JOURNAL OF

AVIAN BIOLOGY
Article
Song but not plumage varies geographically among willow
flycatcher Empidonax traillii subspecies
Sean M. Mahoney, Matthew W. Reudink, Bret Pasch and Tad C. Theimer

S. M. Mahoney (https://orcid.org/0000-0001-8446-423X) ✉ (sean.mahoney@nau.edu), B. Pasch and T. C. Theimer, Dept of Biological Sciences, Northern
Arizona Univ., Flagstaff, AZ, USA. – M. W. Reudink, Dept of Biological Sciences, Thompson Rivers Univ., Kamloops, BC, Canada.

Journal of Avian Biology
00: 1–12,
2020
2020:
e02621
doi: 10.1111/jav.02621

Subject Editor: Martin Paeckert
Editor-in-Chief: Thomas Alerstam
Accepted 29 October 2020

Plumage and song are important signals used by birds to attract mates and repel rivals.
Divergence in sexual signals can lead to reproductive isolation among incipient species,
but the relative importance of each modality may vary among taxa. Tyrannid flycatchers exhibit evolutionarily conservative plumage coloration but distinct song structure
among subfamilies and species. Thus, tyrannids are an interesting group in which to
study the relative role of plumage and song in contributing to population divergence.
In this study, we assessed character divergence among four willow flycatcher Empidonax
traillii subspecies by measuring spectral reflectance of plumage modeled in tetrahedral
colorspace from museum specimens collected on putative breeding grounds. We also
quantified differences in song structure based on publicly available and field-recorded
songs across the species range. Using unsupervised and unbiased clustering algorithms
that assigned group membership independent of a priori taxonomic designations, we
found that currently recognized subspecies did not consistently sort in accordance with
subspecies designation based on plumage color. However, song analyses grouped birds
into two clusters; one that included 89% of all putative E. t. extimus, and another that
included 100% of specimens designated as E. t. adastus, E. t. brewsteri, E. t. traillii
and a small percentage of E. t. extimus (11%). Our results are consistent with previous
hypotheses of conservative plumage evolution in tyrannids and species differentiation
based on song, and support the subspecific status of E. t. extimus.
Keywords: plumage variation, song variation, subspecies, willow flycatcher

Introduction
Many animals use acoustic and visual signals to discriminate between conspecifics
and heterospecifics in various social interactions (Bradbury and Vehrencamp 1998).
Signal divergence among populations resulting from natural selection, sexual selection,
genetic drift, cultural drift, environmental conditions (Hill 2006, Wilkins et al. 2013),
arbitrary mate choice (Prum 2010) or some combination of all these phenomena can
be important in reproductive isolation and speciation (Mayr 1942). Although genomic
techniques are often used to delimit species (Barrowclough et al. 2016), acoustic and
visual characters can be a powerful alternative (Mallet 2005, Tobias et al. 2010). In
birds, plumage differences are often used to delimit species (e.g. Dendroica warblers,
––––––––––––––––––––––––––––––––––––––––

www.avianbiology.org

© 2020 Nordic Society Oikos. Published by John Wiley & Sons Ltd

1

Shutler and Weatherhead 1990, Lovette and Bermingham
1999, Toews et al. 2016) but when plumage variation is limited, song may provide a more robust indicator of species limits (Martens et al. 2003, Päckert et al. 2003, Rheindt et al.
2008).
Plumage evolution in tyrannid flycatchers is evolutionarily
conservative (Zink and Johnson 1984) resulting in a family of many sibling species that are difficult to differentiate
by plumage alone (Stein 1958, 1963, Rheindt et al. 2008).
Morphologically similar species are differentiated instead by
differences in song or behavior (Stein 1963, Johnson and
Cicero 2002, Rheindt et al. 2008). In spite of the general
lack of strong plumage variation among Empidonax flycatchers, subspecies within this genus were often discriminated by
plumage. For example, the original subspecies designations
of the four subspecies of the willow flycatcher currently recognized by the US Fish and Wildlife Service (USFWS; E. t.
traillii in the eastern USA, E. t. brewsteri in the northwestern
USA, E. t. adastus in the interior west and E. t. extimus in
southwest riparian areas) were based on qualitative plumage
descriptions (Brewster 1895, Oberholser 1918, 1932, 1947,
Phillips 1948, Aldrich 1951, Unitt 1987, Browning 1993).
These studies were subsequently criticized due to a lack of
rigorous statistical analyses, small sample sizes, and conclusions based on specimens collected over a relatively small
geographic area (Zink 2015). More recently, a study using a
colorimeter to measure plumage reflectance found differences
among subspecies in mean values on the back and crown but
substantial overlap among individuals (Paxton et al. 2010).
The colorimeter used in that study did not assess UV reflectance, which is an important part of visual signals in birds
generally (Cuthill et al. 2000), and may contribute to sexual
dichromatism in willow flycatchers specifically (Eaton 2007).
Paxton et al. (2010) was subsequently criticized because measurements were based on wild birds captured and released
in the field and therefore specimens were unavailable for
later reanalysis (Zink 2015). Only one study has rigorously assessed geographic song variation in willow flycatchers; Sedgwick (2001) found differences in song structure
between E. t. adastus and E. t. extimus but only compared
two subspecies and songs were not catalogued nor made publicly available. Finally, all studies assessing willow flycatcher
phenotypic variation based analyses on specimens grouped
a priori into presumed subspecies and so comparisons were
not made independent of taxonomic identity. The validity
of subspecific designations in this group have important
implications because one subspecies E. t. extimus is listed as
endangered (US Fish and Wildlife Service, USFWS 1995)
and uncertainty around subspecies designation hinders continued conservation protection (Haig et al. 2006, Zink 2015,
Theimer et al. 2016).
In this study, we quantitatively compare plumage variation
across all four subspecies of willow flycatchers using museum
specimens and methods that include reflectance into the UV
spectrum and incorporate models of avian visual sensitivity.
These methods account for the potential presence of plumage signals that would not be detected by human-biased

2

assessments of color based on features like lightness, saturation and hue used in previous analyses (Paxton et al. 2010).
Investigation of the potential for UV signal divergence is particularly warranted given that Eaton (2007) found evidence
for sexual dichromatism in UV signals from both the back
and belly of Empidonax traillii (Eaton 2007, their Supporting
information). In addition, we expand on Sedgwick’s (2001)
original comparison of songs of E. t. extimus and E. t. adastus by including songs of all four subspecies using publicly
available recordings. Importantly, we utilize unbiased and
unsupervised analyses that assess differences independent of
taxonomic identity.

Methods
Reflectance measurements

To assess plumage differences among willow flycatcher subspecies, we measured plumage reflectance across the avian
visual range (300–700 nm) on the back, belly, breast, crown,
nape and throat of willow flycatchers (E. t. adastus, n = 38; E.
t. brewsteri, n = 30; E. t. extimus, n = 21; E. t. traillii, n = 18)
and, as an outgroup, alder flycatcher (n = 8) museum specimens (Supporting information). Specimens were collected
from across the breeding range of willow and alder flycatchers (Fig. 1A). We restricted the specimens included in our
study to adults collected during the peak breeding season (15
June–25 July). We assigned each specimen to a subspecies
based on its collection location using the putative subspecies
maps from USFWS (1995), Paxton (2000) and Paxton et al.
(2008). Specimens were not genotyped. We measured reflectance using a JAZ spectrometer (Ocean Optics, Dunedin, FL)
with a fiber optic reflectance probe and xenon pulsed light
source. To minimize detection of ambient light, the probe
was housed in a sheath that held the probe at a 90° angle
and at a fixed distance of 5.9 mm from the feather surface.
Specimens were placed flat on a low reflectance matte black
sheet of paper (e.g. construction paper) and we then took 10
replicate readings throughout the area of each plumage patch
(Reudink et al. 2009). Reflectance spectra were recorded
using SpectraSuite (ver. 1.0, Ocean Optics). We standardized
the reflectance measures between each patch by measuring a
white (Ocean Optics WS-1) and dark (sealed, black velvetlined box) standard. Prior to reflectance measurements, reference photos of each specimen were taken (Fig. 1B).
Song recordings

To assess geographic song variation of willow flycatcher subspecies, we used song recordings from across the US willow
flycatcher range (Fig. 2A, E. t. adastus n = 46; E. t. brewsteri n = 32; E. t. extimus n = 37; E. t. traillii n = 26). We
acquired publicly available songs from the Cornell Lab of
Ornithology Macaulay Library and xeno-canto.org libraries (Supporting information). We restricted our recordings
to the ‘fitz-bew’ vocalization, as that is used to attract mates

Figure 1. (A) Specimen collection locations of willow flycatcher subspecies (Empidonax traillii adastus, red (n = 38), E. t. brewsteri, orange
(n = 30), E. t. extimus, blue (n = 21), E. t. traillii, green (n = 18)). Subspecies were assigned a priori based on collection location between 15
June and 25 July. Dots are scaled to sample size. Dashed lines indicate putative subspecies boundaries. (B) Locations of dorsal (left) and
ventral (right) plumage patches of museum specimens on which light reflectance (300–700 nm) was measured to assess subspecific
plumage differences.

and defend territories (Stein 1958, Prescott 1987). To avoid
recordings from non-breeding migrants, only songs that
were recorded during June and July and that included location metadata were included in analyses. We included n = 6
songs in late May (E. t. adastus n = 1, E. t. extimus n = 2, E. t.
traillii n = 3), but only when the recorder noted in the metadata that the individual bird was actively breeding (based
on observing either a mated pair and/or an active nest). We
also recorded singing male willow flycatchers on their breeding territories (Supporting information, locations referenced
from Paxton 2000 and Paxton et al. 2008) during June and
July 2016–2019 using a Sennheisser ME66 shotgun microphone (mono-line) with Rycote handgrip and Rycote Softie
windshield and Marantz PMD661 MKII solid-state recorder.
Our sampling rate (frequency samples per second) was set to
44.1 kHz.
Quantifying song characteristics

We digitized willow flycatcher songs using Raven Pro (Cornell
Lab of Ornithology) using the Hann window (size = 256
samples, 50% window overlap, DFT = 256 samples) with

unsmoothed view (Fig. 2B). We then quantified 44 acoustic
parameters of the ‘fitz-bew’ vocalization (Supporting information): 26 measures of frequency, 16 measures of duration
and two measures of frequency modulations (adapted from
Sedgwick 2001). To minimize background noise, all songs
were high- and low-pass filtered at 1 kHz and 7 kHz respectively. Only songs with high signal-noise ratio were used in
analyses and we did not include recordings that were missing
a parameter or in which the quality was too low to measure
a parameter.
Data analysis
Tetrahedral colorspace analysis

We analyzed raw spectral data using the pavo package
(Maia et al. 2013) for R (<www.r-project.org>). We corrected negative values by adding the minimum reflectance
value to all spectra (Maia et al. 2013). To assess plumage
variation among willow flycatcher subspecies and alder flycatchers, we modeled the reflectance data in ‘tetrahedral colorspace’ using avian visual models that correct for differences
in avian visual systems (Vorobeyev et al. 1998). Although

3

Figure 2. (A) Recording locations and sample sizes of willow flycatcher subspecies (Empidonax traillii adastus, red (n = 46), E. t. brewsteri,
orange (n = 32), E. t. extimus, blue (n = 37), E. t. traillii, green (n = 26)). Subspecies were assigned a priori based on recording location
between 15 June and 25 July. Dots are scaled to sample size. Dashed lines indicate putative subspecies boundaries. (B) Representative spectrograms of willow flycatcher subspecies song. Colored dots indicate subspecies from panel A. Representative songs were selected based on
songs that fell in the approximate center of the subspecies clustering in PCA space.

the visual sensitivity range of willow and alder flycatchers
is unknown, visual systems of other Tyrannidae is biased
towards violet wavelengths (Ödeen and Håstad 2003), so our
tetrahedral models estimated coordinates that correspond to
the average violet sensitive avian photoreceptor sensitivity to
violet (v), short (s), medium (m) and long (l) wavelengths.
We then summarized the tetrahedral data using a non-metric multidimensional scaling (NMDS) analysis. To test for
sexual dichromatism (Eaton 2007) in tetrahedral space and
the effects of specimen age, we used separate mixed effects
models. NMDS1 was our response variable and taxa, sex
(female: n = 41; male: n = 66), specimen age, and the interaction between taxa and sex were included as fixed effects.
Museum was included as a random effect. We treated plumage patches as distinct body regions and therefore separate
hypotheses so we did not adjust our alpha level for multiple
comparisons (Montgomerie 2006). We excluded specimens
from this analysis where sex was unknown (n = 8).
Song structure analysis

Because many of our acoustic parameters were collinear, we
summarized song characteristics with a principal components

4

analysis (PCA), using the prcomp command in R (<www.rproject.org>). Prior to PCA we tested raw data for homoscedasticity using a Levene’s test (F3,137 = 0.33, p = 0.80) and to
meet assumptions of linearity, we log transformed all acoustic
parameters. To help with interpretation of the loadings, we
multiplied −1 to PC scores (Vehrencamp et al. 2003).
Clustering models, and group assignment analyses

To assess whether taxa occupied different tetrahedral color
space or produced unique songs, we used an unbiased and
unsupervised classification approach that is independent of
taxa identification. First, we tested the null hypothesis that
there are no groupings of plumage color and song characteristics – and therefore the data fit best within one group cluster – using the fvis_nbclust function in the factoextra package
for R (Kassambara and Mundt 2017). This analysis assesses
the quality of group clusters (i.e. how well the data fit within
clusters) by calculating the silhouette width for n = 1–5
clusters for plumage color and n = 1–4 clusters for song.
The silhouette width is a measure of confidence for group
membership within a cluster and values range from −1 to +1
with values closer to 1 represent better clustering (Rousseeuw

1987). For all plumage patches and song, groupings were
optimized in n > 1 clusters (Supporting information), suggesting some clustering of plumage color and song characteristics. Next, we determined the appropriate number of group
clusters of NMDS color space scores and song PCA scores
using ClValid (Brock et al. 2008) which evaluates clustering
models and the numbers of clustering groups independent of
taxa identity and subsequently identifies the appropriate clustering algorithm. In our analyses, we evaluated K-means and
hierarchical clustering models with n = 2–5 clustering groups
for plumage color and n = 2–4 for song. ClValid assesses
group clustering based on three indices: connectivity, Dunn
and silhouette width. The connectivity index assigns group
membership of data points based on the proximity to other
samples. Connectivity ranges from 0 to infinity and smaller
values represent well clustered data (Handl et al. 2005). The
Dunn and silhouette indices are measures of the ‘compactness’ and ‘spread’ of clusters. The Dunn metric is the ratio
between the smallest distance between data points from different clusters and the largest intracluster distance (Dunn
1974). Dunn indices range from 0 and infinity and higher
values represent better clustering. Silhouette values estimate
the degree of confidence in membership within a particular
cluster (Rousseeuw 1987). The silhouette indices are estimated by calculating the mean distance of points within a
cluster and the mean distance between points and range from
−1 to +1 and values close to 1 represent better clustering.
Therefore, we chose the number of groups and the clustering
method in our analyses based on models with optimized connectivity, Dunn and silhouette values (Brock et al. 2008). We
selected the best clustering algorithm based on a rank-based
optimization approach using the RankAggreg package in R
(Pihur et al. 2009). RankAggreg generates a rank-based list
corresponding to the performance of a given measure and
then provides an overall ranking from a Spearman’s footrule
analysis, which minimizes distance from all constituent lists
(Bible et al. 2013). Similar clustering analyses have been used
in previous studies with acoustic (Jeon and Hong 2015) and
morphological data (Lucek et al. 2016). Based on the clustering method results, we used either hierarchical clustering
using the ward.D2 function in the hclust package or K-means
clustering using the kmeans function in R to evaluate agreement between clustering group assignment and taxa identity.
To further assess plumage color variation among taxa, we
plotted individual specimens in tetrahedral colorspace which
models specimen color relationships and corrects for avian
visual systems.

Results
Plumage

For all plumage patches, our clustering selection analyses
identified two clustering groups (‘plumage color group 1 and
2’) and hierarchical clustering as the best algorithm (Table 1),
so our classifications are based on those models.

Although the clustering methods identified two groups
based on NMDS scores (Table 1, Supporting information)
for all plumage patches, we found low agreement between
taxonomic identity (based on geographic region where the
specimen was collected) and colorspace grouping (Fig. 3). On
all patches, taxa overlapped in tetrahedral colorspace (Fig. 4).
We found no evidence for sexual dichromatism on the
back (Supporting information, F1,93.3 = 3.5, p = 0.06), belly
(Supporting information, F1,91.2 = 0.07, p = 0.79), breast
(Supporting information, F1,93.0 = 1.66, p = 0.2), nape
(Supporting information, F1,92.5 = 0.05, p = 0.83), however on the crown, males averaged higher colorspace scores
than females (Fig. 5, Supporting information, F1,94.5 = 8.02,
p = 0.006). There was an interaction between taxa and sex on
the throat (Supporting information, F4,91.2 = 2.5, p = 0.04),
meaning the direction of dichromatism differed among taxa.
Table 1. Results from clustering algorithm selection for willow
Empidonax traillii and alder flycatcher Empidonax alnorum plumage
and willow flycatcher song. Clustering group number and algorithm
was selected based on NMDS tetrahedral colorspace scores and
song PCA scores using ClValid and AggRankreg which evaluates
clustering models and the numbers of clustering groups independent of taxa identification. In our analyses, we evaluated K-means
and hierarchical clustering models with n = 2–5 (plumage) and
n = 2–4 (song) clustering groups.

Back
Belly
Breast2
Crown
Nape2
Throat
Song

Index1

Score

Method2

Cluster

Connectivity
Dunn
Silhouette
Connectivity
Dunn
Silhouette
Connectivity
Dunn
Silhouette
Connectivity
Dunn
Silhouette
Connectivity
Dunn
Silhouette
Connectivity
Dunn
Silhouette
Connectivity
Dunn
Silhouette

4
0.18
0.43
4
0.18
0.43
5.8
0.15
0.6
6.13
0.15
0.62
4.2
0.19
0.58
2.9
0.4
0.8
2.92
0.65
0.59

Hierarchical
Hierarchical
Hierarchical
Hierarchical
Hierarchical
Hierarchical
Hierarchical
Hierarchical
Hierarchical
Hierarchical
Hierarchical
Hierarchical
Hierarchical
Hierarchical
Hierarchical
Hierarchical
Hierarchical
Hierarchical
Hierarchical
Hierarchical
Hierarchical

2
2
2
2
2
2
2
3
2
2
2
2
2
5
2
2
2
2
2
2
2

The connectivity index assigns group membership of data points
based on the proximity to other samples, where relatively smaller
connectivity metrics indicate well-clustered groups. The Dunn metric is an estimate of the intercluster distance relative to intracluster
distance of data points, where values > 1 indicate well-clustered
data. And the silhouette index is the mean distance of points within
a cluster and the mean distance between points, where silhouette
values > 1 indicates well-clustered data.
2
Clustering algorithm and cluster number was selected based on
overall rank-based selection process using the RankAggreg package
in R (Pihur et al. 2009). For the breast and nape, RankAggreg identified Hierarchical clustering with two groups (Spearman distances
breast: 6.07; nape: 7.50).
1

5

Figure 3. Proportion of plumage color group assignments for the back, belly, breast, crown, nape and throat (black, ‘plumage color group
1’, gray, ‘plumage color group 2’) of willow flycatcher subspecies (Empidonax traillii adastus, E. t. brewsteri, E. t. extimus, E. t. traillii) and
alder flycatcher E. alnorum museum specimens. Group assignments were based on unsupervised and unbiased classification of non-metric
multidimensional scaling (NMDS) of tetrahedral colorspace modeling that corrects for avian visual systems. Among all taxa, group assignment varied by individual.

Whereas E. t. adastus and E. alnorum males exhibited lower
average throat colorspace scores than females, E. t. brewsteri,
extimus and traillii males averaged higher throat colorspace
scores than females (Fig. 5). There was a significant effect
of specimen age on the breast (Supporting information,
F1,95.9 = 6.59, p = 0.01) and throat (Supporting information,
F1,54.6 = 14.6, p = 0.0003). Museum explained on average
relatively little variation for all patches (Supporting information, mean σ2 = 0.0009, range σ2 = 0.0001–0.004).
Song

From our principal components analysis, PC1 explained
38% of the variation. Positive PC1 scores were associated
with songs exhibiting higher overall minimum frequencies
and higher frequencies in Phrase 1 note 2 and Phrase 2 note
1 (Supporting information). Negative PC1 scores were associated with songs with lower overall minimum frequencies
and lower frequencies in Phrase 1 note 2 and Phrase 2 note 1
(Supporting information). PC2 explained 12% of the variation and positive scores were associated with songs containing more frequency modulations in the terminal portion of
the song and higher frequencies in Phrase 1 note 1. Negative
PC2 scores were associated with songs containing fewer terminal frequency modulations and lower frequencies in Phrase
1 note 1 (Supporting information).

6

Our song clustering analysis identified two groups (‘song
group 1 and 2’) and hierarchical clustering as the best algorithm (Table 1) so our classifications are based on that
model. When we compared group membership to taxonomic
identification based on geographic origin of each song, we
found song group 1 contained exclusively E. t. extimus songs
(n = 33, 89%; Fig. 6). Song group 2 contained high proportions of E. t. adastus (n = 46, 100%; Fig. 6), E. t. brewsteri
(n = 32, 100%; Fig. 6) and E. t. traillii (n = 26, 100%; Fig. 6)
and relatively low proportions of E. t. extimus (n = 4, 11%;
Fig. 6). Song groups separated in song PC space (Fig. 7).

Discussion
Using tetrahedral colorspace models, we found plumage color
did not differ among willow flycatcher subspecies or between
willow and alder flycatchers. Interestingly, we found evidence
for sexual dichromatism on the crown and throat but not
on the back or belly as was previously proposed by Eaton
(2007), suggesting that these plumage patches may be important in social interactions (Hunt et al. 1999, Griffith et al.
2003, Alonso-Alvarez et al. 2004, Limbourg et al. 2004). In
general, males averaged higher crown colorspace scores than
females, but the direction of throat dichromatism differed in
E. t adastus and E. alnorum relative to the other flycatcher

Figure 4. Plumage variation among willow flycatcher Empidonax traillii subpsecies (E. t. adastus, red (n = 38), E. t. brewsteri, orange
(n = 30), E. t. extimus, blue (n = 21), E. t. traillii, green (n = 18)) and alder flycatchers (E. alnorum, gray (n = 8)) as modeled in 3D tetrahedral colorspace relative to an achromatic center (white dot) for back, belly, breast, crown, nape and throat plumage patches. For all plumage
patches, taxa overlap in colorspace.

taxa. Our plumage results differ from previous research that
found color differences based on both qualitative (Phillips
1948, Aldrich 1951, Unitt 1987) and quantitative methods
(Paxton et al. 2010). In part these differences may have been
due to the fact that we used tetrahedral color models corrected for avian visual systems and included UV reflectance,
whereas previous studies focused on differences based on
human visual systems. One caveat of our plumage analysis
was that we used museum study skins rather than live birds,
and plumage spectra can differ between wild and museum
specimens (McNett and Marchetti 2005, Doucet and Hill
2009). Time of year that specimens were collected (Doucet
and Hill 2009), chemicals used to preserve skins (Pohland
and Mullen 2006), specimen age (Armenta et al. 2008,
Doucet and Hill 2009) and anthropogenic landscape conversions (Mason and Unitt 2018) have all been associated with
changes in reflectance in museum skins. After correcting for
avian visual systems, we found color differences among flycatcher taxa on the back, belly and throat. However, there
was a significant effect of specimen age on throat color, so
caution should be observed when interpreting those results.

Although a study of several species of wood warblers found
fading was minimal for specimens collected in the past 50
years (Armenta et al. 2008) only 14 of the 115 specimens we
examined were less than 50 years of age. Therefore, the potential exists that live birds may show differences among subspecies that we failed to detect in museum skins, although the
considerable overlap in reflectance values documented in the
one study that did examine reflectance from live flycatchers
(Paxton et al. 2010) suggests the unsupervised and unbiased
clustering algorithm approach we used would have failed to
detect strong groupings in data from live birds.
In contrast to plumage, song structure varied among flycatcher subspecies with E. t. extimus singing songs different
in structure relative to the other willow flycatcher subspecies.
Although the majority of putative E. t. extimus grouped into
a distinct song category, four (11%) of the E. t. extimus songs
used in our study grouped with the other subspecies. Of
these songs, three were from potential contact zones between
E. t. extimus and E. t. adastus and could have represented
birds of the other subspecies or admixed individuals (Kern
River, CA (n = 2) and Virgin River at St George, UT (n = 1);

7

Figure 5. Mean (± SE) NMDS1 scores representing crown colorspace (top panel) and throat colorspace (bottom panel) for female (f ) and
male (m) willow flycatcher Empidonax traillii subspecies (E. t. adastus, E. t. brewsteri, E. t. extimus, E. t. traillii) and alder flycatchers
Empidonax alnorum. There was a significant effect of sex on the crown (F1, 94.5 = 8.02, p = 0.006) and an interaction between sex and taxa
on the throat (F4, 91.2 = 2.5, p = 0.04). NMDS1 colorspace scores were calculated from tetrahedral color models which model plumage color
after correcting for avian visual systems. Sample sizes are indicated in bottom panel on x-axis. n = 8 specimens were excluded from this
analysis because sex was unknown.

Paxton et al. 2008, Theimer et al. 2016, M. J. Whitfield pers.
comm.). The remaining individual was recorded well within
the E. t. extimus range on the Rio Grande River near Elephant
Butte Reservoir, NM. Future genomic studies and playback
experiments will be important in determining the identity of
these individuals.
Song is proposed to be a reproductive isolating mechanism
in willow flycatchers (Stein 1963) and was previously found
to differ between E. t. adastus and E. t. extimus (Sedgwick
2001). Our results generally agree with Sedgwick (2001).
Although Sedgwick (2001) found differences in song length
between E. t. extimus and E. t. adastus (their Table 2), in our
study, song length was not an important variable in our PCA.
However our song group that contained only E. t. extimus
was distinguished by fewer frequency modulations in part 2
of phrase 3, overall lower frequencies of notes in phrases 1
and 2, and lower frequencies at maximum song amplitude,
consistent with the same key differences between E. t. extimus and E. t. adastus as described by Sedgwick (2001, their
Table 2, 3).
Song in tyrannids is innate rather than learned (Kroodsma
1984), so differences in song arise through genetic rather than
cultural evolution. Several non-exclusive hypotheses may

8

explain the pattern of song differentiation in willow flycatchers that we documented. Differences among populations
could arise through mutations that alter syrinx morphology,
neural circuitry or other physical aspects of song (Stein 1963,
Isler et al. 1997, Robbins and Stiles 1999, Sedgwick 2001,
Podos and Warren 2007). Once established those differences
could be maintained through either selection or reduced
gene flow that prevents genetic homogenization. Empidonax
species and subspecies often show strong habitat preferences
and generally have geographic ranges that are broadly allopatric (Sedgwick 2000, Johnson and Cicero 2002) and this
may reduce gene flow between populations and increase the
role of drift (Sedgwick 2001). Of the four willow flycatcher
subspecies, E. t. extimus arguably inhabits the most distinct
habitat, breeding in lower elevation riparian vegetation in the
arid southwestern United States. Selection pressure to optimize signal transmission in those hotter and drier habitats
may have shaped song structure (Morton 1975, Slabbekoorn
2004, Seddon 2005, Wilkins et al. 2013). In fact, willow
flycatchers song appears to be consistent with this hypothesis, as songs from populations in more arid regions (i.e.
E. t. extimus) are lower in frequency and exhibit fewer frequency modulations (Gonzalez et al. unpubl.). Alternatively,

Figure 6. Proportion of song group assignments (black, ‘song group
1’, gray, ‘song group 2’) for willow flycatcher subspecies (Empidonax
traillii adastus, E. t. brewsteri, E. t. extimus, E. t. traillii). Group
assignments were based on unsupervised and unbiased classification
of song principal components derived from a principal components
analysis. Song group 1 was comprised entirely of E. t. extimus (89%,
n = 33). All E. t. adastus (n = 46), E. t. brewsteri (n = 32), E. t. traillii
(n = 26) and 11% of E. t. extimus (n = 4) songs grouped into song
group 2.

selectively neutral changes in song could be more likely to
be established through founder effects and maintained by
isolation in southwestern population due to the rarity of
appropriate riparian habitat within the broader arid landscape. Whether individuals recognize the vocal differences we
documented remains to be tested but such data would be

an important step in understanding population divergence in
willow flycatchers.
Finally, our results have important implications for taxonomy and conservation of the southwestern populations of the
willow flycatcher. In a recent commentary, Zink (2015) called
into question the validity of the subspecies designation of the
southwestern willow flycatcher, E. t. extimus, and therefore its
protection under the Endangered Species Act. Several concerns
voiced in that paper have been addressed here. Previous studies of plumage either lacked rigorous statistical analyses, used
data from live specimens and thus lacked reference specimens
for subsequent reanalysis and/or assumed subspecies identity
a priori based on presumed subspecies distributions. Our
plumage analysis was based on catalogued museum specimens
(admittedly with the caveats of using museum specimens listed
above) and utilized unbiased classification analyses that did not
assume subspecies identification. Likewise, our song analysis
was based on publicly available songs from all four subspecies,
and our analyses grouped songs independent of any a priori
assumptions about subspecies identity. Zink (2015) also proposed adopting Amadon’s (1949) ‘75% rule’ to justify considering a subset of a species as a distinct subspecies. This rule
requires that ‘75% of a population effectively must lie outside
99% of the range of other populations for a given defining
character or set of characters’ (Patten and Unitt 2002, p. 27).
Zink (2015) went further and recommended characters ‘be
nearly (95%) if not completely diagnosable (Cracraft et al.
1998).’ Our unbiased classification analyses of song structure
separated songs into two groups, with 89% of putative E. t.
extimus falling into song group 1 and 100% of three subspecies
falling into song group 2, thereby exceeding the ‘75% rule’

Figure 7. Willow flycatcher song PC1 (38% of variation) and PC2 (12% of variation). Symbols represent putative subspecies identification,
based on location of field recordings during the breeding season. Colors represent song group assignment (black, ‘song group 1’, gray, ‘song
group 2’), based on unsupervised and unbiased classification of song principal components derived from a principal components analysis.
Positive PC1 scores represent higher frequency songs. Positive PC2 scores represent more frequency modulations in Phrase 3.

9

and approximating the more rigorous recommendation of
Zink (2015). Importantly, this diagnosis was based on song,
a trait essential in diagnosing species limits in Tyrannidae
(Rheindt et al. 2008, Tobias et al. 2010). Thus our song data
support recognition of the southwestern population as a distinct subspecies.
Acknowledgments – We thank the following individuals for
providing access to museum collections: Andy Johnson (Univ. of
New Mexico), Phil Unitt (San Diego Natural History Museum),
Jessika Vasquez (San Bernardino County Museum), Carla Cicero
(Museum of Vertebrate Biology, Univ. of California, Berkeley),
Peter Konstantindis (Oregon State Univ.), Robert Faucett (Burke
Museum), Libby Beckman (Phillip L. Wright Zoologica Museum,
Univ. of Montana), Jeff Stephenson (Denver Museum of Nature
and Science), Elizabeth Wommack (Univ. of Wyoming Museum of
Vertebrates), Eric Rickart (Natural History Museum of Utah). We
also thank the Macaulay Library (Cornell Univ.) and xeno-canto.
org libraries for generously sharing song recordings. Comments
from the Theimer lab group (Northern Arizona Univ.) and C. E.
Aslan and S. M. Shuster greatly improved this manuscript. S. I.
Gonzalez, R. W. Winton, A. N. B. Smith and D. N. Rakestraw
assisted with gathering field recordings. Finally, comments from
three anonymous reviewers greatly improved this paper.
Funding – This work was supported by the T&E Inc. Conservation
Grant (SMM), NAU John Prather Conservation Award (SMM) and
the NAU Landscape Conservation Initiative Fellowship (SMM).
Conflicts of interest – The authors declare no conflict of interest.
Permits – No permits were required for this research.

Author contributions

Sean Mahoney: Conceptualization (equal); Data curation
(lead); Formal analysis (lead); Funding acquisition (lead);
Investigation (lead); Methodology (equal); Resources (equal);
Writing – original draft (lead); Writing – review and editing
(equal). Matthew Reudink: Conceptualization (supporting); Formal analysis (supporting); Methodology (supporting); Resources (equal); Writing – review and editing (equal).
Bret Pasch: Conceptualization (equal); Formal analysis (supporting); Methodology (equal); Resources (equal); Writing
– original draft (supporting); Writing – review and editing
(equal). Tad Theimer: Conceptualization (equal); Formal
analysis (supporting); Methodology (equal); Resources
(equal); Writing – original draft (supporting); Writing –
review and editing (equal).

Transparent Peer Review
The peer review history for this article is available at https://
publons.com/publon/10.1111/jav.02621
Data availability statement

Spectrophotometry data, specimen photographs, song files
used in this study are provided by the authors in the Dryad
data repository.

10

References
Aldrich, J. W. 1951. A review of the races of the Traill’s flycatcher.
– Wilson Bull. 63: 192–197.
Alonso-Alvarez, C., Doutrelant, C. and Sorci, G. 2004. Ultraviolet
reflectance affects male–male interactions in the blue tit Parus
caeruleus ultramarinus. – Behav. Ecol. 15: 805–809.
Amadon, D. 1949. The seventy-five per cent rule for subspecies.
– Condor 51: 250–258.
Armenta, J. K., Dunn, P. O. and Whittingham, L. A. 2008. Quantifying avian sexual dichromatism: a comparison of methods.
– J. Exp. Biol. 211: 2423–2430.
Barrowclough, G. F., Cracraft, J., Klicka, J. and Zink, R. M. 2016.
How many kinds of birds are there and why does it matter?
– PLoS One 11: e0166307.
Bible, J., Datta, S. and Datta, S. 2013. Cluster analysis: finding
groups in data. – In: Informatics for materials science and engineering. Butterworth-Heinemann, pp. 53–70.
Bradbury, J. W. and Vehrencamp, S. L. 1998. Principles of animal
communication. – Sinauer Assoc.
Brewster, W. 1895. Notes on certain flycatchers of the genus Empidonax. – Auk 12: 157–163.
Brock, G., Pihur, V., Datta, S. and Datta, S. 2008. clValid, an R
package for cluster validation. – J. Stat. Softw. 25, <http://
www.jstatsoft.org/v25/i04>.
Browning, M. R. 1993. Comments on the taxonomy of Empidonax
traillii (willow flycatcher). – West. Birds 24: 241–257.
Cracraft, J., Feinstein, J. Vaughn, J. and Helm-Bychowski, K. 1998.
Sorting out tigers Panthera tigris: mitochondrial sequences,
nuclear inserts, systematics and conservation genetics. – Anim.
Conserv. 1: 139–150.
Cuthill, I. C., Partridge, J. C., Bennett, A. T., Church, S. C., Hart,
N. S. and Hunt, S. 2000. Ultraviolet vision in birds. – Adv.
Study Behav. 29: 159–214.
Doucet, S. M. and Hill, G. E. 2009. Do museum specimens accurately represent wild birds? A case study of carotenoid, melanin
and structural colours in long-tailed manakins Chiroxiphia linearis. – J. Avian Biol. 40: 146–156.
Dunn, C. 1974. Well separated clusters and fuzzy partitions. – J.
Cybernet. 4: 95–104.
Eaton, M. D. 2007. Avian visual perspective on plumage coloration
confirms rarity of sexually monochromatic North American
passerines. – Auk 124: 155–161.
Griffith, S. C., Ornborg, J., Russell, A. F., Andersson, S. and Sheldon, B. C. 2003. Correlations between ultraviolet coloration,
overwinter survival and offspring sex ratio in the blue tit. – J.
Evol. Biol. 16: 1045–1054.
Handl, J., Knowles, J. and Kell, D. B. 2005. Computational cluster validation in postgenomic data analysis. – Bioinformatics
21: 3201–12.
Haig, S. M., Beever, E. A., Chambers, S. M., Draheim, H. M.,
Dugger, B. D., Dunham, S., Elliot-Smith, E., Fontaine, J. B.,
Kesler, D. C., Knaus, B. J. and Lopes, I. F. 2006. Taxonomic
considerations in listing subspecies under the US Endangered
Species Act. – Conserv. Biol. 20: 1584–1594.
Hill, G. E. 2006. Environmental regulation of ornamental coloration. – In: Hill, G. E. and McGraw, K. J. (eds), Bird coloration,
Vol. 1. Mechanisms and measurements. Harvard Univ. Press,
pp. 507–560.
Hunt, S., Cuthill, I. C., Bennett, A. T. D. and Griffiths, R. 1999.
Preferences for ultraviolet partners in the blue tit. – Anim.
Behav. 58: 809–815.

Jeon, J. Y. and Hong, J. Y. 2015. Classification of urban park soundscapes through perceptions of the acoustical environments.
– Land. Urban Plan. 141: 100–111.
Johnson, N. K. and Cicero, C. 2002. The role of ecologic diversification in sibling speciation of Empidonax flycatchers
(Tyrannidae): multigene evidence from mtDNA. – Mol. Ecol.
11: 2065–2081.
Kassambara, A. and Mundt, F. 2017. Package ‘factoextra’. Extract
and visualize the results of multivariate data analyses. – <https://
rdrr.io/cran/factoextra>.
Kroodsma, D. E. 1984. Songs of the alder flycatcher Empidonax
alnorum and willow flycatcher Empidonax traillii are innate.
– Auk 101: 13–24.
Isler, M. L., Isler, P. R. and Whitney, B. M. 1997. Biogeography
and systematics of the Thamnophilus punctatus (Thamnophilidae) complex. – In: Remsen Jr, J. (ed.), Studies in neotropical
ornithology honoring ted parker. American Ornithologists’
Union, pp. 355–382.
Limbourg, T., Mateman, A. C., Andersson, S. and Lessells, C. M.
2004. Female blue tits adjust parental effort to manipulated
male UV attractiveness. – Proc. R. Soc. B 271: 1903–1908.
Lovette, I. J. and Bermingham, E. 1999. Explosive speciation in
the New World Dendroica warblers. – Proc. R. Soc. B 226:
1629–1636.
Lucek, K., Kristjánsson, B. K., Skúlason, S. and Seehausen, O.
2016. Ecosystem size matters: the dimensionality of intralacustrine diversification in Icelandic stickleback is predicted by lake
size. – Ecol. Evol. 6: 5256–5272.
Maia, R., Eliason, C. M., Bitton, P.-P., Doucet, S. M. and Shawkey,
M. D. 2013. Pavo: an R package for the analysis, visualization and
organization of spectral data. – Meth. Ecol. Evol. 4: 906–913.
Mallet, J. 2005. Hybridization as an invasion of the genome.
– Trends Ecol. Evol. 20: 229–237.
Martens, J., Eck, S., Päckert, M. and Sun, Y.-H. 2003. Methods of
systematic and taxonomic research on passerine birds: the
timely example of the Seicercus burkii complex (Sylviidae).
– Bonn. Zool. Beitr. 51:109–118.
Mason, N. A. and Unitt, P. 2018. Rapid phenotypic change in a
native bird population following conversion of the Colorado
Desert to agriculture. – J. Avian Biol. 49: jav-01507.
Mayr, E. 1942. Systematics and the origin of species. – Columbia
Univ. Press.
McNett, G. D. and Marchetti, K. 2005. Ultraviolet degradation in
carotenoid patches: live versus museum specimens of wood warblers (Parulidae). – Auk 122: 793–802.
Montgomerie, R. 2006. Analyzing colors. – In: Hill, G. E. and
McGraw, K. J. (eds), Bird coloration, Vol. 1. Mechanisms and
measurements. Harvard Uni. Press, pp. 90–147.
Morton, E. S. 1975. Ecological sources of selection on avian sounds.
– Am. Nat. 109: 17–34.
Oberholser, H. C. 1918. New light on the status of Empidonax
traillii (Audubon). – Ohio J. Sci. 18: 85–98.
Oberholser, H. C. 1932. Descriptions of new birds from Oregon,
chiefly from the Warner Valley region. Sci. Publ. Cleveland
Museum Nat. Hist. 4: 1–12.
Oberholser, H. C. 1947. A new flycatcher from the western United
States. – Proc. Biol. Soc. Washington 60: 77–78.
Ödeen, A. and Håstad, O. 2003. Complex distribution of avian
color vision systems revealed by sequencing the SWS1 opsin
from total DNA. – Mol. Biol. Evol. 20: 855–861.
Päckert, M., Martens, J., Kosuch, J., Nazarenko, A. A. and Veith,
M. 2003. Phylogenetic signal in the song of crests and kinglets
(Aves: Regulus). – Evolution 57: 616–629.

Patten, M. A. and Unitt, P. 2002. Diagnosability versus mean differences of sage sparrow subspecies. – Auk 119: 26–35.
Paxton, E. H. 2000. Molecular genetic structuring and demographic history of the willow flycatcher Empidonax traillii.
– MSc thesis, Northern Arizona Univ., Flagstaff, AZ, USA.
Paxton, E. H., Sogge, M. K., Theimer, T. C., Girard, J. and Keim,
P. 2008. Using molecular markers to resolve a subspecies boundary: the northern boundary of the southwestern willow flycatcher in the Four Corner states. – U.S. Geological Survey
Open File Report 2008-1117.
Paxton, E. H., Sogge, M. K., Koronkiewicz, T. J., McCleod, M. A.
and Theimer, T. C. 2010. Geographic variation in the plumage
coloration of willow flycatchers Empidonax traillii. – J. Avian
Biol. 41: 128–138.
Phillips, A. R. 1948. Variation in Empidonax traillii. – Auk 65:
507–514.
Pihur, V., Datta, S. and Datta, S. 2009. RankAggreg, an R package
for weighted rank aggregation. – BMC Bioinform. 10: 62.
Podos, J. and Warren, P. S. 2007. The evolution of geographic
variation in birdsong. – Adv. Study Behav. 37: 403–458.
Pohland, G. and Mullen, P. 2006. Preservation agents influence
UV-coloration of plumage in museum bird skins. – J. Ornithol.
147: 464–467.
Prescott, D. R. 1987. Territorial responses to song playback in allopatric and sympatric populations of alder Empidonax alnorum and
willow E. traillii flycatchers. – Wilson Bull. 99: 611–619.
Prum, R. O. 2010. The Lande–Kirkpatrick mechanism is the null
model of evolution by intersexual selection: implications for
meaning, honesty and design in intersexual signals. – Evolution
64: 3085–3100.
Reudink, M. W., Marra, P. P., Boag, P. T. and Ratcliffe, L. M. 2009.
Plumage coloration predicts paternity and polygyny in the
American redstart. – Anim. Behav. 77: 495–501.
Rheindt, F. E., Norman, J. A., Christidis, L. 2008. DNA evidence
show vocalizations to be a better indicator of taxonomic limits
than plumage patterns in Zimmerius tyrant-flycatchers. – Mol.
Phylogenet. Evol. 48: 150–156.
Robbins, M. B. and Stiles, F. G. 1999. A new species of pygmy-owl
(Strigidae: Glaucidium) from the Pacific slope of the northern
Andes. – Auk 116: 305–315.
Rousseeuw, P. J. 1987. Silhouettes: a graphical aid to the interpretation and validation of cluster analysis. – J. Comp. Appl. Math.
20: 53–65.
Seddon, N. 2005. Ecological adaptation and species recognition
drives vocal evolution in neotropical suboscine birds. – Evolution 59: 200–215.
Sedgwick, J. A. 2000. Willow flycatcher Empidonax traillii. – In:
Poole, A. and Gill, F. B. (ed.), The Birds of North America,
Inc., Philadelphia, PA, USA.
Sedgwick, J. A. 2001. Geographic variation in the song of willow
flycatchers: differentiation between Empidonax traillii adastus
and E. t. extimus. – Auk 118: 366–379.
Shutler, D. and Weatherhead, P. J. 1990. Targets of sexual selection:
song and plumage of wood warblers. – Evolution 44: 1967–1977.
Slabbekoorn, H. 2004. Singing in the wild: the ecology of birdsong.
– In: Marler, P. and Slabbekoorn, H. (eds), Nature’s music: the
science of birdsong. Elsevier Academic Press, pp. 178–205.
Stein, R. C. 1958. The behavioral, ecological and morphological
characteristics of two populations of the alder flycatcher: Empidonax trailli (Audobon) (no. 371). – Univ. of the State of New
York, State Education Dept, NY, USA.
Stein, R. C. 1963. Isolating mechanisms between populations of
Traill’s flycatchers. – Proc. Am. Phil. Soc. 107: 21–31.

11

Theimer, T. C., Smith, A. D., Mahoney, S. M. and Ironside, K. E.
2016. Available data support protection of the southwestern
willow flycatcher under the Endangered Species Act. – Condor
Ornithol. Appl. 118: 289–299.
Tobias, J. A., Seddon, N., Spottiswoode, C. N., Pilgrim, J. D., Fishpool, L. D. and Collar, N. J. 2010. Quantitative criteria for species delimitation. – Ibis 152: 724–746.
Toews,
D.
P.,
Taylor,
S.
A.,
Vallender,
R.,
Brelsford, A., Butcher, B. G., Messer, P. W. and Lovette, I.
J. 2016. Plumage genes and little else distinguish
the genomes of hybridizing warblers. – Curr. Biol. 26:
2313–2318.
Unitt, P. 1987. Empidonax traillii extimus: an endangered subspecies. – West. Birds 18: 137–162.
USFWS (U.S. Fish and Wildlife Service) 1995. Final rule determining endangered status for the southwestern willow flycatcher.
– Federal Register 60: 10694–10715.

12

Vehrencamp, S. L., Ritter, A. F., Keever, M. and Bradbury, J. W.
2003. Responses to playback of local vs. distant contact calls in
the orange-fronted conure, Aratinga canicularis. – Ethology
109: 37–54.
Vorobeyev, M., Osorio, D., Bennett, A. T. D., Marshall, N. J.
and Cuthill, I. C. 1998. Tetrachromacy, oil droplets
and bird plumage colours. – J. Comp. Phys. A 183: 621–633.
Wilkins, M. R., Seddon, N. and Safran, R. J. 2013. Evolutionary
divergence in acoustic signals: causes and consequences.
– Trends Ecol. Evol. 28: 156–166.
Zink, R. M. 2015. Genetics, morphology and ecological niche
modeling do not support the subspecies status of the endangered southwestern willow flycatcher Empidonax traillii extimus.
– Condor Ornithol. Appl. 117: 76–86.
Zink, R. M. and Johnson, N. K. 1984. Evolutionary genetics of
flycatchers. I. Sibling species in the genera Empidonax and Contopus. – Syst. Biol. 33: 205–216.