Behavioural Processes 129 (2016) 86–93

Contents lists available at ScienceDirect

Behavioural Processes
journal homepage: www.elsevier.com/locate/behavproc

Low frequency dove coos vary across noise gradients in an urbanized

environment
Fengyi Guo a,b , Timothy C. Bonebrake a,b , Caroline Dingle b,∗
a
School of Biological Sciences, University of Hong Kong, Hong Kong
b
Department of Earth Sciences, University of Hong Kong, Hong Kong

a r t i c l e i n f o a b s t r a c t

Article history: Urbanization poses a challenge to bird communication due to signal masking by ambient noise and reflec-
Received 20 November 2015 tive surfaces that lead to signal degradation. Bird species (especially oscines) have been shown to alter
Received in revised form 23 May 2016 their singing behaviour to increase signal efficiency in highly urbanized environments. However, few
Accepted 2 June 2016
studies on the effects of noise on song structure have included birds with low frequency vocal signals
Available online 3 June 2016
which may be especially vulnerable to noise pollution due to significant frequency overlap of their signals
with traffic noise. We compared the perch coos of spotted doves (Streptopelia chinensis), a species with
Keywords:
very low frequency vocalizations, in different background noise levels across urban and peri-urban areas
Anthropogenic noise
Columbidae
in Hong Kong. We documented a 10% upward shift in the minimum frequency of coos of spotted doves
Coo across the noise gradient (a relatively small but significant shift), and a reduced maximum frequency in
Frequency urban habitats with a higher density of built up area. Hong Kong doves had significantly higher mini-
Signals mum and maximum frequencies than doves from throughout their range (from mostly rural sites). Our
Urbanization results indicate that urban species with extremely low sound frequencies such as doves can alter their
vocalizations in response to variable urban acoustic environments.
© 2016 Elsevier B.V. All rights reserved.

1. Introduction
Acoustic signals play crucial roles in bird communication and
convey important information used in territorial defence and mate
attraction (Catchpole and Slater, 2003; Collins, 2004). Anthro-
pogenic modifications to the acoustic environment, especially in
densely populated urbanized areas, have the potential to affect ani-
mal communication in various ways. Low-frequency urban noise
from traffic and industry can mask signals (Halfwerk et al., 2011;
Rabin and Greene, 2002; Slabbekoorn and Peet, 2003; Warren et al.,
2006) while sound reverberation from multiple reflective surfaces
might degrade bird signals due to echoes (Warren et al., 2006). In
dense urban habitats, sound attenuation by the concrete walls of
tall buildings might also limit sound transmission distances (Leader
et al., 2005; Parris and McCarthy, 2013; Warren et al., 2006).
A growing body of research has documented a variety of changes
in the signals of birds in noisy acoustic environments. One of the
most commonly documented responses to urban noise is for birds
to sing songs with higher frequencies, presumably to avoid the
∗ Corresponding author at: Department of Earth Sciences, James Lee Science Build-
ing, University of Hong Kong, Pokfulam Road, Hong Kong.
E-mail address: cdingle@hku.hk (C. Dingle).
masking effects of low frequency urban traffic noise (e.g. great tits
Parus major, Slabbekoorn and Peet, 2003; house finches Carpodacus
mexicanus, Bermúdez-Cuamatzin et al., 2011; silvereyes Zosterops
lateralis, Potvin et al., 2011). Other observed responses include
amplitude shifts (e.g. common nightingale Luscinia megarhynchos,
Brumm, 2004), changing song lengths (e.g. house finches Carpo-
dacus mexicanus Fernández-Juricic et al., 2005) and increasing the
amount of time spent singing (e.g. serins Serinus serinus Díaz et al.,
2011).
Species vocalizing at low sound frequencies are likely to be more
susceptible to the masking effects of noise pollution than those with
higher frequencies vocalizations (4–7 kHz) which may not suffer
from the masking effects of urban noise (Hu and Cardoso, 2009;
Rheindt, 2003; Slabbekoorn, 2013; Slabbekoorn and Ripmeester,
2008, but see Moiron et al., 2015). Species with intermediate mini-
mum signal frequencies (1.5–4 kHz) have been shown to be capable
of adjusting their signal frequencies to minimize the impact of
overlap with anthropogenic noise (Hu and Cardoso, 2010; Parris
and Schneider, 2009). Species vocalizing at very low frequencies
(below 1 kHz), however, will have a high proportion of their sig-
nals masked by low frequency background noise. For such species,
it could be difficult to shift the frequencies of their songs suffi-
ciently to overcome the masking effect of urban noise, and thus
these species might require the adoption of other non-frequency

http://dx.doi.org/10.1016/j.beproc.2016.06.002
0376-6357/© 2016 Elsevier B.V. All rights reserved.
 F. Guo et al. / Behavioural Processes 129 (2016) 86–93 87

Fig. 1. Map indicating the location of the 22 sampling sites used in this study. (10 of the original 32 sites were not included due to small sample sizes.) Black dots in dark
grey areas represent urban sites in dense built-up areas, and white dots represent peri-urban sites in open areas like shrublands and farmlands.

vocal adjustments (Francis et al., 2011; Hu and Cardoso, 2010). In
a model predicting the decrease in active space of avian signals in
noisy areas, Parris and McCarthy (2013) concluded that birds with
frequencies around 1.5 kHz should benefit most from frequency
shifts and thus should be more likely to switch to higher frequen-
cies in noisy urban areas. The Parris and McCarthy (2013) model
applies well to available data on oscine passerines, but the model
is so far untested with birds of frequencies lower than 1.5 kHz. Hu
and Cardoso (2010) found that only three out of nine bird species
with song or call frequencies lower than 1.5 kHz showed signif-
icant increases in minimum frequency in their vocalizations in
noisy urban areas. The species with the lowest frequencies (pied
currawong Strepera graculina: Fmin = 0.55 kHz, Australian magpie
Gymnorhina tibicen: Fmin = 0.61 kHz, and grey butcherbird Cracti-
cus torquatus: Fmin = 0.75 kHz) did not show any significant shift at
all. It therefore remains unclear how non-passerine species with
very low frequency signals respond to noise in urban areas and
what the consequences might be for their success in these novel
environments.
Doves (family Columbidae) are abundant residents of many
cities and seem to have successfully adapted to urban environ-
ments (Chace and Walsh, 2006). The dominant frequencies of urban
anthropogenic noise strongly overlap with the low frequencies of
dove coos which are used for both long-range (perch coo) and
short-range (bow coo) communication (de Kort and ten Cate, 2004;
Patricelli and Blickley, 2006; Rheindt, 2003). This signal masking
seems to contradict their apparent success in urban environments
(Slabbekoorn et al., 1999; Viney et al., 2005). Non-oscine species
like doves are traditionally thought to be incapable of significantly
changing the acoustic features of their signals through behavioural
modification (Patricelli and Blickley, 2006), and thus they might
suffer more than oscines from anthropogenic impacts such as
88 F. Guo et al. / Behavioural Processes 129 (2016) 86–93
urbanization (Ríos-Chelén et al., 2012). Some doves exhibit a strong
negative response to noise; mourning doves (Zenaida macroura), for
example, prefer nesting sites away from noise sources (Francis et al.,
2009, 2011). Recent large comparative studies also suggest that low
frequency vocalizing species tend to avoid noisy areas more than
high frequency species (Cardoso, 2014; Francis, 2015 but see also
Moiron et al., 2015).
To investigate how doves cope in urban environments despite
the very low frequency of their vocalizations, we recorded and ana-
lysed individual dove coos from urban and peri-urban sites across
a noise gradient in Hong Kong. We focused on the perch coo of
spotted doves (Streptopelia chinensis), which is likely used for both
mate attraction and territorial defence (de Kort and ten Cate, 2004;
Slabbekoorn and ten Cate, 1998), usually produced from a high
perch and thus considered a long-range signal (de Kort and ten Cate,
2004). We classified sampling sites based on two factors likely to
affect acoustic characteristics of dove coos, background noise level
and percent urban built up area (to represent the relative density
of reflective surfaces which could lead to reverberations; Dowling
et al., 2012; Warren et al., 2006). We predicted that doves would
increase the minimum frequency of their coos in response to noise
and urbanization. In addition, we compared dove coos from Hong
Kong with those from peri-urban or rural sites outside of Hong
Kong (using recordings available in online libraries) to see if a cor-
relation between noise level/habitat structure and coo frequency is
consistent throughout their range.
2. Materials and methods
2.1. Study site and organism
Hong Kong (22◦ 09 –22◦ 37 N, 113◦ 52 –114◦ 30 E) is a dense
metropolis with a population of more than 7 million people in a
total land area of 1104 km2 . Most urban development is constrained
to 20%–30% of its landmass, due to hilly topography (Lam et al.,
2005). The rest of the land area is protected mainly in the form of
country parks, covered by secondary forests, shrub land, and grass-
land (Jim, 1986; Zhou and Chu, 2012). 32 sample sites were selected
across Hong Kong and classified as urban or peri-urban based on the
density of urban built-up land as classified by the consensus global
land cover dataset of Tuanmu and Jetz (2014). The dataset repre-
sents the percentage of built-up area at the resolution of a 1 km by
1 km grid. Sites were classified as urban if the urban built-up area
was above 10% (Fig. 1). The GIS classification was largely consistent
with the onsite observation. We used ArcGIS version 10.1 for this
analysis.
The spotted dove is a native species to India and Southeast Asia
and has been introduced widely throughout the world (Rasmussen
and Anderton, 2005; Viney et al., 2005). As with other species in
the genus, it is commonly found in open woodland or near human
settlements such as farmlands and urban parks (MacKinnon et al.,
2000; Slabbekoorn et al., 1999; Viney et al., 2005). Although some-
times found at the edge of open secondary forests, they are scarcely
seen in deep forests (King et al., 1975; Yong et al., 2013). In Hong
Kong, spotted doves are an abundant resident, widely distributed in
habitats such as countryside, urban parks, and wooded areas (Viney
et al., 2005).
2.2. Field recordings and noise measurements
We recorded doves in the field from June 13th to August 6th,
2014. We aimed to record at least three individuals at each site dur-
ing this period. To avoid the confounding factor of changes in dove
activity levels at different times of the day, only one site was vis-
ited per day, between 7:00 am and 11:00 am, during which time the
spotted doves were the most vocal (Zhou et al., 2004). Recordings
were collected using a Marantz PMD-660 field recorder connected
to a Sennheiser ME66 short-shotgun microphone, with the micro-
phone pointing directly towards the individual dove at distances
of 3–15 m. Each dove was recorded for 2–3 min or until the dove
stopped calling. All individuals were recorded in natural conditions
and no disturbance was made to their behaviours. No recordings
were made in bad weather such as heavy rain, wind, or typhoon
conditions. Information on individual sex and body size could not
be obtained given the remote recordings.
Background noise levels were measured immediately after each
recording session was finished, using a SL-4001 Sound Level Meter
(slow response time, A-weighted, 20 ␮ Pa as a reference value) held
horizontally at breast height (approximately 1.1 m). First, the meter
was pointed in a random direction to get an average reading over
a five second period, then the sound level meter was rotated 90◦
to the right and another reading was obtained. This procedure was
repeated until four measurements from each of the four directions
(one set) were obtained (Minor and Urban, 2010). Immediately
after the first set of noise recordings were completed, the procedure
was repeated to get a second set of noise recordings. In cases when
an unusual sound occurred while measuring background noise lev-
els (e.g. a car occasionally driving by in a quiet place or when cicadas
began singing loudly), the first set of four recordings was completed
and the second set was not initiated until the intermittent noise
had ceased, allowing us to obtain a more representative reading of
average noise levels (Hu and Cardoso, 2010).
All eight measurements (two for each direction) were averaged
to produce a single noise level (AdB) for each location where a dove
was recorded. Since the decibel (dB) is a logarithmic unit, each of
the eight noise measurements was first converted to a linear scale
before taking the average (Jacobsen and Juhl 2013). Finally, the site
noise level (Noise) was determined as the median value of all the
AdB values from each recording location in that site, so that the
effects of extreme noisy points such as roadsides of urban parks
would be avoided (Lam et al., 2005).
2.3. Acoustic analysis
Recordings were digitized at a sampling rate of 44.1 kHz, con-
verted to spectrograms and analysed using Raven Pro 1.5 (www.
birds.cornell.edu/raven), using a Hann window with window size
of 2282, and a fast Fourier transformation (FFT) length of 4092,
resulting in a fine frequency resolution of 10.8 Hz (Charif et al.,
2010). Recordings with faint spectrograms or those greatly masked
by some external noise sources (e.g. sounds from other animals or
human conversation) were not included for the sake of accuracy.
Recordings that included more than one dove interacting were also
discarded as birds might modify coo characteristics when interact-
ing with each other (Brumm and Todt, 2004). In cases where one
individual switched its coo type during a single recording, one coo
type was randomly selected for the analysis.
We defined a coo as the shortest repeating unit in a record-
ing set (de Kort and ten Cate, 2004). Perch coos of spotted doves
are low in frequency (450–1000 Hz) and have narrow bandwidths
(Slabbekoorn et al., 1999). Spotted doves use multiple coo types
that vary in the number of notes (especially the terminal notes)
and overall length (Ali and Ripley, 1981; Rasmussen and Anderton,
2005). In Hong Kong, we recorded four coo types, two of which were
predominant perch coos and were the focus of acoustic analysis:
Type A (four notes) and Type B (three notes) (Fig. 2).
We measured three frequency features: minimum frequency,
maximum frequency and peak frequency. Both pitch perception
and modulation of formant frequencies function on a logrithmic
scale and therefore, using linear scale measurement can over-
estimate frequency shifts in high frequency signals, especially
 F. Guo et al. / Behavioural Processes 129 (2016) 86–93
Fig. 2. Spectrogram of the two dominant coo types of spotted dove perch coos; (a)
Type A coo with four notes, and (b) Type B coo with three notes (more common than
Type A).
in comparisons across large frequency ranges between species
(Cardoso, 2013). However, as our focus is on within-species com-
parison here, where the impact of that bias is expected to be small
(Cardoso, 2013), we measured the frequency variables linearly for
convienence. Peak frequency was determined automatically using
Raven as the frequency at which maximum power occurs (Charif
et al., 2010), while other acoustic data were extracted from spec-
trograms, using manual placement of the cursor. This method has
been criticized due to the potential for inaccuracies in measuring
absolute minimum frequencies especially when measuring songs
recorded in noise (Grace and Anderson, 2015; Zollinger et al., 2012).
However, transmission experiments have shown that the magni-
tude of this error is likely to be relatively small (Verzijden et al.,
2010). Throughout our measurement processes, the artefacts were
greatly reduced by dynamically adjusting the greyscale of spec-
trogram to fully visualize the entire frequency range (Cardoso and
Atwell, 2012). We directly measured the potential for measure-
ment error caused by an overlap between background noise and
the dove coos by calculating the signal-to-noise ratio (SNR) of each
coo measurement and correlating that with the difference between
peak and minimum frequency of each coo (an indication of mea-
surement error). As we found a significant correlation between SNR
and this difference, we included SNR in the statistical analysis as an
additional explanatory factor.
Dove vocalization consisted of bouts of consecutive coos,
repeated for up to several minutes. We aimed to analyse 10 con-
89
secutive coos of a single type (either 3 or 4 notes, Fig. 2) within a
single bout of vocalization for each individual. Occasionally a dove
would drop the first one or two short notes from an individual coo
within the bout. As these notes never had the minimum, maximum
or peak frequency for the coo, we included these coos in our anal-
ysis, measuring the remaining notes of those incomplete coos as
we did in normal cases, and the coo type was determined based
on other complete coos measured in the same recording. Coo mea-
surements were averaged first by individual, then by site to provide
an overall value for each site.
We first compared the frequency variables between the two coo
types (Type A/four notes and Type B/three notes) using Welch two
sample t-tests. To test whether the use of coo type varied by habitat
or by level of background noise, we then used a generalized linear
model (GLM) with binomial logistic regression with coo type as the
dependent variable and both habitat type (urban vs. peri-urban)
and background noise as the factors. We used a Welch two sam-
ple t-test to compare background noise levels between urban and
peri-urban sites. To test the effect of habitat type on the frequency
characteristics of dove coos, we used Welch two sample t-tests (for
minimum frequency) and Wilcoxon rank sum tests (for maximum
and peak frequency, as these were not normally distributed). We
constructed linear models to assess the potential impacts of habi-
tat type, noise levels, and SNR on each of the frequency variables.
We first ran models for each three acoustic variables using only
habitat and noise level as the explanatory factors. Then we re-ran
the models for minimum and peak frequency including SNR as an
additional factor as we found that SNR had an impact on our mea-
surements of minimum frequency (see results below). We ran this
model first for all coos combined, and then for each coo type sepa-
rately. All statistical analyses were conducted in R version 3.2.2 (R
Core Team, 2014).
2.4. Distribution-wide analysis
As spotted doves have a wide distribution, we were interested
to see if urban and rural coos also differed in sites outside of Hong
Kong. Twenty other spotted dove recordings (see SI Table 1 for the
distributions of 20 world dove sampling sites) were obtained from
Xeno-Canto (http://www.xeno-canto.org/) and compared to coos
recorded in Hong Kong. Only recordings of the same subspecies
(Streptopelia chinensis chinensis) with good quality sonograms were
chosen for the analysis, following the same criteria used to select
recordings from Hong Kong. We converted recordings to spectro-
grams and measured the coo features of world doves with Raven
using the same protocol as the Hong Kong doves. We applied the
same urban and peri-urban classification method with urban built-
up density data (Tuanmu and Jetz, 2014) to the sites of recordings
taken from Xeno-Canto. Most sites were classified as peri-urban (3
urban sites and 17 peri-urban sites). Therefore, we only included
the peri-urban sites (which would more accurately be classified
as “rural” because they had 0% built-up area) of world doves to
compare with those in Hong Kong.
Note that background noise levels could not be obtained for the
world recordings from Xeno-Canto. However, based on qualitative
examination of the “anthropogenic noise band” on spectrograms,
low frequency background noise generally appeared to be much
greater in Hong Kong sites (including peri-urban sites) than world
sites. This suggests, not surprisingly, that the Hong Kong sites are
noisier and more urbanized than the sites outside of Hong Kong
where the doves were recorded. The coo feature datasets of 120
birds from Hong Kong and the 17 “world rural doves” were com-
bined but were not averaged across sites in this analysis. The limited
sample size of world doves also made separate analyses on coo
types not possible and thus all coos were analysed together. We
used One-way ANOVA with Tukey contrasts to test the differences
90 F. Guo et al. / Behavioural Processes 129 (2016) 86–93

Table 1 Table 3
Comparison of frequency features between distinct coo types (Type A and Type B) Summary of linear model results examining the effects of habitat, noise and signal
by Welch two sample t-test. to noise ratio (SNR) on frequency features as well as minumum frequency in each
coo type separately.
Type A (n = 43) Type B (n = 77) t P
Response variable Factor DF Estimate t P
Minimum frequency (Hz) 542.2 ± 34.7 523.7 ± 35.8 2.77 0.007
Maximum frequency (Hz) 880.4 ± 48.7 861.9 ± 45.8 2.03 0.045 Maximum frequency Habitat 1 −38.57 −4.10 <0.001
Peak frequency (Hz) 749.8 ± 48.3 719.2 ± 55.1 3.16 0.002 Noise 1 0.75 1.15 0.26
Error 19
Minimum frequency Habitat 1 6.02 0.81 0.43
Table 2 Noise 1 1.54 3.02 0.007
Summary of noise level and spotted dove coo features in urban and peri-urban Hong Error 19
Kong compared by Welch two sample t-test (for noise and minimum frequency) or Peak frequency Habitat 1 −5.20 −0.41 0.69
Wilcoxon rank sum test (for maximum and peak frequency). Noise 1 0.39 0.44 0.66
Error 19
Peri-urban (n = 7) Urban (n = 15) t W P Minimum frequency Habitat 1 −13.33 −2.70 0.015
Noise (dB) 52.6 ± 7.7 54.4 ± 6.8 −0.51 – 0.62 Noise 1 1.27 4.55 <0.001
Minimum frequency (Hz) 524.6 ± 19.1 533.3 ± 19.1 −0.99 – 0.34 SNR 1 −3.75 −6.81 <0.001
Maximum frequency (Hz) 893.6 ± 25.8 856.3 ± 18.0 – 95 0.002 Error 18
Peak frequency (Hz) 734.5 ± 22.0 730.0 ± 28.9 – 65.5 0.38 Peak frequency Habitat 1 −14.80 −0.96 0.351
Noise 1 0.25 0.29 0.775
SNR 1 −1.86 −1.08 0.295
Error 18
in dove coo characteristics between world rural sites with Hong Minimum frequency (Type A) Habitat 1 −15.08 −1.22 0.24
Kong urban and peri-urban sites. All statistical analyses were done Noise 1 1.28 1.53 0.15
in R version 3.2.2 (R Core Team, 2014). SNR 1 −4.65 −2.85 0.01
Error 16
Minimum frequency (Type B) Habitat 1 −13.22 −2.29 0.034
3. Results Noise 1 1.44 4.28 <0.001
SNR 1 −2.93 −4.79 <0.001
Error 18
3.1. Hong Kong doves

Our analysis included recorded coos from 120 individual doves

from 22 sites in Hong Kong (Fig. 1). Ten sites with fewer than
three recorded individuals were not included in the analysis. We
analysed 7.4 ± 2.93 (Mean ± S.D.) coos on average per individual
dove. Among all 120 samples, Type A coos (n = 43) had significantly
higher frequencies (for all three frequency variables) than Type
B coos (n = 77) (Table 1). There was no evidence that doves use
the coo types differently in different habitats or noise levels—both
coo types were used equally in each habitat type and in quiet
and noisy sites (GLM: ZHabitat = 0.67, PHabitat = 0.51, ZNoise = −0.087,
PNoise = 0.93).
Unexpectedly, the amplitude of background noise did not dif-
fer significantly between urban and peri-urban sites (Table 2).
Background noise levels did vary across sampling sites, but the
variation could not be explained by percent built-up area in each
site, the measure we used to categorize our sampling sites. This
may reflect a high degree of variation in noise levels within each
site, or some other factors affecting noise level which we did not
measure. Despite the lack of significant noise differences between
urban and peri-urban sites, we found that maximum frequency
of dove coos was significantly higher in peri-urban sites than in
urban sites, but did not vary with noise (Tables 2 and 3; F = 8.58,
R2 = 0.42, P = 0.002). In contrast, minimum frequency increased sig-
nificantly with increasing levels of background noise, but did not
differ between urban and peri-urban sites (Tables 2 and 3; Fig. 3;
F = 5.25, R2 = 0.29, P = 0.015). Our test for measurement bias showed
that the difference between peak and minimum frequency was pos-
itively correlated with SNR (Linear Regression: F118 = 18.7, R2 = 0.13,
P < 0.001), and the SNR was negatively correlated with background
noise level (Linear Regression: F118 = 12.35, R2 = 0.09, P < 0.001).
This indicates that recording quality generally decreased with an
increase in background noise and that this may have resulted in
measurement error in the direction predicted by our hypothesis.
We therefore re-ran the analysis with SNR as a covariate in the
minimum frequency model and found that a significant portion
of the variability in minimum frequency across sites could still be
explained by differences in noise levels, with minimum frequency
increasing with increasing levels of background noise (Table 3;
F = 27.29, R2 = 0.79, P < 0.001). In this model, we also found a small
effect of habitat in that the minimum frequency of urban coos were
slightly lower than peri-urban coos (Table 3). Peak frequency did
not correlate with any of the explanatory variables (Tables 2 and 3;
without SNR: F = 0.16, R2 = 0, P = 0.85; with SNR: F = 0.50, R2 = 0,
P = 0.69).
When we analysed each coo type separately, the maximum
frequency of both coo types was higher in peri-urban habitats.
However, for Type A coos (which had higher minimum frequen-
cies than Type B coos), the minimum frequency did not correlate
with background noise level nor with habitat, but only with SNR
(Table 3; F = 3.15, R2 = 0.25, P = 0.05). For Type B coos, the minimum
frequency was significantly correlated with both urban habitat and
with background noise level (Table 3; F = 18.29, R2 = 0.71, P < 0.001).
3.2. Distribution-wide analysis
This analysis combined recordings from 120 doves from
Hong Kong and 17 doves from other parts of the world (SI
Table 1). Compared to doves in Hong Kong (from either urban
or peri-urban sites), doves from other parts of the range (from
primarily rural sites) had significantly lower minimum frequencies
(Fig. 4; ANOVA: F2 = 20.8, P < 0.001; Pairwise(World–HKPeri-urban) :
t = −4.65, P < 0.001; Pairwise(World–HKUrban) : t = −6.44, P < 0.001)
and maximum frequencies (Fig. 4; ANOVA: F2 = 9.56,
P < 0.001; Pairwise(World–HKPeri-urban) : t = −3.80, P < 0.001;
Pairwise(World–HKUrban) : t = −1.49, P = 0.29). We found no sig-
nificant difference for peak frequency across populations (Fig. 4;
ANOVA: F2 = 1.04, P = 0.36).
4. Discussion
We found that the minimum frequency of the coos of spotted
doves correlates with levels of background noise: coos of doves
in noisy areas had significantly higher minimum frequencies than
those living in quieter environments. This response is consistent
with a common (although not ubiquitous) trend of an upward fre-
quency shift in noisy habitats documented in other urban bird
species (reviewed in Brumm and Zollinger, 2013; Slabbekoorn and
 F. Guo et al. / Behavioural Processes 129 (2016) 86–93 91

575

Minimum Frequency (Hz)

550

525

500

50 60 70
Noise (dB)

Fig. 3. Significant linear relationship of minimum frequency with background noise across the 22 sampling sites, with ± 95% standard error bars.
Minimum Frequency (Hz)

mum frequency of Hong Kong dove coos were significantly higher

600 than those recorded in sites outside Hong Kong further supports
this hypothesis.
550
Spotted dove coos have a minimum frequency of about 500 Hz,
500 1 kHz below the frequency range modelled by Parris and McCarthy
450 (2013). Extrapolating from their model, we would predict that the
minimum frequency of spotted dove coos should increase by 1.23%,
400
or 6.15 Hz per dB of noise increase. After correcting for potential
HK Peri-urban HK Urban World Rural
measuring artefacts induced by low SNRs in noisy habitats, we
observed that spotted doves exhibited increased minimum fre-
quencies of only 1.27 Hz dB−1 (Table 3). The results of our study
Maximum Frequency (Hz)

could therefore help contribute to the general applicability of their

1000 model. Although doves may benefit greatly from shifting their coos
above the frequency range of traffic noise, they may be constrained
900 by their relatively simple syrinx morphology that is only capa-
ble of limited frequency variation (Beckers et al., 2003; Elemans
800 et al., 2008; Slabbekoorn and ten Cate, 1998). The relatively small
shift in frequencies observed could also be a consequence of lim-
700
HK Peri-urban HK Urban World Rural
ited coo plasticity and a lack of learning ability (Kroodsma, 2004;
Ríos-Chelén et al., 2012; Secondi et al., 2002; Slabbekoorn and ten
Cate, 1998). However, there is some evidence that even a small shift
in coo frequency may still be important for dove communication.
Playback experiments with collared doves (Streptopelia decaocto)
Peak Frequency (Hz)

850
revealed that doves can perceive the difference in coos modulated
800 by as little as 11% from their own coos (Slabbekoorn and ten Cate,
750 1997), and rock doves (Columba livia) can discriminate frequency
700 changes of 5–10 Hz when frequency is lower than 500 Hz (Sinnott
650 et al., 1980).
600 We found that the response to noise differed by coo type.
HK Peri-urban HK Urban World Rural Type B coos had higher minimum frequencies associated with
noise levels while no such shift could be found for Type A coos.
Fig. 4. Boxplot of coo features (top; minimum frequency, middle; maximum fre- This might be related to the different information conveyed in
quency, bottom; peak frequency) for spotted doves recorded in Hong Kong (urban
either type and the different function associated with coo type (Gil
and peri-urban) and recordings from throughout its natural distribution range (rural
sites). and Gahr, 2002), and thus not all types change consistently with
environmental properties (Potvin et al., 2011). However, as there
was no significant preference of either coo type in a given habi-
Ripmeester, 2008) and provides evidence that species with very tat or with background noise levels, we attribute the correlations
low frequency vocalizations may also be capable of adjusting their between frequency characteristics and environmental features to
signals in noisy urban conditions. The observation that the mini-
92 F. Guo et al. / Behavioural Processes 129 (2016) 86–93
direct modifications within coo types, rather than replacement of
different coo types (Cardoso and Atwell, 2011a).
The frequency increase observed in several avian species could
be a coupled result of the Lombard effect, in which animals increase
sound amplitude in response to background noise (Brumm and
Naguib, 2009; Brumm and Todt, 2002; Brumm and Zollinger,
2011; Cynx et al., 1998; Potash, 1972). Elemans et al. (2008) also
demonstrated that ring doves (Streptopelia risoria) cannot control
coo frequency and amplitude independently due to their sim-
ple syringeal morphology. However, one important prediction of
increasing frequency as by-product of Lombard effect is that the fre-
quency of the entire vocalization will shift (Brumm and Zollinger,
2011; Cardoso and Atwell, 2011b). Yet, in our results, neither
peak frequency nor maximum frequency of dove coos shifted with
increasing noise level. Therefore, instead of a by-product of singing
with increased amplitude, the increase in minimum frequency that
we observed is more likely to be an adaptive response to anthro-
pogenic low frequency noise (Cardoso and Atwell, 2011b).
The results from the urban and peri-urban habitats are more
difficult to interpret. In Hong Kong, maximum frequency was sig-
nificantly lower in urban sites than in peri-urban sites. Minimum
frequency was only related to habitat type in the model that
included SNR as a variable, and minimum frequency was lower
in urban habitats. Since noise levels did not differ significantly
between urban and peri-urban sites, these frequency changes are
likely related to other habitat features, such as multiple scatter-
ing surfaces present in urban habitats which would affect higher
frequencies more than lower frequencies (e.g. through sound atten-
uation; Slabbekoorn et al., 2002). Reduction in coo frequency of
urban doves might therefore be a strategy of reducing frequency-
dependent attenuation and thus optimizing sound transmission
efficiency and broadcast range in a closed habitat, as predicted by
the acoustic adaptation hypothesis (Boncoraglio and Saino, 2007;
Brumm and Naguib, 2009; Dowling et al., 2012; Ey and Fischer,
2009; Leader et al., 2005; Morton, 1975; Wiley and Richards, 1982).
Our Hong Kong results differ from other studies, which have found
that maximum or peak frequencies are higher rather than lower
in urban birds (Nemeth and Brumm, 2009). In many of these stud-
ies, urban sites were compared with relatively more dense forested
sites, such that the urban areas represented the more open habi-
tats (Nemeth and Brumm, 2009; Slabbekoorn and den Boer-Visser,
2006). In our study, the urban sites were the least open of the
two habitat types. Based on the results from the study within
Hong Kong, we might have predicted that dove coos from the
rural areas outside of Hong Kong would have the highest maxi-
mum frequencies. Non-Hong Kong doves were recorded in sites
with 0% urban built-up density and thus were likely to be more
open than both urban and peri-urban Hong Kong sites. However,
our results revealed the opposite pattern: the frequency of spot-
ted dove coos from other parts of their range were lower than in
Hong Kong (especially peri-urban) doves. It is likely that there are
other factors influencing song structure that our method of catego-
rizing sites as urban and peri-urban, based on percentage of built
up area, did not capture. To further explore the potential effect of
frequency-dependent attenuation on urban songs, more detailed
measurements of habitat structure should be included in future
studies (Patten et al., 2004).
In our study we could not control for possible confounding vari-
ables between urban and peri-urban habitats and thus cannot rule
out that factors other than noise or habitat density may have an
impact on the dove coos (Francis et al., 2011; Slabbekoorn, 2013).
A few recent studies have been conducted in single habitats where
the only factor that varied between sites was noise level, due to
noise from compressor stations and well pads (Bayne et al., 2008;
Francis et al., 2009) or due to the addition of artificial noise or “phan-
tom roads” (McClure et al., 2013). While these allow for the isolation
of the effects of single factors on song, thus providing strong evi-
dence for the effect of noise on song structure, they may offer
limited insights into the multiple factors that may affect signals
in dense and noisy urban habitats.
Our study provides evidence that low frequency non-passerine
species vocalize differently in areas with varying levels of back-
ground noises and urban built-up densities. Spotted doves shifted
their frequencies as predicted if they were avoiding signal masking
by low frequency background noise common in highly urbanized
areas. We demonstrate that the vocal signals of extremely low fre-
quency vocalizing species do differ in noisy and quiet areas, a novel
result in line with the results from studies on higher frequency
birds. To determine whether these signal shifts are adaptive, future
studies will require direct investigation of the sound transmission
efficiency as well as reproductive success of doves with different
coo features in the same habitats (as with great tits in Halfwerk
et al., 2011). In combination with the current study, this could give
new insights into the evolution of animals in the novel habitats of
the Anthropocene.
Acknowledgements
We would like to thank Biyu Du, Jiayue Zhang, Yixian Zhou, and
the many undergraduates at HKU who helped with the collection
of data in the field. Thanks also go to Tsang Pak Nok Toby and
anonymous reviewers for helpful comments on early versions of
this manuscript. This study was funded by the Summer Research
Fellowship (SRF) Scheme of the Faculty of Science, The University
of Hong Kong.
Appendix A. Supplementary data
Supplementary data associated with this article can be found,
in the online version, at http://dx.doi.org/10.1016/j.beproc.2016.
06.002.
References
Ali, S., Ripley, S.D., 1981. Handbook of the Birds of India and Pakistan: Together
with Those of Bangladesh, Nepal, Bhutan and Sri Lanka Volume 2: Megapodes
to Crab Plover. Oxford University Press, Delhi, India.
Bayne, E.M., Habib, L., Boutin, S., 2008. Impacts of chronic anthropogenic noise
from energy-sector activity on abundance of songbirds in the boreal forest.
Conserv. Biol. 22, 1186–1193.
Beckers, G.J., Suthers, R.A., ten Cate, C., 2003. Mechanisms of frequency and
amplitude modulation in ring dove song. J. Exp. Biol. 206, 1833–1843.
Bermúdez-Cuamatzin, E., Ríos-Chelén, A.A., Gil, D., Garcia, C.M., 2011.
Experimental evidence for real-time song frequency shift in response to urban
noise in a passerine bird. Biol. Lett. 7, 36–38.
Boncoraglio, G., Saino, N., 2007. Habitat structure and the evolution of bird song: a
meta-analysis of the evidence for the acoustic adaptation hypothesis. Funct.
Ecol. 21, 134–142.
Brumm, H., Naguib, M., 2009. Environmental acoustics and the evolution of bird
song. Adv. Study Behav. 40, 1–33.
Brumm, H., Todt, D., 2002. Noise-dependent song amplitude regulation in a
territorial songbird. Anim. Behav. 63, 891–897.
Brumm, H., Todt, D., 2004. Male–male vocal interactions and the adjustment of
song amplitude in a territorial bird. Anim. Behav. 67, 281–286.
Brumm, H., Zollinger, S.A., 2011. The evolution of the Lombard effect: 100 years of
psychoacoustic research. Behaviour 148, 11–13.
Brumm, H., Zollinger, S.A., 2013. Avian vocal production in noise. In: Brumm, H.
(Ed.), Animal Communication and Noise. Springer, Berlin, Heidelberg, pp.
187–227.
Brumm, H., 2004. The impact of environmental noise on song amplitude in a
territorial bird. J. Anim. Ecol. 73, 434–440.
Cardoso, G.C., Atwell, J.W., 2011a. Directional cultural change by modification and
replacement of memes. Evolution 65, 295–300.
Cardoso, G.C., Atwell, J.W., 2011b. On the relation between loudness and the
increased song frequency of urban birds. Anim. Behav. 82, 831–836.
Cardoso, G.C., Atwell, J.W., 2012. On amplitude and frequency in birdsong: a reply
to Zollinger et al. Anim. Behav. 84, e10–e15.
Cardoso, G.C., 2013. Using frequency ratios to study vocal communication. Anim.
Behav. 85, 1529–1532.
 F. Guo et al. / Behavioural Processes 129 (2016) 86–93
Cardoso, G.C., 2014. Nesting and acoustic ecology, but not phylogeny: influence
passerine urban tolerance. Glob. Change Biol. 20, 803–810.
Catchpole, C.K., Slater, P.J., 2003. Bird Song: Biological Themes and Variations.
Cambridge University Press, Cambridge, U.K.
Chace, J.F., Walsh, J.J., 2006. Urban effects on native avifauna: a review. Landscape
Urban Plann. 74, 46–69.
Charif, R.A., Waack, A.M., Strickman, L.M., 2010. Raven Pro 1.4 User’s Manual.
Cornell Lab of Ornithology, Ithaca, NY.
Collins, S., 2004. Vocal fighting and flirting: the functions of birdsong. In: Marler, P.,
Slabbekoorn, H. (Eds.), Nature’s Music: The Science of Birdsong. Elsevier
Academic, Boston, U.S.A, pp. 39–79.
Cynx, J., Lewis, R., Tavel, B., Tse, H., 1998. Amplitude regulation of vocalizations in
noise by a songbird, Taeniopygia guttata. Anim. Behav. 56, 107–113.
de Kort, S.R., ten Cate, C., 2004. Repeated decrease in vocal repertoire size in
Streptopelia doves. Anim. Behav. 67, 549–557.
Díaz, M., Parra, A., Gallardo, C., 2011. Serins respond to anthropogenic noise by
increasing vocal activity. Behav. Ecol. 22, 332–336.
Dowling, J.L., Luther, D.A., Marra, P.P., 2012. Comparative effects of urban
development and anthropogenic noise on bird songs. Behav. Ecol. 23, 201–209.
Elemans, C.P., Zaccarelli, R., Herzel, H., 2008. Biomechanics and control of
vocalization in a non-songbird. J. R. Soc. Interface 5, 691–703.
Ey, E., Fischer, J., 2009. The acoustic adaptation hypothesis—a review of the
evidence from birds anurans and mammals. Bioacoustics 19, 21–48.
Fernández-Juricic, E., Poston, R., De Collibus, K., Morgan, T., Bastain, B., Martin, C.,
et al., 2005. Microhabitat selection and singing behavior patterns of male
house finches (Carpodacus mexicanus) in urban parks in a heavily urbanized
landscape in the Western US. Urban Habitats 3, 49–69.
Francis, C.D., Ortega, C.P., Cruz, A., 2009. Noise pollution changes avian
communities and species interactions. Curr. Biol. 19, 1415–1419.
Francis, C.D., Ortega, C.P., Cruz, A., 2011. Noise pollution filters bird communities
based on vocal frequency. PLoS One 6, 1–8.
Francis, C.D., 2015. Vocal traits and diet explain avian sensitivities to
anthropogenic noise. Glob. Change Biol. 21, 1809–1820.
Gil, D., Gahr, M., 2002. The honesty of bird song: multiple constraints for multiple
traits. Trends Ecol. Evol. 17, 133–141.
Grace, M.K., Anderson, R.C., 2015. No frequency shift in the D notes of Carolina
chickadee calls in response to traffic noise. Behav. Ecol. Sociobiol. 69, 253–263.
Halfwerk, W., Bot, S., Buikx, J., van der Velde, M., Komdeur, J., ten Cate, C., et al.,
2011. Low-frequency songs lose their potency in noisy urban conditions. Proc.
Natl. Acad. Sci. U. S. A. 108, 14549–14554.
Hu, Y., Cardoso, G.C., 2009. Are bird species that vocalize at higher frequencies
preadapted to inhabit noisy urban areas? Behav. Ecol. 20, 1268–1273.
Hu, Y., Cardoso, G.C., 2010. Which birds adjust the frequency of vocalizations in
urban noise? Anim. Behav. 79, 863–867.
Jacobsen, F., Juhl, P.M., 2013. Fundamentals of General Linear Acoustics. John Wiley
& Sons, Chichester, U.K, pp. 29–40.
Jim, C., 1986. The country parks programme and countryside conservation in Hong
Kong. Environmentalist 6, 259–270.
King, B.F., Woodcock, M., Dickinson, E.C., 1975. A Field Guide to the Birds of
South-east Asia: Covering Burma, Malaya, Thailand, Cambodia, Vietnam, Laos,
and Hong Kong. Collins, London, U.K.
Kroodsma, D., 2004. The diversity and plasticity of birdsong. In: Marler, P.,
Slabbekoorn, H. (Eds.), Nature’s Music: The Science of Birdsong. Elsevier
Academic Press, Boston, U.S.A, pp. 108–131.
Lam, K.C., Ng, S.L., Hui, W.C., Chan, P.K., 2005. Environmental quality of urban parks
and open spaces in Hong Kong. Environ. Monit. Assess. 111, 55–73.
Leader, N., Wright, J., Yom-Yov, Y., Nelson, D., 2005. Acoustic properties of two
urban song dialects in the orange-tufted sunbird (Nectarinia osea). Auk 122,
231–245.
MacKinnon, J., Phillipps, K., He, F.Q., Showler, D., MacKinnon, D., 2000. A Field
Guide to the Birds of China. Oxford University Press, Oxford, U.K.
McClure, C.J., Ware, H.E., Carlisle, J., Kaltenecker, G., Barber, J.R., 2013. An
experimental investigation into the effects of traffic noise on distributions of
birds: avoiding the phantom road. Proc. R. Soc. Lond. B: Biol. Sci. 280,
20132290.
Minor, E., Urban, D., 2010. Forest bird communities across a gradient of urban
development. Urban Ecosyst. 13, 51–71.
Moiron, M., González-Lagos, C., Slabbekoorn, H., Sol, D., 2015. Singing in the city:
high song frequencies are no guarantee for urban success in birds. Behav. Ecol.
26, 843–850.
Morton, E.S., 1975. Ecological sources of selection on avian sounds. Am. Nat. 109,
17–34.
Nemeth, E., Brumm, H., 2009. Blackbirds sing higher pitched songs in cities:
adaptation to habitat acoustics or side-effect of urbanization? Anim. Behav. 78,
637–641.
93
Parris, K.M., McCarthy, M.A., 2013. Predicting the effect of urban noise on the
active space of avian vocal signals. Am. Nat. 182, 452–464.
Parris, K.M., Schneider, A., 2009. Impacts of traffic noise and traffic volume on birds
of roadside habitats. Ecol. Soc. 14, 29.
Patricelli, G.L., Blickley, J.L., 2006. Avian communication in urban noise: causes and
consequences of vocal adjustment. Auk 123, 639–649.
Patten, M.A., Rotenberry, J.T., Zuk, M., 2004. Habitat selection, acoustic adaptation:
and the evolution of reproductive isolation. Evolution 58, 2144–2155.
Potash, L.M., 1972. Noise-induced changes in calls of the Japanese quail. Psychon.
Sci. 26, 252–254.
Potvin, D.A., Parris, K.M., Mulder, R.A., 2011. Geographically pervasive effects of
urban noise on frequency and syllable rate of songs and calls in silvereyes
(Zosterops lateralis). Proc. R. Soc. Lond. B: Biol. Sci. 278, 2464–2469.
Core Team, R., 2014. R: A Language and Environment for Statistical Computing. R
Foundation for Statistical Computing, Vienna, Austria http://www.R-project.
org/.
Ríos-Chelén, A., Salaberria, C., Barbosa, I., Macías Garcia, C., Gil, D., 2012. The
learning advantage: bird species that learn their song show a tighter
adjustment of song to noisy environments than those that do not learn. J. Evol.
Biol. 25, 2171–2180.
Rabin, L.A., Greene, C.M., 2002. Changes to acoustic communication systems in
human-altered environments. J. Comp. Psychol. 116, 137–141.
Rasmussen, P.C., Anderton, J.C., 2005. Birds of South Asia: The Ripley Guide.
Smithsonian Institution, Washington, D.C.
Rheindt, F.E., 2003. The impact of roads on birds: does song frequency play a role in
determining susceptibility to noise pollution? J. Ornithol. 144, 295–306.
Secondi, J., de Bakker, M.A., ten Cate, C., 2002. Female responses to male coos in the
collared dove Streptopelia decaocto. Behaviour 139, 1287–1302.
Sinnott, J., Sachs, M., Hienz, R.D., 1980. Aspects of frequency discrimination in
passerine birds and pigeons. J. Comp. Physiol. Psychol. 94, 401–405.
Slabbekoorn, H., Peet, M., 2003. Birds sing at a higher pitch in urban noise. Nature
424, 267.
Slabbekoorn, H., Ripmeester, E.A., 2008. Birdsong and anthropogenic noise:
implications and applications for conservation. Mol. Ecol. 17, 72–83.
Slabbekoorn, H., den Boer-Visser, A., 2006. Cities change the songs of birds. Curr.
Biol. 16, 2326–2331.
Slabbekoorn, H., ten Cate, C., 1997. Stronger territorial responses to frequency
modulated coos in collared doves. Anim. Behav. 54, 955–965.
Slabbekoorn, H., ten Cate, C., 1998. Perceptual tuning to frequency characteristics
of territorial signals in collared doves. Anim. Behav. 56, 847–857.
Slabbekoorn, H., de Kort, S., ten Cate, C., 1999. Comparative analysis of perch-coo
vocalizations in Streptopelia doves. Auk 116, 737–748.
Slabbekoorn, H., Ellers, J., Smith, T.B., 2002. Birdsong and sound transmission: the
benefits of reverberations. Condor 104, 564–573.
Slabbekoorn, H., 2013. Songs of the city: noise-dependent spectral plasticity in the
acoustic phenotype of urban birds. Anim. Behav. 85, 1089–1099.
Tuanmu, M.N., Jetz, W., 2014. A global 1-km consensus land-cover product for
biodiversity and ecosystem modelling. Glob. Ecol. Biogeogr. 23, 1031–1045.
Verzijden, M.N., Ripmeester, E.A.P., Ohms, V.R., Snelderwaard, P., Slabbekoorn, H.,
2010. Immediate spectral flexibility in singing chiffchaffs during experimental
exposure to highway noise. J. Exp. Biol. 213, 2575–2581.
Viney, C., Phillipps, K., Lam, C.Y., 2005. The Birds of Hong Kong and South China.
Information Services Department, Hong Kong.
Warren, P.S., Katti, M., Ermann, M., Brazel, A., 2006. Urban bioacoustics: it’s not just
noise. Anim. Behav. 71, 491–502.
Wiley, R.H., Richards, D.G., 1982. Adaptations for acoustic communication in birds:
sound transmission and signal detection. In: Kroodsma, D.E., Miller, E.H.,
Ouellet, H. (Eds.), Acoustic Communication in Birds. Academic Press, New York,
pp. 131–181.
Yong, D.L., Lim, K.C., Lee, T.K., 2013. A Naturalist’s Guide to the Birds of Singapore.
John Beaufoy Publishing, Oxford, U.K.
Zhou, D., Chu, L., 2012. How would size, age, human disturbance, and vegetation
structure affect bird communities of urban parks in different seasons? J.
Ornithol. 153, 1101–1112.
Zhou, Y.B., Zhang, J.X., Li, H., Zhang, J., Hu, J.C., 2004. Behaviors and characteristics
of occupying territorial song of spotted dove (Streptopelia chinensis) during
breeding season. Zool. Res. 25, 153–157.
Zollinger, S.A., Podos, J., Nemeth, E., Goller, F., Brumm, H., 2012. On the relationship
between, and measurement of: amplitude and frequency in birdsong. Anim.
Behav. 84, e1–e9.