Flight calls of Eastern and Western Purple Finches (Haemorhous purpureus) are diagnostically different

Purple Finch (Haemorhous purpureus) populations in eastern and western North America differ in morphological traits and genetic characteristics, yet their vocal divergence has not been quantified. We analyzed flight-call structure in Eastern and Western birds using measurements of frequency and call duration. Multivariate analyses revealed strong acoustic differentiation between subspecies, driven primarily by differences in maximum frequency, minimum frequency, and frequency span. Duration was not significantly different. Variation among subspecies in flight call characteristics was much greater than variation within individuals or within subspecies. Discriminant function analyses correctly identified calls in 93% of samples that were originally identified by the recordist or by assignment based on known geographic ranges. The five misclassified birds were all recorded in the West, had been assigned subspecies status based on range, were detected only outside the typical breeding season and were statistically Easterns. After re-assignment, the chance of misclassifying a bird to the incorrect subspecies was near zero. Only one bird in our sample, recorded in the putative zone of overlap in the Pacific Northwest, was marginally intermediate in its frequency characteristics. The reliable identification of flight calls is useful because Eastern and Western Purple Finches are highly distinctive genetically and may warrant full species status. In addition, the growing popularity of public archives for bird sound recordings increases the possibilities to track movements of infraspecifically diverse taxa like Purple Finches that are highly vocal but whose plumage variation is subtle. Finally, the existence of structurally divergent calls suggests the possibility of broader ecological and evolutionary differentiation.

Flight calls are typically a unique class of short vocalizations given by birds during flight, occasionally while perched. Interest in flight calls of songbirds as markers of infraspecific phenotypic, and possibly genotypic, variability has increased because of extensive work on Red Crossbills (Loxia curvirostra) and Evening Grosbeaks (Hesperiphona vespertinus) (Groth, 1993; Sewall et al., 2004). Both of those finch species have a high diversity of diagnosably different flight calls that align with different populations within each species despite minor differences in external morphological features. Each population of birds (often but not always recognized as a subspecies) giving a distinctive flight call is referred to as a call type. Evening Grosbeaks, for example, may have 6 or more distinctly different call types whose flight calls have been shown to be reliable markers of different infraspecific groups despite birds within those 6 groups being nearly identical in external appearance (Sewall et al., 2004; Kirsch et al., 2025). Likewise, Red Crossbills form infraspecific groups with at least 12 recognizably different call types in North America as well as many others in Eurasia (Groth, 1993; Benkman et al., 2009; Martin et al., 2019; Young et al., 2024).

Structural characteristics of flight calls (e.g., frequency range, frequency modulation and duration) appear to be largely consistent temporally within call types and so represent effective markers of call-type group identity. In rare cases, call-switching may occur or structural characteristics of the main flight calls of a call-type may drift over time. In general, however, call types give flight calls that are diagnostically different from the flight calls of birds of other call types. Such markers are convenient and effective ways to identify birds that are otherwise so similar morphologically that field identification of the groups either requires exceptional conditions allowing careful inspection of plumage or bill characteristics, or is impossible as accurate identification can only be achieved by direct comparison in series of museum specimens. Diagnostically different call types provide logistical opportunities to simplify documentation of geographic ranges, track seasonal movement patterns of call-types and characterize irruptions and patterns of vagrancy, simply by recording flight calls and analyzing their structural characteristics (Robinson et al., 2024). The continuing growth of online archives where birders and other naturalists contribute digital recordings has facilitated the rapid improvement of our understanding of these topics.

Purple Finches (Haemorhous purpureus) are distributed widely across northern and western North America (Wootton, 2020). Like Evening Grosbeaks and Red Crossbills, they wander in search of tree seeds during the non-breeding season and routinely give call notes while flying, sometimes while perched. Historic examination of specimens led to identification of at least two subspecies (Ridgway, 1910). Eastern Purple Finches (H. p. purpureus Gmel.) breed across southern Canada and the northeastern United States from the Atlantic Ocean west to British Columbia. Western Purple Finches (H. p. californicus Baird) breed from southwestern British Columbia along the Pacific coast to southern California. Each has external morphological characteristics that allow field identification but require close inspection (Ridgway, 1910; Blake, 1955). Eastern adult males are more brightly colored on the crown, rump and flanks than are Western adult males, and have more obvious streaking on the back than Westerns. Eastern females are more boldly streaked ventrally with a crisper facial pattern than Western females, which tend to be washed with yellow-olive ventrally and have less clearly-defined ventral streaking. Eastern birds have a longer quill in the first than fourth primary, whereas Westerns exhibit the opposite pattern (Ridgway, 1910). Some early work suggested that two additional subspecies may occur, including one in Nova Scotia and Newfoundland and another in British Columbia. Recent publications argue that the most parsimonious interpretation is that two subspecies (H. p. purpureus and H. p. californicus) should be recognized while the other two simply exhibit aspects of clinal variation (Wootton, 2020). Recognition of two subspecies aligned with prior work on allozymes from a small sample of Purple Finches (Marten and Johnson, 1986).

A recent phylogenetic study of 267 individuals across 10 sites in northern and western North America, based on comparisons of mtDNA and microsatellite variation, indicated a strong genetic difference between the Eastern and Western subspecies (Macfarlane et al., 2016). Nucleotide differences between the subspecies averaged 2.4% and the two subspecies did not share haplotypes, forming two distinct clusters. A narrow zone of admixture in the Pacific Northwest was also detected. The distinctiveness of the two taxa aligns with the conclusion that the two currently recognized subspecies may warrant elevation to species status but such determinations remain complicated (Winker, 2021).

Here, we consider that Purple Finches form two lineages currently recognized as subspecies. We asked if flight calls, which are known to be markers of population-level differences among other finch species, are diagnostically different in Eastern and Western Purple Finches. Although no detailed study of vocal variation has been conducted previously, literature suggests that songs and calls may be recognizably different (Wootton, 2020). Given the recent genetic evidence of divergence between Eastern and Western Purple Finches, we predicted that the two groups should have diagnosably different flight calls. We evaluated the prediction by measuring and comparing recordings of Purple Finch flight calls across their North American range. We focused on whether the amount of variation among subspecies was sufficiently greater than variation within individuals and within subspecies to allow definitive identification to subspecies.

Sound recordings

We downloaded recordings of flight calls from Xeno-canto (XC; xeno-canto.org) and Macaulay Library (ML; macaulaylibrary.org; date of access: November 2025). Calls were recorded and uploaded by enthusiasts. Therefore, they represent ad hoc sampling across the North American range of Purple Finches (Figure 1). Because flight calls are brief sounds, we verified that sampling rate from the diverse sets of recorders used by contributors was at least 44 kHz before proceeding with measurements. Although use of community-sourced recordings limits standardization of spectrogram parameters, we only utilized calls when we subjectively determined that sufficient detail was preserved to provide opportunities for accurate measurements. We measured spectrograms by cursor with Raven Pro software (FFT length= 512 points, frame size 100%, overlap 50%, temporal resolution typically 1 ms and Hann window). As we discuss later, calls of the Western group, especially, included weak higher-frequency elements that were typically visible (and therefore measurable) only in the highest-quality recordings. We measured all of the best quality recordings of flight calls we could obtain and attempted to balance sample sizes of individual birds across the Eastern and Western groups. Our operational definition of quality included presentation of the spectrogram of the relevant call in sufficient detail that the characteristics we wished to measure (initial, terminal, maximum and minimum frequencies, for example) could be unambiguously determined. Thus, calls with excessive smearing or poor details of structure suggesting the calling bird was perhaps too distant from the recorder were excluded. We initially identified birds as Eastern or Western based on the identifications provided by the recordists or by geographic locations. We allowed the statistical analyses (see below) to suggest reclassifications.

We focused our measurements on frequency characteristics (initial, minimum, maximum and ending frequencies in Hz) and duration (s). We calculated change in frequency (delta frequency) as the difference between maximum and minimum frequency. When multiple calls were present within a given recording, we selected calls at random with an online random number generator (typically only 1–6 calls were within each recording) with the constraint that we measured the best available calls when a clear range in quality was present. For example, when calls were given by flying birds, the signal strength varied as a finch moved toward and away from the recordist. Furthermore, calls sometimes overlapped other sounds obscuring opportunities for accurate measurements so we selected calls whose characteristics could be clearly viewed. We chose up to 4 calls within each recording to facilitate characterization of within-individual variability.

Data analysis

We compared call structure of the Eastern and Western subspecies using multivariate analysis of variance (MANOVA; (Wilks, 1932)) with Fmax (maximum frequency), Fmin (minimum frequency), Fs (initial frequency), Fe (terminal frequency) and delta frequency (Fmax-Fmin) as response variables and Subspecies as the predictor. Duration was dropped from multivariate analyses because Welch’s t-test showed no differences among Subspecies. Significance of MANOVA was evaluated using Wilks’ λ. Welch’s t-tests were used for univariate comparisons due to unequal variances (Welch, 1938; Derrick and White, 2016). In all comparisons, we considered p<0.05 to be statistically significant. Effect sizes were quantified as Cohen’s d (Cohen, 1962).

We performed hierarchical cluster analysis (Ward’s method) to visualize groupings based on similarities in frequency measurements Fmax, Fmin, Fs, and Fe (Ward, 1963; Bridges, 1966). We then labeled calls based on the state or province where each was recorded. We looked for discordance in geographic groupings. In five cases (discussed in detail below), birds occurring in locations traditionally assumed to have only Western birds clustered with Eastern birds. Those individuals were either unlabeled to subspecies by the recordists and initially assigned by us as Westerns based on geography or were labeled as Westerns by the recordists. We inspected and re-measured recordings of those five birds to ensure no measurement errors were made.

To evaluate diagnosability of types based on the traits we measured (Fmax, Fmin, Fs, Fe), we used discriminant function analysis (DFA (Fisher, 1936)) on the set of 68 recordings (one per individual finch). Duration was excluded because it was nearly identical across subspecies and delta frequency was redundant with other variables for purposes of this analysis. Prior to analysis, we compared group-specific covariance matrices to assess whether a linear DFA (common covariance) or quadratic DFA (group-specific covariance) was most appropriate. Variances differed between subspecies, and correlation structure was not identical, indicating violation of the strict equal-covariance assumption. However, the quadratic DFA requires estimating separate covariance matrices for each group, and sample sizes were modest (East: n = 30; West: n = 38), which yielded unstable group-specific covariance estimates. Cross-validation via leave-one-out cross-validation showed that quadratic DFA did not improve classification accuracy over the linear model. For these reasons, we adopted the linear DFA with a pooled covariance matrix as the primary and most robust method.

To quantify repeatability of call structure within versus among individuals, we used one-way random-effects models treating Individual as a random intercept (calls nested within individuals) and calculated intraclass correlation coefficients (ICC; ICC(1,1)) for each variable and subspecies. Variance components and ICCs were estimated using a one-way ANOVA approach, with 95% confidence intervals computed using the corresponding F distributions. ICC values represent the proportion of total variance attributable to differences among individuals. Within-individual variability was summarized using per-individual standard deviations for each variable. All analyses were conducted in JMP Pro version 17 (JMP, 2023).

To avoid pseudoreplication in comparisons between subspecies, we used a dataset containing one randomly selected high-quality call per individual (N = 68 birds) for all MANOVA, Welch’s t-test, DFA and hierarchical cluster analyses. A second dataset including all measured calls (N = 155 calls) was used for estimating repeatability (ICC) to characterize within-individual and within-subspecies variability.

The flight calls of Eastern and Western Purple Finches are different. Eastern flight calls may be phoneticized as peet, a somewhat ringing note occurring within a relatively narrow frequency range (3–5 kHz). Westerns give a drier sounding pit that begins at a much lower frequency and ranges across a distinctly wider range of frequencies (1.2–11 kHz). Flight calls from both the Eastern and Western groups were consistently brief (Figure 2). Eastern and Western calls averaged 0.021 sec (SD = 0.009 and SD = 0.006, respectively). Duration did not differ significantly by group (Table 1). A MANOVA including Fmax, Fmin, Fs, Fe, and delta frequency revealed a highly significant effect of Purple Finch subspecies (Wilks’ λ = 0.214, F 4,63 = 58.3, p < 0.0001), indicating broad multivariate divergence in call structure. Initial frequencies (Fs) differed significantly. Eastern calls initiated at a mean higher frequency, 3217 (SD = 206) Hz whereas the mean initial frequency of Western calls was 1989 (525) Hz. In contrast, terminal frequencies (Fe) were similar. Eastern calls ended at a mean frequency of 3190 (252 Hz) similar to their initial frequencies while Western calls ended at 3343 (521) Hz, substantially higher than the mean initial frequency. Thus, the delta frequency, which described the range of frequencies covered by each call, was less for Eastern birds (617 Hz versus 2657 Hz for Westerns).

The frequency at which Western calls ended was much more variable, as revealed by the larger standard deviations. Some Western calls reached very high frequencies (up to 11.5 kHz) whereas others did not. Nine of 38 Western calls reached maximum frequencies in excess of 7 kHz. The variability could possibly result from the extreme brevity (usually less than 0.01 sec) of the highest-frequency elements of those Western calls or sampling error associated with distance between recorders and birds. Despite the variability, mean maximum frequencies of Western calls were still significantly higher than the maximum frequencies of Eastern calls (Table 1). Eastern birds in our sample were consistently less variable in all measurements except duration as indicated by lower coefficients of variation than those of Western call characteristics (Table 1). Univariate effect sizes (Cohen’s d) indicated extremely large differences between subspecies in Fmin (d = 2.67), Fs (d = 2.94), Fmax (d = −1.30), and delta frequency (d = −1.17), whereas differences in Fe (d = −0.36) and duration (d = 0.03) were small and not statistically significant. These results confirm that most of the separation is carried by frequency parameters prior to the ending frequency and not flight call duration.

Discriminant function analysis showed 63 of 68 (93%) predicted identifications aligned with assignments to Eastern versus Western made by the recordists or by our assignment based on geographical location of the recording (Supplementary Figure 1). Cross-validated accuracy of the linear DFA was ~92.6%, and a quadratic DFA (allowing unequal covariance matrices) did not improve classification. Because group-specific covariance matrices were unstable for quadratic DFA, largely due to modest sample sizes and multicollinearity in Westerns of Fs and Fmin, linear DFA provided a more parsimonious and robust solution. The five misclassifications were birds recorded in British Columbia (2) or California (3). All flight call variables measured from those five birds were not significantly different from values from the entire set of flight calls labeled as Easterns (Supplementary Table 1). Results from hierarchical cluster analysis confirmed the DFA results, including the grouping of those same 5 birds labeled as Western into the Eastern cluster (Supplementary Figure 2). The two British Columbia birds were recorded on Vancouver Island during September and may have been migrants. Likewise, two of the California birds were recorded during April and may have been migrants. Two other California recordings with Eastern characteristics were correctly labeled as Eastern by the recordists, so evidence indicates birds giving Eastern flight calls occur in California. The fifth bird, also in California, was recorded in July and could possibly be a migrant as well as evidenced by eBird records of Easterns occurring south of their breeding range in July and August even as far as the southeastern United States. Intermediate flight calls were very rare in our dataset. Only one flight call might be considered intermediate between the two subspecies. A bird from Washington in October was assigned an 81% probability of being Western and 19% probability of being Eastern. Finally, calls from Nova Scotia and Newfoundland at the far eastern extent of the Purple Finch range, an area where previous work suggested a third subspecies may reside, fell well within the sets of quantities characterizing other Eastern Purple Finches.

To partition variation in call structure within and among individuals, we estimated repeatability (intraclass correlation coefficients; ICC) separately for the two subspecies using one-way random-effects models with Individual as a random factor (Table 2). For Eastern flight calls, ICCs ranged from 0.47 to 0.79 (95% CI ≈ 0.24–0.82) across the six acoustic variables (Fmax, Fmin, Fs, Fe, delta frequency, duration), indicating that 47–79% of the total variance in call structure was attributable to differences among individuals. Repeatability was highest for Fmax and duration, moderate for Fmin, delta frequency, and Fe, and lowest for Fs, although still positive. For Western flight calls, ICCs were high for all variables except for Fe (≈0.17, CI overlapping zero). Thus, in both subspecies, call structure was consistently more variable among individuals than within individuals, with Westerns exhibiting particularly strong individual distinctiveness in several frequency-based traits (Supplementary Figure 3). Individual variation was present within each subspecies but the multivariate divergence between subspecies greatly exceeded within-subspecies variability. Thus, individual differences do not obscure the strong diagnostic separation of Eastern versus Western flight calls.

Eastern and Western Purple Finches both give short duration (0.02 sec) flight calls that differ consistently in a small number of characteristics including initial frequency, maximum frequency and the span of frequencies within each call. Eastern flight calls are more constrained, spanning about 23% of the range of frequencies spanned by an average Western call. Easterns start at a higher frequency (averaging 3.2 kHz) than Westerns (2.0 kHz) and reach lower maximum frequencies (3.8 kHz) than most Westerns (6.0 kHz). The maximum frequencies reached by flight calls of Westerns is highly variable, with some reaching values typical of Easterns and others exceeding 11 kHz. The high variability of Western calls, generally, is revealed by their higher coefficients of variation for all characteristics we compared except for duration. Whether variability in maximum frequency of Western calls reflects true variation or sampling error associated with effects of distance between a calling bird and a recordist or recorder characteristics remains unclear as the elements of Western calls reaching the highest frequencies are extremely brief and may be too weak to propagate well at distance. Despite that source of variation, the characteristics we measured allow identification of flight calls to the two currently recognized subspecies with a very high level of confidence.

The variables showing the highest repeatability (strongest individual signatures) were also those that best distinguished Eastern from Western flight calls in multivariate space. Specifically, minimum and maximum frequency and delta frequency, which had strong intraclass correlation coefficients and large effect sizes, were the primary contributors to separation of subspecies in the DFA and accounted for most of the multivariate divergence detected by MANOVA. This convergence of results indicates that the same acoustic dimensions that differentiate individual birds also differentiate the two subspecies. The effect sizes we observed (|d| up to 2.9) far exceed the standard diagnosability thresholds used in avian taxonomy (Patten and Unitt, 2002). Taken together, these analyses demonstrate that Purple Finch flight calls show clear, statistically robust subspecies-level divergence.

Aside from the Eastern versus Western differences, we saw little evidence for additional geographic variation. Hierarchical cluster analysis discerned two major groups. The Western group formed two subclusters with individuals scattered from California, Oregon, Washington and British Columbia across both clusters, indicating lack of geographic partitioning within the Western subspecies. Likewise, the Eastern cluster showed no obvious evidence of geographic structure. We found just three recordings of flight calls from Nova Scotia and Newfoundland, an area formerly suggested to be home to a different subspecies (Wootton, 2020). Those recordings were extremely similar to the other Eastern flight calls and provided no evidence for additional separation.

The distinctive flight call differences we describe here confirm that migration and movement characteristics of the two subspecies can be effectively studied by recording their calls. This offers an important logistical opportunity to track subspecies as flight calls are given year-round without the limited seasonality of songs. Although distinctive plumage differences do occur, they are subtle and require either close inspection by experienced observers or photographic verification (Ridgway, 1910; Blake, 1955). Furthermore, Purple Finches often occupy the canopy of forests and, except for time spent at bird feeders, can be difficult to visually inspect carefully (Wootton, 2020). Our understanding of seasonal movement patterns and of the geographic ranges, especially in the zone of overlap in the Pacific Northwest, can be quickly improved with archival in public databases of flight call recordings associated with accurate geolocation and date information. We recommend that flight calls deposited in archives such as xeno-canto.org and Macaulay Library be tagged with the search term “flight call” to facilitate discovery by analysts and that location information be as precise as possible.

We found at least seven examples of birds in the West whose calls fit better as Eastern Purple Finches. Two were already correctly labeled by the recordists as Easterns. Five misclassified recordings were considered misclassified because they were labeled as Westerns either by us or the recordists based on geography. Statistical tests revealed they were indistinguishable from Easterns (Supplementary Table 1). Two potential explanations seem possible here. One, some Western Purple Finches may give calls that are nearly identical to those of Eastern flight calls or, two, Easterns occur occasionally in the Pacific states. We consider the latter to be the most parsimonious explanation given the statistical similarities with all other Eastern recordings. We did not find any recordings in the Eastern range that fit Western better. The range of Easterns extends as far west as British Columbia and the genetic study indicated that some birds in southwestern British Columbia and coastal Oregon were intermediate, suggesting some hybridization (Macfarlane et al., 2016). More sampling of calls in these areas would help determine if calls are intermediate between more typical Eastern and Western calls outside that zone of genetic intergradation. We found only one call that was slightly intermediate, a bird recorded in Washington during October. Given the date, we cannot determine if it was a migrant or local resident. Therefore, we did not see obvious evidence of intermediate calls that might be attributed to hybridization effects but recordings of flight calls during the breeding season from the Pacific Northwest contact zone in British Columbia are very limited.

The two currently recognized subspecies appear to be genetically distinct enough, with minimal evidence of genetic admixture in a narrow zone of overlap, to potentially be considered different species (Macfarlane et al., 2016). The distinctiveness of their flight calls aligns with this view. Although we know of no formal studies comparing the songs of the two subspecies yet, anecdotal information suggests that the advertising songs also differ in consistent ways associated with typical frequency ranges, duration and pacing of notes within songs (Wootton, 2020). The only previous mention in the literature that Purple Finches may have two different call types is in the recent Stokes and Young book (Stokes and Young, 2024), where they labeled Westerns as type 1 and Easterns as type 2. If the two current subspecies are eventually recognized as genetically distinct enough to warrant recognition as species, we recommend consideration be given to re-naming them with geographic descriptors aligning more closely with their distributions. Because the range of Eastern Purple Finch includes North America from Newfoundland to British Columbia, the name Northern Purple Finch is more accurate as their breeding range currently spans across northern North America. Likewise, given that Easterns occur also in western Canada, an alternative and more accurate name for Western Purple Finch could be Pacific Purple Finch.

The flight calls of Eastern and Western Purple Finches are distinctly different. Flight calls of Easterns could be phoneticized as peet as they occur within a relatively narrow frequency range and have a brief, but somewhat ringing quality to them, whereas Westerns give a dry pit that begins at a much lower frequency and ranges across a distinctly wider range of frequencies. The variation among subspecies is substantially greater than the variation within subspecies or individuals, indicating that field identification can be highly reliable. More high-quality recordings are needed. In particular, calls from the Pacific Northwest zone of overlap should be prioritized to evaluate the possibility of hybridization effects on call traits. In addition, flight calls from Nova Scotia and Newfoundland, a region thought historically to have another distinctive subspecies but could simply be part of a cline within the Eastern subspecies, would be informative to obtain. Finally, still virtually absent from public archives of ornithological data are records of Purple Finches where we have recordings of flight calls and photos of the same individuals (and of sufficient quality to identify subspecies). To definitively determine associations between flight call characteristics and subspecies identity, we need recordings from birds that are collected and subsequently genotyped. Short of that goal, recordings combined with either video or diagnostic photographs can be useful as well. Such additional information would confirm what is already a clear signal of subspecific differences in flight calls.