be sorted clearly into a type were labeled “type 4” (n=8), and calls which were inseparable between flight calls and disjunct song elements were labeled “type 0” (n=16).

All three analytical approaches (qualitative classification, hierarchical clustering, and DFA) converged on the same primary categorization structure, with only minor exceptions (Figure 4). Each type formed its own distinct cluster, with type 3 the most distinct. Types 1 and 2 were more similar, with type 1 displaying more within-type variation. When included, calls classified as “type 4” were clustered closest to the type 2 or type 3 branches. Cluster analysis by region revealed a close alignment between the core group of type 1s and the southern portion of the species’ range and some geographic overlap between types 2, 3, and the unknown “4” recordings, matching the results of our qualitative analysis (Figure 5). Notably, the several calls assigned to the unknown “type 4” group clustered together despite clear sonographic variability. Similarly, canonical discriminant function analysis sorted all of the calls in the same pattern as qualitative analysis (Figure 6), save a single type 3 which more closely matched the unknown “4” calls. The entropy R-squared of this analysis was very high, at 0.94.

Among the three putative types, recordings of type 1 were geographically isolated from those of types 2 and 3, whereas the latter two showed significant range overlap. All type 1s were recorded in Guatemala, Honduras, El Salvador, and the southernmost Mexican state of Chiapas (Figure 5). All type 2 recordings were from the northern Mexican states of Sonora, Chihuahua, Coahuila and San Luis Potosí. Type 3 recordings were the same northern Mexican states as type 2, with the addition of the Central states of Guerrero, Mexico City, and the State of Mexico. Of the eight unsorted recordings, seven were from states not represented among those sorted into call types. Of the 16 recordings which could not be conclusively classified as songs, trills, or flight calls, seven were from states not represented among the sorted flight calls.

Hooded Grosbeak trills

Vocal elements that consist of rapid sequences of sub-elements that may modulate sharply and repetitively in frequency (producing a uniformly harsh or buzzy tone) were considered trills. Trills were identified in 13 recordings, all from northern and central Mexico (Figure 7). Compared to Evening Grosbeak trills, Hooded Grosbeak trills were longer and tended to contain a greater total number of modulated peaks. While there was notable variation between trills on different recordings, we did not sort them into types due to the small sample size and lack of clear, discrete groupings. Additional vocal elements which could not be easily categorized were or contained both trills and clear elements in possibly sequences were considered unknown vocalizations (Figure 1).

Our examination of available Hooded Grosbeak recordings reveals a broad vocal repertoire, including an array of songs, flight calls and trills. Furthermore, we found three quantitatively distinct call types of Hooded Grosbeak, each of which differed due to a combination of measured flight call duration and frequency variables. Despite the quantitative similarities of the “type 4” group, we do not propose it as a call type here, since these recordings appear to show within-group sonographic variability comparable to the between-group variation among the three described call types. However, the presence of several unique calls that cluster together in “type 4” suggests that the description of additional call types is likely with a larger recording sample size. Mapping the location of the call types revealed a distinct clustering of type 1 birds in the southern extent of the Hooded Grosbeak range, while range limits between types 2–4 remain unclear and require further study.

Interestingly, the Hooded Grosbeak’s vocal repertoire is apparently more diverse than that of the Evening Grosbeak. The point of greatest fundamental contrast between the two species is the prevalence of a song with plastic elements clearly distinct from social calls in Hooded Grosbeaks. One possible explanation for the observed vocal differences in these species is differences in reproductive life history. Notably, type 4 Evening Grosbeaks in Colorado, USA apparently arrive on the breeding grounds already paired and do not defend a territory (Scott and Bekoff, 1991). Because complex songs may fulfill competitive or territorial purposes, the reduction of song complexity in Evening Grosbeak could suggest a reduced need for these functions compared to Hooded Grosbeak. However, the extent to which Evening Grosbeaks use semi-plastic call sequences as a “song” (Kirsch et al., 2025; Pieplow, 2019) and the degree of difference in breeding behavior between the two species remains unclear. Regardless, the presence of complex phrases in Hooded Grosbeak song clearly distinct from social calls is striking, especially given that the closely related Hawfinch (Coccothraustes coccothraustes) along with the majority of Nearctic Fringillidae also possess distinct songs (Kirsch et al., unpublished data, Pieplow, 2019). Thus, the simplification or possible absence of Evening Grosbeak song may represent an innovation unique to that species’ vocal behavior and warrants further investigation.

Despite this marked difference in vocal behavior, both qualitative and quantitative analyses suggest that the two species of Hesperiphona may exhibit a similar system of flight call differentiation among cryptic geographic subpopulations. This is evidenced by the apparent presence in Hooded Grosbeaks of at least three discrete flight call variants which are delivered individually or in couplets at irregular intervals (similar to the flight calls of Evening Grosbeaks and other cardueline finches). Additionally, we found that at least one type corresponds clearly to a unique geographic range, adding to its potential ecological and evolutionary significance. The geographic distinction between recordings containing putative call type 1 and those containing other types is especially distinct, with the vast majority of flight calls recorded in Chiapas, Mexico and Central America consistent with this type. This pattern suggests that type 1 could represent a distinct subpopulation or subspecies, though more data is needed to verify these patterns. Geographic or phenological distinctions may also exist between putative types 2 and 3, but given the presence of both types in northern Mexico and the small available dataset, elucidating these differences would require targeted field research in both ranges of the Sierra Madre north of the Isthmus of Tehuantepec and in the apparent overlap zone in the states of Hidalgo, Mexico, Queretaro, and San Luis Potosí.

The most salient limitation to our work here is the rarity of Hooded Grosbeak audio recordings and the lack of literature on the species’ vocal behavior. In the absence of direct field observations that verifiably connect audio recordings to specific territorial or social behaviors, classification of vocal elements as songs, flight calls, and trills was based on comparisons to the calls of related taxa and the limited behavioral elements discernable from recordings (e.g. changes in distance from the microphone during calling suggests a bird is in flight). Without the critical context of field observations, it is possible that a very small number of vocalizations classified as “flight calls” could represent unusual song variations or that other flight call variants could have been mislabeled as song elements. Nonetheless, based on many collective years of field experience with Evening Grosbeaks and other fringillids, we are confident that our work here approaches the most accurate account of the species’ vocalizations possible based on the available data.

Our findings augment a growing body of work which demonstrates the prevalence and potential evolutionary significance of intraspecific vocal diversity in fringillid finches and suggests that the phenomenon of discrete flight call types is not limited only to Evening Grosbeaks and the Red Crossbill (Loxia curvirostra) complex. Call types can illustrate the ecological or geographic boundaries between subpopulations and have been used to describe new cryptic species, underscoring the importance of further work to understand vocal variation in Hooded Grosbeaks and other species. From nearly 150 recordings of this neotropical species, we were able to distinguish at least three putative vocal subpopulations. Family Fringillidae contains more than 200 species, and most of them occur outside of the regions with the largest library of audio recordings (Europe and temperate North America) (Winkler et al., 2020). With additional field work, it is likely that many more subpopulations or call types are waiting to be described. Recognition of such infraspecific groups expands our understanding of each species’ ecology, evolution, and conservation. Field studies on flight call variation, genetics, and ecology of Hooded Grosbeak and other Neotropical finches are clearly needed to illuminate the nature, extent, and implications of vocal diversity in this unique family.