Content uploaded by Mar Sánchez
Author content
All content in this area was uploaded by Mar Sánchez on Mar 02, 2023
Content may be subject to copyright.
Vocal expression of emotional arousal across two call types in young
rhesus macaques
Jay W. Schwartz
a
,
b
,
*
, Mar M. Sanchez
c
,
d
, Harold Gouzoules
a
a
Department of Psychology, Emory University, Atlanta, GA, U.S.A.
b
Psychological Sciences Department, Western Oregon University, Monmouth, OR, U.S.A.
c
Emory National Primate Research Center, Emory University, Atlanta, GA, U.S.A.
d
Department of Psychiatry and Behavioral Sciences, Emory University, Atlanta, GA, U.S.A.
article info
Article history:
Received 12 September 2021
Initial acceptance 26 November 2021
Final acceptance 4 April 2022
MS. number: A21-00518R
Keywords:
acoustics
arousal
communication
emotion
expression
Macaca mulatta
primate
vocal
As Darwin first recognized, the study of emotional communication has the potential to improve scientific
understanding of the mechanisms of signal production as well as how signals evolve. We examined the
relationships between emotional arousal and selected acoustic characteristics of coo and scream vo-
calizations produced by female rhesus macaques, Macaca mulatta, during development. For coos, arousal
was assessed through measures of stress-induced elevations of plasma cortisol exhibited in response to
the human intruder test. In the analysis of screams, arousal was evaluated from the intensity of
aggression experienced by the vocalizer during natural social interactions. Both call types showed a
positive relationship between arousal and overall fundamental frequency (F0, perceived as pitch in
humans). In coos, this association was dampened over development from infancy (6 months) to the
juvenile, prepubertal period (16 months) and further to menarche (21.3e31.3 months), perhaps reflecting
developmental changes in physiology, anatomy and/or call function. Heightened arousal was also
associated in coos with increases in an acoustic dimension related to F0 modulation and noisiness. As
monkeys matured, coos showed decreases in overall F0 as well as increased noisiness and F0 modulation,
likely reflecting growth of the vocal apparatus and changes in vocal fold oscillation. Within screams, only
one acoustic dimension (related to F0 modulation) showed developmental change, and only within one
subclass of screams within one behavioural context. Our results regarding the acoustic correlates of
arousal in both call types are broadly consistent with findings in other species, supporting the hypothesis
of evolutionary continuity in emotion expression. We discuss implications for broader theories of how
vocal acoustics respond to selection pressures.
©2022 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
While there is broad consensus with Darwin's (1872) contention
that emotion plays an important role in animal vocal production
(Owren et al., 2011;Seyfarth &Cheney, 2003;Wheeler &Fischer,
2012), the precise manner in which emotion influences the voice
remains largely unstudied in many taxa (Briefer, 2012). Research
into the proximate mechanisms of vocal production, including the
role of emotion, can provide insights about phylogenetic history
(Zimmermann et al., 2013), as well as general processes in the
evolution of vocal communication, for example the role of con-
straints and correlated response to selection (Fitch &Hauser, 2006).
Thus, an overarching goal for this research is to understand vocal
emotion expression at both the proximate (mechanistic) and
ultimate (evolutionary) levels, and to integrate these levels of
analysis (see Mayr, 1961).
Central to the relationship between emotional states and vocal
production is arousal (Briefer, 2012). Arousal represents a spectrum
of internal states ranging from an inactive, calm or tranquil status
(low arousal) to a highly alert or excited condition (high arousal)
(Mendl et al., 2010). Emotional arousal comprises a suite of physi-
ological conditions and processes that generally (but imperfectly)
co-occur (see Scarantino &Griffiths, 2011) as part of a broader set of
internal states associated with reactions to stimuli perceived as
threatening, repulsive, attractive or exciting. In mammals, these
processes include heightened activation of the sympathetic ner-
vous system, resulting in increased heart rate, respiration, muscle
tension, and the activation of the hypothalamicepituitaryeadrenal
(HPA) axis, resulting in the production and release of stress hor-
mones such as cortisol (Ralph &Tilbrook, 2016;Romero, 2004).
*Corresponding author.
E-mail address: schwartzj@wou.edu (J. W. Schwartz).
Contents lists available at ScienceDirect
Animal Behaviour
journal homepage: www.elsevier.com/locate/anbehav
https://doi.org/10.1016/j.anbehav.2022.05.017
0003-3472/©2022 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
Animal Behaviour 190 (2022) 125e138
Arousal partly functions to potentiate behaviours such as reward-
seeking behaviour or approach (in positive contexts), or (in nega-
tive contexts) vigilance, aggression or fleeing (Mendl et al., 2010).
From a proximate perspective, the effects of emotion and arousal
on vocal production are thought to result largely from physiological
processes acting on the vocal apparatus (Briefer, 2012;Scherer, 1986;
Zei Pollermann &Archinard, 2002). The question of how arousal
manifests in vocal acoustics has been addressed in an array of
mammalian taxa (Briefer, 2012), but many species remain unstudied.
One of the best-established acoustic correlates of arousal is vocal
fundamental frequency (F0; determined by the rate of oscillation of
the vocal folds and commonly perceived as pitch by humans) with
greater arousal generally associated with increases in mean F0, as
well as F0 modulation (for example, the F0 range or slope) (Briefer,
2012). Arousal might achieve these effects on F0 by increasing ten-
sion and action in the diaphragm and/or laryngeal muscles, resulting
in faster vibration of the vocal folds and greater changes in vibration
frequency (Briefer, 2012;Titze, 1994). Effects of arousal on the
acoustic structure of vocalizations also include changes in vocal
duration, although duration may often be a stronger indicator of
emotional valence (the positive/pleasurable or negative/aversive
quality of emotions) (Briefer, 2012;Friel et al., 2019), as well as
increased noisiness (although see Blumstein &Chi, 2012) and up-
ward shifts in the distribution of energy across the frequency spec-
trum (Briefer, 2012).
A potentially complicating factor is that this literature includes
both studies comparing different call types and those looking at
acoustic variation within a particular call type. Mammals exhibit
acoustic gradation within and between call types (Fischer et al.,
2016;Price et al., 2015;Wadewitz et al., 2015), and the ways that
internal states affect vocal acoustics in each condition might differ
(Schamberg et al., 2018;Schwartz et al., 2020). Senders often use
call types in functionally distinct contexts (Schamberg et al., 2018)
and base calling on the composition or (in rare cases) knowledge
state of the audience (Crockford et al., 2012;Fichtel &Manser,
2010), suggesting between-call type variation depends not only
on emotion but also on other cognitive processes, including exec-
utive function. In contrast, the more fine-grained acoustic structure
of a vocalization (i.e. within-call type variation) might generally be
more strongly influenced by the effects of physiological compo-
nents of emotions (especially arousal) on the mechanisms of vocal
production (Schwartz et al., 2020).
From an ultimate perspective, the physiological mechanisms
of arousal are largely conserved across vertebrate taxa, as are
some aspects of vocal production, resulting in broad homology in
the ways that arousal manifests in vocal acoustics (Briefer, 2012;
Zimmermann et al., 2013). For example, the positive correlation
between arousal and vocal F0 seems to have been evolutionarily
conserved in mammals, and perhaps vertebrates more broadly
(Briefer, 2012;Filippi et al., 2017). The breadth of this homology
will become clear as more taxa and call types are studied. Thus,
one approach to gaining insight into the evolution of vocal
emotion expression is to assess the degree to which acoustic
correlates of arousal are consistent across species, both within
and across lineages, as well as among call types and contexts
within a species. Overall, although scientific understanding of the
vocal expression of emotion is growing, many taxa remain un-
studied, and few studies have compared multiple call types
within a species (for exceptions, see: Bastian &Schmidt, 2008;
Gogoleva et al., 2010;Linhart et al., 2015;Yeo n et al ., 2011),
resulting in rather limited knowledge of the degree of consis-
tency in vocal emotion expression. Further research into vocal
emotion expression across call types in additional taxa can
improve scientific understanding of vocal production at both the
proximate and ultimate levels.
Of additional interest at both the proximate and ultimate levels
is the question of whether and how patterns of vocal emotion
expression change over the course of development. The commu-
nicative functions of calls can change as an individual matures. For
example, separation calls, which are exhibited by infants and ju-
veniles across an array of mammalian taxa and function to help the
mother locate and reunite with the caller, have greater functional
significance for younger, more vulnerable animals who depend
more on their mother's presence, while that significance wanes as
individuals mature (Lingle et al., 2012). Theoretically these devel-
opmental changes could lead to selection favouring changes in the
acoustic structure of calls, perhaps including the acoustic correlates
of emotional states within calls at various life stages. On the other
hand, given the phylogenetically widespread nature of patterns of
vocal emotion expression, one might expect them to remain
consistent across development.
A primary aim of the present study was to compare the acoustic
correlates of arousal across two call types in rhesus macaques,
Macaca mulatta ecoos and screams ein order to contribute to-
wards ongoing efforts to test the hypothesis that there are broad
and general effects of emotional arousal on vocal production
(Briefer, 2012;Filippi et al., 2017;Zimmermann et al., 2013). Coos, a
tonal call typically with a rising, then falling F0 contour ranging
from 400 to 3000 Hz, function as a separation call as defined above
and are also produced in a variety of other contexts, including
anticipation of food and affiliative social interaction, and are
thought to function to maintain or stimulate friendly contact and/
or convey the location of the sender (Bayart et al., 1990;Hansen,
1976 ;Hauser &Marler, 1993). Macaque screams are relatively
high-F0 (3e10 kHz) vocalizations generally produced during
agonistic interactions, functioning to recruit aid, usually from kin;
rhesus screams fall into broad acoustic classes conveying infor-
mation about the intensity of aggression received by the vocalizer
and the relative rank of the opponent (H. Gouzoules, 2005;
Gouzoules &Gouzoules, 1995,2000;H. Gouzoules et al., 1985,
1998;S. Gouzoules et al., 1984). Although coos and screams are
acoustically disparate and occur in distinct contexts, both are
generally characterized by high, but variable, levels of arousal,
making them an interesting pair of call types in which to compare
and contrast the acoustic correlates of arousal.
In the present study, coos and screams were each recorded in
distinct and separate contexts. Screams were recorded during
natural social interactions; measures of the intensity of agonism
experienced by a vocalizer served as a proxy for arousal. Earlier
research has revealed acoustic variation among rhesus screams
relating to differing degrees of aggression (as well as opponent
rank), although this variation fell into broadly distinct scream
classes (S. Gouzoules et al., 1984). No previous studies have, to our
knowledge, examined the acoustic correlates of arousal in rhesus
screams while accounting for this between-scream class variation.
Coos were recorded during a behaviour test entailing maternal
separation (and not necessarily representative of other contexts in
which coos naturally occur, such as feeding) and were preceded
and followed by cortisol measurements to assess HPA activation.
HPA activation occurs as part of a broader arousal response, and
although HPA activation can reflect other processes unrelated to
emotions, it has been used to gauge arousal in studies of vocal
communication (e.g. Bayart et al., 1990;Blumstein &Chi, 2012;
Schrader &Todt,1998;S
ebe et al., 2012). Relationships between coo
acoustics and cortisol concentrations have been reported previ-
ously for rhesus macaques, although in a qualitative rather than
quantitative way (Bayart et al., 1990). In accord with the hypothesis
that the acoustic correlates of arousal are largely shared across taxa
(Briefer, 2012;Filippi et al., 2017;Zimmermann et al., 2013), we
predicted that coos and screams would each exhibit arousal-related
J. W. Schwartz et al. / Animal Behaviour 190 (2022) 125e138126
acoustic changes represented in the literature, including increases
in F0, F0 modulation, noisiness and peak frequency and upward
shifts in energy distribution.
In addition to the primary aim of testing hypotheses about the
overall association between arousal and vocal acoustics, a second-
ary aim was to explore whether and how the acoustic structure of
coos and screams, and patterns of emotion expression therein,
change as individuals mature. As discussed above, the importance
for survival of separation calls (e.g. coos) and of arousal during
separation is greatest in infancy and changes as individuals gain
independence (Lingle et al., 2012;Sanchez et al., 2015), while aid
recruitment and status maintenance functions of screams change
as individuals integrate into society and acquire their social rank (S.
Gouzoules et al., 1984;H. Gouzoules &Gouzoules, 1995), leading to
an interest in examining the developmental trajectories of these
two call types. Previous research has shown decreases in coo F0 and
F0 modulation and increases in coo duration in this species over the
first several weeks of development (Hammerschmidt et al., 2000).
In the closely related pigtail macaque, Macaca nemestrina, screams
undergo acoustic changes as individuals matured (H. Gouzoules &
Gouzoules, 1989). We predicted that we would observe similar
patterns over the first few years of subjects’lives, especially de-
creases in F0 due to growth of the body and larynx (e.g. Fitch &
Hauser, 2006;H. Gouzoules &Gouzoules, 1990). We also investi-
gated whether the acoustic correlates of arousal within each call
type might differ for monkeys at different stages of development.
METHODS
Subjects and Housing
This study was limited to female monkeys in order to avoid the
potentially confounding effects of sex, given logistical constraints
and the number of subjects available, and since the social life his-
tory patterns and behaviours of male and female macaques are very
different (Thierry, 2010). Subjects were followed from infancy (5e6
months of age) through menarche to examine and account for
developmental transitions in vocal acoustics and behaviour. Forty-
three female rhesus macaques served as the sample. All subjects
were concurrently involved in longitudinal studies to investigate
the synergistic effects of stress and diet on neurobehavioural
development of female rhesus macaques and were selected for
study at birth based on relatedness and dam social rank.
Subjects were housed in three outdoor enclosures, each
measuring 0.3 ha with an attached indoor area, at the Emory Na-
tional Primate Research Center (ENPRC) Field Station in Lawren-
ceville, Georgia, U.S.A. (Nper enclosure: 18, 13, 14). Each enclosure
housed a social group composed of one to two adult (i.e. >5 years of
age) males, 40e60 adult females with an established and docu-
mented linear matrilineal dominance hierarchy (based on data on
the direction and outcome of agonistic encounters; for details, see
Howell et al., 2014) and their immature offspring. The subject pool
included 21 high-ranking macaques (within the top 40% of in-
dividuals in their enclosure, based on observed agonistic in-
teractions ending in submissive behaviour by the subject) and 22
low-ranking macaques (within the bottom 40%). Subjects had ad
libitum access to water and low-fat, high-fibre pellets (Purina Mills
Int., Lab Diets, St Louis, MO, U.S.A.), and were provisioned with fresh
fruit and vegetables daily. As part of the aforementioned studies of
diet, 21 subjects (10 high-ranking) were provided additional ad
libitum access to high-calorie pellets (D14051502B, Research Diets,
Inc., New Brunswick, NJ, U.S.A.). Social rank and diet were not
directly relevant to hypotheses regarding the acoustic correlates of
arousal, which were the main focus of the present study, and were
thus excluded from analysis.
Ethical Note
The Emory University Institutional Animal Care and Use Com-
mittee approved all procedures in accordance with the Animal
Welfare Act and the U.S. Department of Health and Human Services’
Guide for Care and Use of Laboratory Animals. Human intruder
tests and blood draws were part of longitudinal studies at the
ENPRC funded by the U.S. National Institutes of Health (NIH) NIH/
NICHD R01 grant HD077623 (investigators: M. M. Sanchez and M. E.
Wilson) and took place independent of the present investigation of
vocal emotion expression.
Coo Recording, Preparation and Selection
Data collection for coos took place during a total of 111 human
intruder laboratory tests (HITs; Kalin &Shelton, 1989) conducted
from October 2014 through to December 2017, in which subjects
were first alone, then joined in the testing room by an unfamiliar
human. The HIT elicits HPA activation, resulting in increases in
plasma cortisol as well as cooing and other behaviours, yielding
behavioural profiles that vary among individuals depending in part
on temperament and other characteristics (Gottlieb &Capitanio,
2013;Hamel et al., 2017;Kalin, Larson, et al., 1998;Kalin, Shelton,
et al., 1998). Subjects underwent up to three tests, at 6 months of
age, at 16 months of age and at menarche; 4 of the 43 subjects were
removed from the project prior to 16 months of age, and an addi-
tional 10 subjects were removed prior to menarche, due to health-
related issues and/or the ending of the larger research project on
stress, obesity and diabetes. These subjects underwent only the first
one or two tests, respectively, leaving 29 subjects who completed
all three tests, although data from all 43 subjects were analysed.
Menarche was assessed using previously published methods
(Wilson et al., 2013) and occurred at ages ranging from 21.3 to 31.3
months.
At each age point, at sunrise (to control for circadian variation in
cortisol), and following training, the subject was prompted to enter
the indoor area for an unanaesthetized blood draw from the
saphenous vein, using procedures to which subjects were habitu-
ated and which have been described previously (Howell et al.,
2014;McCormack et al., 2009;Wilson et al., 2013). This blood
draw occurred within 10 min of the initial group disturbance (a
time frame within which effects of disturbance on circulating
cortisol are not yet detectable in papionin primates; Sanchez et al.,
2010;Sapolsky, 1982) and served as the baseline for analysis of
stress-induced cortisol increase at the end of the HIT (see below).
Pearson's correlation tests were used to test for a possible corre-
lation between the latency from the initial disturbance to the blood
draw at each age point; all were nonsignificant (P>0.05).
Following the blood draw, the subject was transported to a cage
within a 3.1 3.1 m testing room where the HIT was conducted.
Details of the HIT protocols used in the present study are
available in previous publications (Howell et al., 2014;Wilson et al.,
2013). Briefly, the HIT entailed three conditions each lasting 10 min:
Alone (subject by itself), Profile (human stranger sitting 3 m from
the cage containing the subject, at a 90
angle, presenting the face
in profile) and Stare (stranger facing the subject maintaining a
direct gaze at the subject's eyes). The testing room contained a
HoMedics SS-2000 noise generator (Commerce Township, MI,
U.S.A.) set to the ‘waterfall’setting (brown noise) at high volume
(approx. 80e85 dB), to create a controlled and constant auditory
background for testing and to mask any sounds of other activities
occurring within the testing building (e.g. opening/closing doors,
voices, monkeys vocalizing or shaking a cage in a nearby testing
room). The end of the 30 min HIT was immediately followed by a
second blood draw. The increase in cortisol levels present in this
J. W. Schwartz et al. / Animal Behaviour 190 (2022) 125e138 127
blood sample from baseline levels served as a relative measure of
cumulative physiological stress reactivity in response to the three
HIT conditions (without certainty as to the relative contribution or
arousal level of each stage). Plasma cortisol assays were performed
by the ENPRC Biomarker Core Laboratory using protocols described
in detail elsewhere (Pincus, 2018, pp. 124e125). For two out of 111
HIT trials, pre- and/or post-HIT cortisol concentrations could not be
obtained due to problems with blood sample collection or cortisol
assays. Three additional HIT trials were associated with anoma-
lously small or large measurements of increases in cortisol con-
centrations from before to after the HIT (i.e. <0
m
g/ml, or >50
m
g/
ml), likely reflecting assay problems or other methodological is-
sues. Those five HITs were removed from further consideration,
leaving 106 trials by 41 monkeys analysed.
Video and audio recordings of HITs were obtained with a video
camera (Sony DCR-SR85, Tokyo, Japan). Raw audio was extracted
from the video and converted to a mono, 44.1 kHz, 24-bit WAV
format using Adobe Audition 3.0 (Adobe Systems, San Jose, CA,
U.S.A.). To mitigate effects of the noise machine present in the
testing room, prior to coo extraction, audio was prepared using the
‘Noise Reduction (process)’function in Audition. The noise was
consistent across trials, but to account for any possible variation, a
spectral profile was generated for each trial from a 5 s sample of
uninterrupted noise occurring immediately after a coo (the closest
coo to the middle of the HIT), and then the whole sound file un-
derwent reduction of noise fitting that specific spectral profile,
theoretically leaving acoustic energy due to calls unaltered, even
though the frequencies of the noise and of the calls overlapped. The
function was conservatively set to reduce noise by 50% and by
20 dB, resulting in substantial reduction but not complete elimi-
nation of background noise. More aggressive noise reduction set-
tings occasionally interfered with some aspects of the spectral
structure of the coos. The high amplitude and acoustic homoge-
neity of background noise generated by the HoMedics SS-2000 unit
made noise reduction implementation both necessary and,
generally, effective. However, to be confident that noise reduction
did not confound acoustic measurements, we focused acoustic
analyses on the F0 contour of coos, which, based on visual in-
spection of spectrograms with and without noise reduction, was
clearly conserved (Fig. 1).
Thirty-eight out of 41 subjects produced coos during 82 of the
106 HITs. The number of coos produced in each stage of the HIT
(Alone, Profile, Stare) was tabulated. Up to five coos from each of
the three stages of each HIT were randomly selected and isolated
for acoustic analysis. If fewer than five coos were produced during a
single stage, then all coos from that stage were isolated. In no case
were two coos both sampled that were produced within a 1 s
intercall interval of one another. Eight hundred forty-two coos were
isolated in total. One or more acoustic parameters could not be
obtained from nine additional calls, leaving 833 calls analysed from
38 subjects during 82 HITs (6 months: 380 calls from 33 HITs, 16
months: 264 calls from 30 HITs, menarche: 189 calls from 19 HITs).
Scream Recording, Preparation and Selection
Data collection for screams took place over 111 h from
September 2016 through to December 2017 at an enclosure con-
taining a subset of 18 of the 43 subjects (the largest number of
subjects housed in any of the three enclosures) and was limited to
these 18 subjects. Ages associated with recorded screams ranged
from 5.4 to 40 months. The procedures used to record and analyse
screams are described in detail elsewhere (Supplementary Mate-
rials in Schwartz et al., 2020). In brief, recording was done 3 m away
from the enclosure fence. Screams and accompanying behavioural
observations were collected using an all-occurrences sampling
procedure (Altmann, 1974): each time one or more screams were
produced by one of the 18 subjects, the recording was accompanied
by a verbal notation of the identity of the screamer and the nature
of the eliciting behaviour (Table 1). In total, we recorded 224
scream bouts with positive identities (IDs) from among the 18
15
10
5
15
10
5
0.1 s
Time
Frequency (kHz)
Figure 1. Spectrograms of exemplars of coos after (left) and before (right) noise reduction.
J. W. Schwartz et al. / Animal Behaviour 190 (2022) 125e138128
subjects; 181 of these were accompanied by complete behavioural
observations. Subsequent analyses were limited to these 181 bouts.
To count the numbers of screams in each bout, we examined
recordings aurally and visually, through spectrograms, using Adobe
Audition 3.0. We classified each scream into one of five subclasses
(following S. Gouzoules et al., 1984). Analyses were limited to tonal
and pulsed screams due to sample size constraints for other sub-
classes. Tonal screams have a clearly identifiable F0 contour and a
duration lasting 0.25e1 s; pulsed screams are distinguished by their
short duration (<0.25 s) and range in tonality from atonal (noisy,
cough-like) to tonal (clearly identifiable F0) (Fig. 2). One hundred
forty-four scream bouts contained at least one tonal and/or one
pulsed scream that was suitable for acoustic analyses (no clipping,
uninterrupted by other vocalizations or loud sounds, e.g. a monkey
striking an enrichment structure or the fence during a display). Up to
two screams of each subclass (tonal and pulsed) were selected per
bout eon the basis of recording quality or, if more than two screams
of a subclass were of comparable recording quality, then randomly e
resulting in 362 screams (178 tonal screams and 184 pulsed screams)
sampled. One or more acoustic parameters could not be obtained
from 38 tonal screams and 24 pulsed screams (generally due to vocal
noise masking the F0), leaving 300 screams (140 tonal, 160 pulsed)
from 134 bouts by 17 individuals analysed (calls per bouts per in-
dividual: tonal: no threat ¼26/16/8, noncontact threat ¼54/37/15,
contact threat ¼38/26/13, chase ¼13/9/6, attack ¼9/6/4; pulsed: no
threat ¼30/19/6, noncontact threat ¼74/47/15, contact threat ¼38/
25/12, chase ¼13/7/7, attack ¼5/3/3). The maximum time elapsed
between the onset of a screaming bout (usually coincident with the
eliciting threat) and a sampled scream was 26 s.
Acoustic Analyses
The selected vocalizations were isolated from the larger
recording and saved as individual WAV files, using Adobe Audi-
tion. The intentionally generated background noise and its sub-
sequent reduction procedure used with the HIT recordings
precluded valid measurements of amplitude or fine spectral
features of coos (e.g. peak frequency, energy distribution).
However, the coo recordings were well suited to analysis of
duration and F0 contour. In accordance with the recommended
practice of adopting a hypothesis-testing (rather than an
exploratory) approach in acoustic analyses (Fischer et al., 2013),
we analysed 11 parameters that have been identified by multiple
studies in the literature as a correlate of emotional arousal or
valence in other species (Table 2).
To analyse these parameters, spectrograms were generated from
the coos using fast Fourier transform in Praat 6.0.29 (Boersma &
Ween in k, 2 013). Coos were highlighted with the manual selection
tool, excluding any visually detected reverberation, and the spec-
trograms and waveforms were examined, while listening to the
vocalization, to obtain its duration. F0 measurements were then
obtained from the highlighted portion using the ‘Quantify Source’
command in the GSU Praat Tools package v.1.9 (Owren, 2008). The
default settings for this command were used, with the exception that
the pitch ceiling was set to 3000 Hz to capture the full F0 range of
juvenile rhesus macaque coos. The command uses Praat's ‘To Pitc h’
autocorrelation function to estimate a F0 contour, which the user
then inspects and manually corrects if necessary (e.g. octave
correction, removal of any unvoiced segments) (Owren, 2008).
Automatic mean F0, minimum F0 and maximum F0 measurements
were obtained in this way. The Quantify Source command also pro-
duces a mean harmonic-to-noise ratio (HNR) measurement based on
Praat's ‘To Harmonicity’autocorrelation function, and a jitter mea-
surement based on Praat's ‘PointProcess: Get jitter’function. HNR
represents the relative amounts of tonal as opposed to chaoticenergy
in a vocalization, with lower values representing a noisier call; the
generated noise and noise reduction process likely affected the val-
idity of HNR as an absolute parameter; however, given the uniformity
of these procedures within and across trials, HNR measurements
served as a relative measure of call noisiness. Jitter represents the
cycle-to-cycle variation in F0; we used the default measurement
option for jitter, relative average perturbation (RAP).
In addition to the above automatic measurements, one
researcher manually measured onset F0, offset F0 and the time of
the maximum F0. Onset F0 and offset F0 were measured by placing
the cursor at the onset or offset of the vocalization and recording
the F0 estimate produced by the To Pitch autocorrelation function
at that time point. Finally, the time of the maximum F0 was
measured by highlighting the segment of the coo beginning at
onset and ending at the F0 maximum and recording the duration of
this segment. Onset F0 slope was calculated by dividing the fre-
quency difference between start and maximum F0 values by the
time distance between these two points; coos where this time
distance comprised less than 5% of the total call duration were
excluded from this measurement. Although coos in a broader set of
contexts often end with a descending F0 (Hauser &Marler, 1993),
many of the coos analysed here ended at the F0 maximum, and
therefore we did not analyse offset F0 slope.
Procedures used to process and analyse screams have been
described in detail elsewhere (Supplementary Materials in
Schwartz et al., 2020). The same parameters measured for the coos
(Table 2) were obtained for the screams, excluding the onset F0
slope variable (since tonal and pulsed scream classes do not typi-
cally exhibit the rising, then falling F0 contour characteristic of
coos), using the Quantify Source command within the GSU Tools
package in Praat. In addition, we measured five spectral parameters
associated with arousal across many mammalian species (Table 3),
by applying a 1024-point fast Fourier transform and obtaining
automated measurements available in Raven Pro 1.5 sound analysis
software (Center for Conservation Bioacoustics, 2014).
Table 1
Categories of aggression accompanying screams, along with numbers of incidents accompanied by each behaviour type analysed versus recorded during the data collection
period
Intensity Behaviour Description Nanalysed (observed)
1 No threat Subject received no visible aggressive or dominance behaviour 24 (33)
2 Noncontact threat Subject received a visible threat (e.g. forward lunge, open-mouth stare) by another
individual, no physical contact
61 (81)
3 Threat with contact Subject received a visible threat including physical contact, often grabbing 34 (44)
4 Chase Subject ran from and was pursued by another individual for at least 2 s 11 (15)
Attack Subject was physically pinned or restrained by another individual for at least 2 s,
often involving biting
6 (8)
J. W. Schwartz et al. / Animal Behaviour 190 (2022) 125e138 129
Statistical Analyses
Aiming to constrain the number of significance tests and avoid
type I errors, we used principal components analysis to scale and
reduce the measured acoustic parameters to a small number of
orthogonal acoustic dimensions (principal components, hereafter
PCs), using the ‘prcomp’package in R (R Core Team, 2018). Coos and
screams were analysed separately. In subsequent analyses, rather
than raw acoustic variables, we tested the effects of arousal and
other predictors on acoustic PC scores, specifically scores on those
PCs explaining at least 10% of the total acoustic variance. To test our
hypotheses about the relationship between arousal and acoustics in
coos and screams, while accounting for acoustic variation due to
other sources such as individual identity and age, we fitted linear
mixed models (LMM), each with scores on one acoustic principal
component as the outcome variable. We adopted a significance-
testing approach to modelling, since our main goal was to test for
the presence of general associations between vocal acoustics and
arousal along with other predictors.
For the coo analyses, the increase in plasma cortisol from
baseline (pre-HIT) to after the HIT served as a primary proxy for a
monkey's stress and emotional arousal reaction to the test and was
thus included as a fixed effect in the LMMs. In addition to cortisol
increase, fixed effects included age in months (as a continuous
covariate) and stage of the HIT (Alone, Profile, Stare), primarily to
hold these variables constant while estimating the effect of cortisol
15
10
5
15
10
5
(a)
(b)
Time
Frequency (kHz)
0.1 s
Figure 2. Representative spectrograms of (a) tonal and (b) pulsed screams.
Table 2
Acoustic parameters obtained from coos and selected sources indicating a relationship with arousal
Parameter Description Source
Duration (s) Time elapsed between onset and end of a single vocalization Bayart et al., 1990 (macaques); Briefer, Maigrot, et al., 2015
(horses); Sugiura, 2007 (Japanese macaques)
Mean F0 (Hz) Mean value of the fundamental frequency (F0) over the
vocalization
Briefer, Maigrot, et al., 2015 (horses); Briefer, Tettamanti, &
McElligott, 2015 (goats); Rendall, 2003 (baboons); Szipl et al.,
2017 (ravens); Yeon et al., 2011 (cats)
Max. F0 and min. F0 (Hz) Maximum and minimum values of the F0 over the
vocalization; F0 range is calculated as the difference
Bayart et al., 1990 (macaques); Briefer, Maigrot, et al., 2015
(horses); Lingle et al., 2012 (review); Sugiura, 2007 (macaques);
Yamaguchi et al., 2010 (marmosets)
Onset F0 and end F0 (Hz) F0 values at onset and offset of the vocalization Briefer, Maigrot, et al., 2015 (horses); Briefer, Tettamanti, &
McElligott, 2015 (goats); Rendall, 2003 (baboons)
Relative time of max. F0 (proportion) The timing of the maximum value of the F0 as a proportion
of total duration
Ord
o~
nez-G
omez et al., 2019 (spider monkeys); Rendall, 2003
(baboons); Yamaguchi et al., 2010 (marmosets)
Onset F0 slope (Hz/s) The difference between max. and onset F0 divided by the
time distance between the two points
Rendall, 2003 (baboons); Sugiura, 2007 (macaques)
Mean harmonics-to-noise ratio (dB) A measure of the relative amounts of tonal and noisy
(chaotic) energy in a vocalization
Blumstein &Chi, 2012 (marmots); Liao et al., 2018
(marmosets); Linhart et al., 2015 (pigs); Siebert et al., 2011
(goats)
Jitter (proportion) Cycle-to-cycle variation in period length; a measure of F0
instability
Lehoczki et al., 2019 (dogs); Puppe et al., 2005 (pigs); Rendall,
2003 (baboons); Siebert et al., 2011 (goats); Szipl et al., 2017
(ravens)
See also Briefer (2012) for review.
J. W. Schwartz et al. / Animal Behaviour 190 (2022) 125e138130
reactivity, and also secondarily to test whether coo acoustics would
differ over the course of the HIT and over development. To control
for potential differences in cortisol reactivity depending on baseline
cortisol levels (e.g. high baseline cortisol blunting cortisol reac-
tivity), baseline cortisol and its interaction with cortisol increase
were both included as fixed effects. An interaction between age and
cortisol increase was also included as a fixed effect, to investigate
whether and how the vocal expression of arousal might change
over development. To account for repeated observations of in-
dividuals within and across the three age points and HIT stages, we
included a nested random effect of HIT stagewithin HIT trial within
individual, as well as a random slope effect of age category by in-
dividual (Schielzeth &Forstmeier, 2009).
The analysis of screams was also carried out using LMM with
acoustic PCs as outcome variables. The eliciting behavioural context
(Table 1) served as a proxy for emotional arousal and was included
as a categorical fixed effect. Attack and chase were combined into a
single category, due to low frequency counts for these two situa-
tions and the assumption that attacks and chases both evoke
greater arousal in recipients of aggression than other documented
eliciting behaviours. Other fixed effects included age, scream class
(tonal or pulsed) and the interaction between eliciting behaviour
and scream class, to determine whether and how the effects of
arousal on scream acoustics might differ between tonal and pulsed
screams. Scream bout and subject were included as nested random
intercept effects.
LMM were fitted by restricted maximum likelihood, with a
Gaussian distribution and identity link, using the ‘lmer’function
(‘lme4’package; Bates et al., 2015) in R. Degrees of freedom and P
values were estimated using Satterthwaite's method via the
‘summary()’and ‘anova()’functions in the ‘lmerTest’package
(Kuznetsova et al., 2017). All continuous predictors were scaled and
centred to allow meaningful interpretation of main effects in the
presence of interactions (Schielzeth, 2010). The interaction terms
specified above, when nonsignificant, were removed from the
models because they limit estimation of main effects to one point
along the covariate rather than estimating an overall average effect
(Engqvist, 2005). They were then individually re-added to estimate
the effect size and significance of the interaction. We conducted
post hoc pairwise comparisons between levels of categorical fixed
effects in R using the KenwardeRoger method, with Tukey HSD P
adjustment, via the ‘emmeans’package (Lenth, 2016). Residuals of
each LMM were checked visually for homoscedasticity and
normality. Counts of numbers of vocalizations included many zeros
(HIT trials lacking any coos and scream instances lacking one
scream class) and were therefore tested with zero-inflated Poisson
GLMM, using the ‘glmmTMB’package (Brooks et al., 2017), instead
of LMM, with Pvalues for categorical predictors calculated using
likelihood ratio tests. All significance testing used an alpha of 0.05.
RESULTS
For a full report of the model outputs, see the Supplementary
Material.
Coos
Principal components analysis of coo acoustics yielded three
PCs, each explaining at least 10% of the total acoustic variance
among coos, labelled CooPC1, CooPC2 and CooPC3 to distinguish
the separate analyses used for coos and screams. Inspection of the
loadings (Table 4) suggested these three PCs could be conceptual-
ized as generally mapping onto three biologically coherent acoustic
dimensions: CooPC1 represented overall F0, CooPC2 represented F0
modulation along with noisiness, and CooPC3 represented tem-
poral parameters.
Higher arousal, as measured by greater increases in cortisol from
before until after the HIT, was associated with coos with signifi-
cantly higher overall F0 (CooPC1) and significantly greater F0
modulation and noisiness (CooPC2) (Table 5,Fig. 3). Most param-
eters also changed significantly over the course of the HIT, with the
Alone stage producing coos with higher overall F0 (CooPC1) than
the Profile and Stare conditions and longer call duration (CooPC3)
than the Stare condition, and the Stare condition yielding signifi-
cantly more coos than the other conditions (Table 6). As monkeys
matured from infancy (6 months) to the juvenile, prepubertal
period (16 months) to menarche, their coos on average exhibited
shifts towards a lower overall F0 (CooPC1) and greater F0 variability
and/or more noise (CooPC2) (Table 5,Fig. 3). Finally, we observed a
significant negative interaction between cortisol reactivity and age
on CooPC1, suggesting that the positive association between
cortisol reactivity and overall F0 of coos was dampened as in-
dividuals matured (Table 5,Fig. 3). Because the variables were
centred, the significant positive main effect of cortisol reactivity on
CooPC1 indicates that this association was statistically significant
for individuals of the mean age of monkeys in our sample
(approximately 14 months of age) (Schielzeth, 2010).
Screams
Principal components analysis of scream acoustics yielded three
PCs, each explaining at least 10% of the total acoustic variance
Table 3
Additional acoustic parameters obtained from screams, and selected sources indicating a relationship with emotion
Parameter Description Source
Mean peak frequency (Hz) The frequency with the most energy on average over the
vocalization
Puppe et al., 2005 (pigs); Yeon et al., 2011 (cats)
DFA25, DFA50, DFA75 (Hz) Distribution of frequency amplitudes: frequency values of the
upper limits of the first three quartiles of energy, on average
over the vocalization
Briefer, Maigrot, et al., 2015 (horses); Gogoleva et al., 2010
(foxes); Linhart et al., 2015 (pigs); Yeon et al., 2011 (cats)
Interquartile range (Hz) Difference between DFA75 and DFA25; width of energy
distribution across the frequency spectrum
Puppe et al., 2005 (pigs); Schrader &Todt, 1998 (pigs); Siebert
et al., 2011 (goats)
Table 4
Results of principal components analysis of coo acoustics, including the percentage
of variance explained and loadings by each acoustic parameter
CooPC1 CooPC2 CooPC3
% Variance 51.23 20.56 11.08
Call duration 0.16324 0.013004 0.623777
Mean F0 0.414545 0.044036 0.075839
Min. F0 0.357457 0.306653 0.033809
Max. F0 0.410116 0.09491 0.093322
F0 range 0.286373 0.42145 0.10852
Onset F0 0.354305 0.291149 0.053625
Offset F0 0.408035 0.03336 0.007716
Time of max. F0 0.100775 0.091 0.74585
Onset F0 slope 0.297523 0.33828 0.153923
Jitter 0.17526 0.45316 0.017331
Mean HNR 0.0459 0.550704 0.01719
Acoustic parameters loading highly (>0.3) are presented in bold.
J. W. Schwartz et al. / Animal Behaviour 190 (2022) 125e138 131
among screams. As with coos, ScreamPC1 represented overall F0 or
overall call frequency, with loadings from mean F0, maximum F0,
mean peak frequency and all three measures of distribution of
frequency amplitude (DFA) (Table 7). ScreamPC2 mapped roughly
onto the acoustic characters distinguishing tonal from pulsed
screams, representing a dimension ranging from relatively short,
noisy screams with a small frequency range (F0 range and inter-
quartile range, IQR) (more ‘pulsed-like’), to longer, more tonal
screams with a larger frequency range (more ‘tonal-like’)(Table 7).
High values on ScreamPC3 denoted a screamwith a high maximum
F0, large F0 range, high DFA75 value and a small IQR (Table 7).
The results of a different analysis using the same data were
previously presented in a review article (Schwartz et al., 2020).
Within both pulsed and tonal screams, more intense attack and
chase contexts were associated with higher F0 and overall fre-
quency (ScreamPC1) (Table 8,Fig. 4). Within pulsed but not tonal
screams, attack/chase contexts were associated with increases in
ScreamPC2 (Table 8,Fig. 4), suggesting those contexts yielded
pulsed screams that were shorter and noisier, i.e. more stereotyp-
ically ‘pulsed-like’as opposed to ‘tonal-like’. With regard to
development, pulsed screams showed significant decreases on
ScreamPC3 as monkeys matured (Table 8). Furthermore,
ScreamPC3 showed a significant interaction between age and
context in pulsed screams: with the developmental trend described
above seemingly limited to chase and attack contexts (slope of age
for chase/attack contexts ±SE: 1.27 ±0.43; chase/attack eno
threat: 1.35 ±0.48, P¼0.030; chase/attack enoncontact
threat: 1.34 ±0.46, P¼0.024; all other pairs: P>0.05). No other
significant developmental changes in scream acoustics were
observed.
Scream usage varied significantly depending on context, age and
the interaction between the two (Table 8,Supplementary
Materials). At the average age across observed interactions (22.4
months), significantly more tonal screams were produced during
attack and chase contexts compared to other contexts (attack/chase
eno threat: estimate ±SE ¼0.76 ±0.19, P<0.001; attack/chase e
noncontact threat: 0.85 ±0.15, P<0.001; attack/chase econtact
threat ¼0.67 ±0.16, P<0.001; all other pairs: P>0.05); as in-
dividuals matured, they produced fewer tonal screams per bout
overall (Table 8), particularly during the no-threat context (slope of
age in noncontact threat category: 0.54 ±0.12; differences in
slope of age between contexts: no threat enoncontact
threat: 0.57 ±0.17, P¼0.006; no threat econtact
threat: 0.46 ±0.16, P¼0.018; all other pairs: P>0.05). Pulsed
screams were produced in significantly greater abundance during
noncontact threats than in the absence of any threat, at the average
age (estimate ±SE ¼0.53 ±0.14, P<0.001; all other pairs:
P>0.05). As individuals matured, they also produced fewer pulsed
screams overall (Table 8), but began to produce more pulsed
Table 5
Coo LMM results
Acoustic variable Coeff ±SE P
Number of calls (log)
a
Cortisol increase (
m
g/ml) 0.007±0.005 0.165
Baseline cortisol (
m
g/ml) 0.080±0.008 <0.001
Increase*baseline ee
Age (months) 0.121±0.019 <0.001
Cortisol increase*age ee
CooPC1 (overall F0)
Cortisol increase (
m
g/ml) 0.40±0.18 0.029
Baseline cortisol (
m
g/ml) 0.01±0.11 0.910
Increase*baseline 0.08±0.10 0.419
Age (months) 1.51±0.18 <0.001
Cortisol increase*age 0.64±0.14 <0.001
CooPC2 (F0 variability)
Cortisol increase (
m
g/ml) 0.41±0.13 0.004
Baseline cortisol (
m
g/ml) 0.13±0.09 0.156
Increase*baseline 0.01±0.07 0.849
Age (months) 0.74±0.12 <0.001
Cortisol increase*age 0.03±0.11 0.807
CooPC3 (temporal)
Cortisol increase (
m
g/ml) 0.08±0.11 0.447
Baseline cortisol (
m
g/ml) 0.13±0.07 0.069
Increase*baseline 0.10±0.06 0.077
Age (months) 0.14±0.10 0.164
Cortisol increase*age 0.05±0.10 0.594
P<0.05 presented in bold. Effect size and Pvalue estimates for main effects were
obtained by first removing nonsignificant interaction effects from the model
(increase*baseline and/or cortisol increase*age). LMMs also included human
intruder laboratory test (HIT) stage as a fixed effect; results are presented in Table 6.
a
Raw counts analysed using GLMM with Poisson distribution, log link and zero-
inflation parameter (glmmTMB function, glmmTMB package). Estimates presented
on a log scale. Interaction effects and scaling/centering caused failure to converge
and are thus excluded.
3
2
1
0
CooPC2
–1
–3
10 20
Cortisol increase (µ
g
/ml)
30
–2
6Age
3
0
CooPC1
–3
–6
6 months
16 months
Menarche
Figure 3. Relationships between arousal, age and acoustic dimensions of coos (rep-
resenting overall F0 and overall F0 modulation, respectively). Points represent the
mean value for all coos sampled from a single human intruder test, HIT (N¼1e5);
error bars represent SEs within the HIT. Age is presented categorically for visual clarity,
but was analysed as a continuous covariate in LMMs.
J. W. Schwartz et al. / Animal Behaviour 190 (2022) 125e138132
screams in response to threats with contact (slope of age in contact
threat category: 0.29 ±differences in slope of age between con-
texts: contact threat eno threat: 0.32 ±0.11, P¼0.016; contact
threat enoncontact threat: 0.58 ±0.10, P<0.001; contact threat e
attack/chase: 0.65 ±0.19, P¼0.004; all other pairs: P>0.05).
DISCUSSION
The present study identified similarity in the acoustic correlates
of emotional arousal for subsets of two rhesus macaque call types,
coos and screams: within each call type, greater arousal was
associated with increases in an acoustic dimension representing
overall F0, although this association decreased in coos as monkeys
matured. These similar patterns are interesting in light of the highly
disparate contexts in which the coos and screams were recorded
and the different methods used to measure arousal for each,
namely, cortisol measurement in a laboratory behavioural stress
test (HIT) for coos and intensity of aggression received in naturally
occurring social interactions for screams. Coos also showed
arousal-dependent variation in an additional acoustic dimension
relating to F0 modulation parameters. These results along with
those of other studies (reviewed in Briefer, 2012) are consistent
with the hypothesis of evolutionary homology in emotional
mechanisms of vocal production (Filippi et al., 2017;Zimmermann
et al., 2013), both between these two call types and between rhesus
macaques and other species. They also carry broader implications
regarding general processes in vocal evolution, discussed below.
Vocal development was seen in maturational changes to the overall
acoustics of, and acoustic correlates of arousal within coos, but only
to a very limited extent within screams. These contrasts could be
due to differences in social function and/or proximate mechanisms
of vocal production between call types and across development
and/or differences in the methods used to study the two call types.
These possible interpretations are discussed below.
Acoustic Correlates of Arousal in Coos and Screams: Evolutionary
Implications
Increased calling is a well-documented indicator of arousal in a
diverse array of mammals (Briefer, 2012). Consistent with this
trend, we found a significant positive association between cortisol
increase and the number of coos produced, while highly intense
attack and chase contexts generally yielded more tonal (but not
pulsed) screams than other contexts. In addition to call usage,
arousal can also influence the precise acoustic structure of a call,
largely due to its broad effects on muscular action and tension
including within the vocal apparatus. Numerous studies have
pointed to a positive association between emotional arousal and
vocal F0, likely due to increased tension and action in laryngeal
muscles (Briefer, 2012;Titze, 1994). However, very few studies have
explicitly examined the effects of arousal on acoustic variation
within different call types of rhesus macaques (Bayart et al., 1990;
Patel &Owren, 2007), and to our knowledge, no previous study has
done so across call types. In line with our hypotheses, results
showed considerable within-call type variation in overall F0 and a
positive relationship between arousal and overall F0 in both coos
and screams (including both scream classes), suggesting homology
with other previously studied mammalian taxa (Briefer, 2012)
including other primates (e.g. Rendall, 2003).
The results of the present study and others (reviewed in Briefer,
2012) suggest that arousal is strongly linked to within-call type
variation in vocal F0, and this link is consistent across taxa as well
as across call types within a species, even call types with disparate
associated contexts and functions. Consequently, we propose that
the F0 (and perhaps other acoustic parameters) of some vocaliza-
tions may be best understood as having been shaped not only by
selection pressures directly related to communicative function per
se, but also by forces related to emotional states and their other
functions. Specifically, the essentially simultaneous and correlated
outputs of emotional arousal eincluding muscle tension in the
vocal apparatus and resulting acoustic changes (especially in the
F0), physiological changes such as mobilization of energy stores and
behavioural changes such as heightened vigilance eall share a
common proximate mechanism, and therefore might represent the
cumulative and aggregate outcome of the many selective pressures
acting on them, a common situation for highly correlated traits (see
Lande &Arnold, 1983). Thus, scientific understanding of the evo-
lution of the acoustic structure of vocalizations may benefit from
further research into the proximate role of emotion in vocal
production.
Note, however, that the correlation between vocal F0 and
arousal found in this study likely does not capture all dimensions of
vocal emotion expression in screams. Tonal and pulsed screams are
just two of the scream classes exhibited by rhesus macaques and
thus might not be representative of screams as a whole in this
species, especially since they elicit weaker responses from allies in
the group, usually matrilineal kin, than do ‘noisy’and ‘arched’
scream classes (corresponding to severe aggression and rank
challenge, respectively) (S. Gouzoules et al., 1984). Many screams
represented in this species do not include an identifiable F0 contour
(likely due to chaotic vibration of the vocal folds; Fitch et al., 2002),
let alone show a relationship between vocal F0 and arousal. In
addition, it is important to reiterate here that macaque coos occur
not only during social separation but also preceding friendly con-
tact and in anticipation of food (Bayart et al., 1990;Hansen, 1976;
Hauser &Marler,1993), each of which is probably associated with a
different emotional state; the acoustic correlates of arousal in coos
within these other contexts remain open to investigation.
In addition to overall F0, the literature also indicates F0 modu-
lation (the variability of the F0 within a single call) as a significant
indicator of arousal (e.g. Lehoczki et al., 2019;Lingle et al., 2012).
Consistent with this trend, and in support of our hypothesis, the
coos of subjects showing stronger arousal reactions to the HIT (i.e.
greater cortisol increase) exhibited significantly higher scores on an
Table 6
Results of post hoc pairwise comparisons of coo acoustics and human intruder laboratory test (HIT) stages with Tukey adjustment
Acoustic dimension Test statistic Overall PProfile eAlone Stare eAlone Stare eProfile
Coeff ±SE PCoeff ±SE PCoeff ±SE P
Number of calls
a
c
2
¼1394.6 <0.001 0.02±0.02 0.652 0.61±0.02 <0.001 0.59±0.02 <0.001
CooPC1 F¼10.94 <0.001 0.54±0.17 0.004 0.62±0.14 <0.001 0.09±0.16 0.852
CooPC2 F¼2.23 0.112 dddddd
CooPC3 F¼4.98 0.008 0.28±0.14 0.093 0.35±0.11 0.008 0.06±0.13 0.881
P<0.05 presented in bold.
a
Raw counts analysed using GLMM with Poisson distribution, log link, and zero-inflation parameter (glmmTMB function, glmmTMB package). Estimates presented on a log
scale. Inclusion other fixed effects alongside human intruder laboratory test (HIT) stage caused failure to converge, hence other fixed effects were excluded while estimating
effects of HIT stage.
J. W. Schwartz et al. / Animal Behaviour 190 (2022) 125e138 133
acoustic dimension relating to F0 modulation. Lingle et al. (2012)
pointed out that a higher maximum F0 (and therefore greater
overall F0 range) is one of the major features distinguishing contact
calls produced during social separation or isolation from those
produced within the group, and suggested that, in general, the
acoustic differences between calls produced in these contexts
might be attributable to arousal differences. In rhesus macaques,
coos are produced in a wide range of contexts that appear to vary in
arousal level (e.g. Bayart et al., 1990;Hauser &Marler, 1993). The
correlation observed in the present study between F0 modulation
and cortisol increase appears to show the graded acoustic variation
hypothesized by Lingle et al. (2012) at a small scale, i.e. within the
class of coos that might be interpreted based on context as ‘sepa-
ration calls’.
Previous studies of Japanese macaques, Macaca fuscata (Sugiura,
2007), and common marmosets, Callithrix jacchus (Yamaguchi et al.,
2010), suggested that greater maximum F0 and/or F0 modulation in
the context of social isolation might enhance the localizability of
calls, enhancing the odds of reunion with groupmates. Indeed, in
the present study, coos produced when the monkeys were alone
were significantly greater in overall (including maximum) F0
compared to those produced in the presence of an unfamiliar hu-
man, although the same was not true of the acoustic dimension
relating to F0 modulation per se. Another nonmutually exclusive
evolutionary possibility has to do with selection favouring a
broader arousal response, based on, for example, increased
antipredator vigilance when individuals are most vulnerable and
mobilization of energy stores that would aid in escape. These se-
lection pressures could result in the evolution of an aroused
emotional response to social isolation, which could in turn result in
changes to proximate mechanisms of vocal production (e.g. muscle
tension in the larynx), thereby yielding a change in vocal acoustics.
Our results suggest the degree of F0 modulation in rhesus macaque
coos is influenced by arousal, making this evolutionary scenario
more plausible.
While it has been established that the F0 of a call can correlate
with and potentially communicate the sender's arousal level, that is
of course not the only or even the primary communicative function
of many call types. Acoustically distinct call types can mediate so-
cial interactions or communicate complex information to receivers,
raising the question of whether and how acoustic variation related
to arousal maps to variation delineating the boundaries among call
types (Schwartz et al., 2020). In addition to an acoustic dimension
relating to overall F0 or call frequency (ScreamPC1), our principal
components analysis of screams identified another dimension
(ScreamPC2) mapping roughly onto the acoustic parameters
delineating pulsed and tonal screams (the two scream subclasses
analysed in the present study). One possible interpretation is that
these two orthogonal acoustic dimensions might relate to semi-
separate acoustic channels, one serving to delineate the boundaries
among scream classes, while the other relates to sender arousal in a
similar manner across scream classes and other call types
(Schwartz et al., 2020).
Indeed, while both pulsed and tonal screams showed within-
scream class variation in overall F0 and call frequency that was
significantly associated with our measure of sender arousal, the
same was not true of the second acoustic dimension. Rather, tonal
screams showed no significant association between agonistic in-
tensity and ScreamPC2. Pulsed screams produced during highly
intense chase and attack contexts were characterized by signifi-
cantly higher values on ScreamPC2, meaning generally more ste-
reotypically ‘pulsed-like’, almost as if more intense aggression
necessitated more clearly delineating pulsed screams from other
scream classes. The originally reported context in which pulsed
screams disproportionately occur was that of aggression by a
higher-ranking matrilineal relative (S. Gouzoules et al., 1984). To
speculate, one possibility is that producing more stereotypically
‘pulsed-like’pulsed screams in the context of especially severe
aggression might function to abate that aggression by communi-
cating to the aggressor and/or bystanders that the screaming
Table 7
Results of principal components analysis of scream acoustics, including percentage
of variance explained and loadings by each acoustic parameter
ScreamPC1 ScreamPC2 ScreamPC3
% Variance 43.24 14.83 10.25
Call duration 0.12001 0.32809 0.215005
Mean F0 0.35742 0.106177 0.232443
Min. F0 0.24252 0.437853 0.1106
Max. F0 0.33291 0.07855 0.372365
F0 range 0.22022 0.3357 0.465179
Onset F0 0.26957 0.266387 0.152005
Offset F0 0.28478 0.130282 0.094369
Time of max. F0 0.04256 0.27496 0.08855
Jitter 0.04783 0.25514 0.288804
Mean HNR 0.06715 0.390798 0.21224
Mean peak frequency 0.34665 0.04655 0.12135
DFA25 0.30816 0.179073 0.08592
DFA75 0.32796 0.16301 0.35072
DFA50 0.33702 0.00255 0.23365
IQR 0.20262 0.35824 0.40553
Acoustic parameters loading highly (>0.3) presented in bold.
Table 8
Scream LMM results
Acoustic dimension Context Age Context*age
For
c
2
PCoeff ±SE PFor
c
2
P
Pulsed screams
Number of calls (log)
a
c
2
¼42.0 <0.001 0.36±0.17 0.031
c
2
¼41.9 <0.001
ScreamPC1 F¼3.04 0.033 0.04±0.25 0.877 F¼0.47 0.701
ScreamPC2 F¼4.84 0.036 0.01±0.12 0.906 F¼2.01 0.119
ScreamPC3 F¼0.66 0.581 1.27±0.43 0.004 F¼3.02 0.034
Tonal screams
Number of calls (log)
a
c
2
¼31.7 <0.001 0.31±0.14 0.025
c
2
¼13.6 0.003
ScreamPC1 F¼5.04 0.003 0.47±0.32 0.150 F¼1.34 0.267
ScreamPC2 F¼0.88 0.455 0.07±0.14 0.629 F¼0.42 0.737
ScreamPC3 F¼1.98 0.125 0.08±0.14 0.564 F¼0.56 0.643
Effect size and Pvalue estimates for main effects were obtained by first removing nonsignificant interaction effects (context*age) from the model. P<0.05 presented in bold.
For detailed results of post hoc tests, see Supplementary Materials.
a
Raw counts analysed using GLMM with Poisson distribution, log link and zero-inflation parameter (glmmTMB function, glmmTMB package). Estimates presented on a log
scale.
c
2
obtained via likelihood ratio test.
J. W. Schwartz et al. / Animal Behaviour 190 (2022) 125e138134
monkey is the aggressor's relative. We lacked data on the identity of
the aggressor, which would help to elucidate this matter. More
broadly, we recommend that future studies continue to differen-
tiate between acoustic variation between versus within call types
and to test the ongoing hypothesis that these two kinds of acoustic
variation generally occupy semiseparate acoustic dimensions and
serve distinct communicative functions.
Effects of Age
The primary aim of the present study was to test hypotheses
about the acoustic correlates of arousal, while accounting for
acoustic variation related to development. Nevertheless, examining
the correlations between age and vocal behaviour yielded inter-
esting results, providing mixed support for our hypotheses. We
found developmental changes in the number of calls produced, for
all call types, including changes in which screams were associated
with which agonistic contexts. This amounts to a developmental
change in call usage, which are generally well documented in other
primates, more so than changes in the general acoustic structure of
calls (Egnor &Hauser, 2004). With regard to vocal acoustics, we
predicted that we would observe decreases in overall F0 and F0
modulation and increases in duration of coos over the first few
years of development, in line with past research (Hammerschmidt
et al., 2000). These predictions were supported by significant de-
creases in overall F0 of coos as monkeys matured. These changes
are likely partly attributable to developmental growth of the vocal
apparatus, with larger structures producing lower frequencies. We
additionally observed changes along an acoustic dimension
amounting to increased noisiness (and contrary to our hypotheses,
F0 modulation) over development, perhaps reflecting develop-
mental decreases in the periodicity of vocal fold oscillation (see
Fitch et al., 2002), at least in the context of coos produced during
the HIT.
In addition to overall changes in coo acoustics over the course of
development, the acoustic correlates of arousal also appeared to
undergo developmental changes, specifically, the positive correla-
tion between cortisol increase and overall F0, including maximum
F0, was dampened as monkeys matured (although still significantly
positive at the average age studied). This may be related to devel-
opmental changes in how monkeys perceive and react to the HIT,
including separation from the mother. Infant primates' regulation
of emotional and physiological responses to stressors depends in
part on the presence of the mother, i.e. ‘maternal buffering’, with
youngsters gaining the capacity to self-regulate the stress response
as they develop (Sanchez et al., 2015). Thus, it is possible that
development and experience might alter the relationships between
arousal, vocal production and glucocorticoid concentrations. It is
also possible that the relationship between arousal and vocal
acoustics is easier to detect statistically in younger monkeys simply
because they generally show stronger HPA activation in response to
the HIT than older monkeys (Fig. 3), presumably due to the greater
importance of maternal buffering at younger ages and/or repeated
exposure to the HIT.
From an ultimate perspective, it is worth considering the pos-
sibility that the functional significance of separation calls, such as
coos, might change over development, reflected by changes not
only in the acoustics of the call type overall but also in the corre-
lates of arousal within the call type. Since maternal separation
presumably represents a more significant and immediate threat to
fitness in vulnerable infants compared to older monkeys, it is
plausible that a positive link between arousal and the maximum F0
of coos could be more beneficial for infants, as it might make the
calls more localizable, increasing the odds of reunion with the
mother or other groupmates (Lingle et al., 2012;Sugiura, 2007);
this function might become less critical as monkeys age, and
changing selection pressures on coo acoustics in older monkeys
might have resulted in an ontogenetic dampening of the effects of
arousal on the overall F0 of coos. On the other hand, we did not
observe a significant interaction effect between age and cortisol
increase on the acoustic dimension relating to coos’F0 range, which
is thought to be important for localizability (Sugiura, 2007).
Among screams, the only significant acoustic change over
development was in a single dimension showing decreased F0
range and greater bandwidth as monkeys matured; this change was
limited to pulsed screams, particularly those produced during
attack/chase contexts. No significant developmental changes were
observed in any other acoustic dimensions, scream subclasses or
contexts. This result contrasts with past research showing acoustic
changes in the screams of pigtail macaques over the course of
development (H. Gouzoules &Gouzoules, 1989). Why we observed
more substantial developmental changes in the acoustics of coos
ScreamPC2
–2.5
No
threat
Noncontact
threat
Contact
threat
Context
Chase/
attack
0
2.5
5
ScreamPC1
–5
0
5
10 Scream class
Pulsed
Tonal
Figure 4. Acoustic parameters of screams by context and class. Bold error bars show
means ±SE. Horizontal lines denote significant pairwise differences within each
scream class (P<0.05).
J. W. Schwartz et al. / Animal Behaviour 190 (2022) 125e138 135
(both overall acoustics and acoustic correlates of arousal) compared
to screams is unclear. One possible explanation has to do with the
different methods used to study the two call types: while a similar
age range was captured for coos and screams (6e31 months for
coos, 5e40 months for screams), coos were recorded at three
specific age points e6 months, 16 months, and menarche ewhile
screams were recorded on an all-occurrences schedule during
regular weekly observations. It is possible that this difference in
methodology might have influenced the results. Alternatively, the
differences in observed developmental trajectory could have to do
with differences in mechanisms of production and/or call function.
Further research and analysis specifically aimed at modelling the
development of the acoustic structures of coos and screams could
potentially shed further light on these issues.
Methods for Measuring Arousal
Methods used to gauge the emotional states of animals include
inferences based on contextual factors, such as the intensity of
agonism (Filippi et al., 2017;Szipl et al., 2017), and direct mea-
surement of physiological components of emotions, such as cortisol
released as a result of activation of the HPA axis in response to a
perceived stressor (Bayart et al., 1990;Blumstein &Chi, 2012;Ralph
&Tilbrook, 2016;Schrader &Todt, 1998;S
ebe et al., 2012). The
present study utilized both methods ethe former for screams and
the latter for coos. Neither method is without limitations.
Context-based inferences about emotional states are convenient
and noninvasive, but they rely on assumptions about how an ani-
mal perceives and reacts to a particular situation. In particular,
monkeys of different ages, sex or rank most likely exhibit different
emotional and behavioural reactions to similar social situations
(e.g. Cheney, 1977), warranting caution in any extrapolation of our
results to adult or male rhesus macaques. Another limitation is that
agonistic intensity likely correlated not only with caller arousal but
also with physical activity; it is possible this variation in physical
activity may have generated differences in call acoustics due to
changes in a monkey's posture, respiration, etc. Thus, it is not
possible in the present study to fully disentangle the acoustic cor-
relates of arousal in screams from those of physical activity,
although the results do align with hypotheses surrounding the ef-
fects of arousal on the voice. We are not aware of previous studies
into the vocal expression of arousal directly accounting for physical
activity; future studies may benefit from measuring and controlling
for this potential third variable. Other factors we were not able to
observe, such as the identity of the aggressor and the proximity of
potential aiders, likely added unaccounted-for variation in animals'
emotional states as well as behaviour, although we do not see a
reason that this might have confounded our results or invalidated
our basic assumption that, on average, receiving an attack or chase
elicits greater arousal in animals of all ages than, for example, a
mere lunge forward (noncontact threat).
Cortisol levels are also limited as indicators of emotion: HPA
activation occurs not only in response to emotionally arousing
stimuli, but also in response to physiological challenges, such as
exercise, temperature extremes, immune challenge and pain, and
additionally exhibits circadian and ultradian rhythms that are
regulated differently from the responses to the above stressors
(Charmandari et al., 2005;Ralph &Tilbrook, 2016). We partially
controlled for this by measuring the absolute increase in cortisol
concentrations from baseline (pre-HIT) to after the HIT, which
occurred at the same time of day across trials; this method effec-
tively gauges the degree of HPA activation in response to a stressor
(Romero, 2004). Under such controlled conditions, stress-related
cortisol increases are reasonably interpreted as an indicator of an
animal's emotional arousal reaction.
Operational assessment of arousal, and call recording, involved
different methods for coos and screams, so drawing contrasts be-
tween call types requires caution. Another complicating factor is
the subjects’familiarity with agonistic encounters, which occur
very frequently in the large social groups at the ENPRC, compared
with the novelty of the HIT. As a result, subjects may have been
more highly aroused by the latter on average. However, despite
differences in method and context, we found general similarity
between coos and screams regarding the relationship between
arousal and overall F0, a pattern that rhesus macaques share with
other species (Briefer, 2012).
Conclusions
In the present study, rhesus macaque coos and pulsed and tonal
screams alike show evidence of a positive association between
arousal and F0, despite differences between the contexts in which
these vocalizations were recorded. This similarity likely represents
homology with other mammalian taxa. It also suggests that the F0
of vocalizations produced in high-arousal contexts is integrated to
an extent with arousal and its other physiological and behavioural
concomitants, supporting the notion that emotion is a significant
factor influencing the evolution of the acoustic structure of vocal-
izations. On the other hand, contrasts between coos and screams
raise interesting, if currently speculative, possibilities regarding the
interplay between emotion, natural selection and vocal production.
Namely, perhaps the different functions of rhesus macaque screams
and coos, within the disparate contexts in which they occur, have
resulted in the evolution of adaptive differences in the effects of
arousal and/or age on acoustic variation within each of these two
call types. These hypotheses require further study before any
definitive adaptive conclusions can be drawn, but they illustrate
how future research and discussion on the evolution of the acoustic
structure of animal vocalizations stand to benefit from careful
consideration of proximate mechanisms, especially the role of
emotional states.
Author Contributions
Jay W. Schwartz: Conceptualization, Methodology, Investiga-
tion, Formal Analysis, Writing eOriginal Draft, Writing eReview &
Editing. Mar M. Sanchez: Conceptualization, Methodology, Inves-
tigation, Resources, Writing eReview &Editing, Supervision,
Project Administration, Funding Acquisition. Harold Gouzoules:
Conceptualization, Methodology, Resources, Writing eOriginal
Draft, Writing eReview &Editing, Supervision, Project
Administration.
Declarations of Interest
None.
Acknowledgments
We thank Dr Mark E. Wilson for the design and supervision of
the human intruder and cortisol testing. Thanks also to Natalie
Brutto, Kelly Bailey, Manuel Bautista, Patrice Rando, Jalani Paul,
Jonathan Engelberg, Jodi Godfrey and Desiree De Leon for their
assistance in data collection. Finally, we thank two anonymous
referees for their feedback and suggestions for revisions to this
manuscript. This study was supported by the U.S. National Science
J. W. Schwartz et al. / Animal Behaviour 190 (2022) 125e138136
Foundation Graduate Research Fellowship under Grant No. DGE e
1343012, by the U.S. National Institutes of Health (NIH) grant
number 1R01HD077623 and the NIH Office of the Director, Office of
Research Infrastructure Programs, P51OD011132 (Emory National
Primate Research Center -ENPRC- base grant). The ENPRC is fully
accredited by AAALAC, International.
Supplementary Material
Supplementary materialassociated with this article is available, in
the online version, at https://doi.org/10.1016/j.anbehav.2022.05.017.
References
Altmann, J. (1974). Observational study of behavior: Sampling methods. Behaviour,
49(3), 227e267. https://doi.org/10.1080/14794802.2011.585831
Bastian, A., & Schmidt, S. (2008). Affect cues in vocalizations of the bat, Megaderma
lyra , during agonistic interactions. Journal of the Acoustical Society of America,
124(1), 598e608. https://doi.org/10.1121/1.2924123
Bates, D., M€
achler, M., Bolker, B., & Walker, S. (2015). Fitting linear mixed-effects
models using lme4. Journal of Statistical Software, 67(1), 1e48. https://doi.org/
10.18637/jss.v067.i01
Bayart, F., Hayashi, K. T., Faull, K. F., Barchas, J. D., & Levine, S. (1990). Influence of
maternal proximity on behavioral and physiological responses to separation in
infant rhesus monkeys (Macaca mulatta). Behavioral Neuroscience, 104(1),
98e107. https://doi.org/10.1037/0735-7044.104.1.98
Blumstein, D. T., & Chi, Y. Y. (2012). Scared and less noisy: Glucocorticoids are
associated with alarm call entropy. Biology Letters, 8(2), 189e192. https://
doi.org/10.1098/rsbl.2011.0832
Boersma, P., & Weenink, D. (2013). Praat: Doing phonetics by computer Version
5.3.51. http://www.praat.org/.
Briefer, E. F. (2012). Vocal expression of emotions in mammals: Mechanisms of
production and evidence. Journal of Zoology, 7(1), 1e20. https://doi.org/10.1098/
rsbl.2011.0832
Briefer, E. F., Maigrot, A. L., Mandel, R., Freymond, S. B., Bachmann, I., & Hillmann, E.
(2015). Segregation of information about emotional arousal and valence in
horse whinnies. Scientific Reports, 4,1e11. https://doi.org/10.1038/srep09989
Briefer, E. F., Tettamanti, F., & McElligott, A. G. (2015). Emotions in goats: Mapping
physiological, behavioural and vocal profiles. Animal Behaviour, 99,131e143.
https://doi.org/10.1016/J.ANBEHAV.2014.11.002
Brooks, M. E., Kristensen, K., Van Benthem, K. J., Magnusson, A., Berg, C. W.,
Nielsen, A., Skaug, H. J., M€
achler, M., & Bolker, B. M. (2017). glmmTMB balances
speed and flexibility among packages for zero-inflated generalized linear mixed
modeling. R Journal, 9(2), 378e400. https://doi.org/10.3929/ethz-b-000240890
Center for Conservation Bioacoustics. (2014). Raven Pro: Interactive sound analysis
software. Ithaca, NY: Cornell Lab of Ornithology Version 1.5. http://
ravensoundsoftware.com/.
Charmandari, E., Tsigos, C., & Chrousos, G. (2005). Endocrinology of the stress
response. Annual Review of Physiology, 67(1), 259e284. https://doi.org/10.1146/
annurev.physiol.67.040403.120816
Cheney, D. L. (1977). The acquisition of rank and the development of reciprocal
alliances among free-ranging immature baboons. Behavioral Ecology and So-
ciobiology, 2(3), 203e318. https://doi.org/10.1007/BF00299742
R Core Team. (2018). R: A language and environment for statistical computing. Vienna:
R Foundation for Statistical Computing. http://www.r-project.org/.
Crockford, C., Wittig, R. M., Mundry, R., & Zuberbühler, K. (2012). Wild chimpanzees
inform ignorant group members of danger. Current Biology, 22(2), 142e146.
https://doi.org/10.1016/J.CUB.2011.11.053
Darwin, C. (1872). The expression of the emotions in man and animals. London: John
Murray. https://doi.org/10.1037/h0076058
Egnor, S. E. R., & Hauser, M. D. (2004). A paradox in the evolution of primate vocal
learning. Trends in Neurosciences, 27(11), 649e654. https://doi.org/10.1016/
J.TINS.2004.08.009
Engqvist, L. (2005). The mistreatment of covariate interaction terms in linear model
analyses of behavioural and evolutionary ecology studies. Animal Behaviour,
70(4), 967e971. https://doi.org/10.1016/j.anbehav.2005.01.016
Fichtel, C., & Manser, M. (2010). Vocal communication in social groups. In
P. Kappeler (Ed.), Animal behaviour: Evolution and mechanisms (pp. 29e54).
Berlin: Springer. https://doi.org/10.1007/978-3-642-02624-9.
Filippi, P., Congdon, J. V., Hoang, J., Bowling, D. L., Reber, S. A., Pa
sukonis, A.,
Hoeschele, M., Ocklenburg, S., de Boer, B., Sturdy, C. B., Newen, A., &
Güntürkün, O. (2017). Humans recognize emotional arousal in vocalizations
across all classes of terrestrial vertebrates: Evidence for acoustic universals.
Proceedings of the Royal Society B: Biological Sciences, 284(1859), Article
20170990. https://doi.org/10.1098/rspb.2017.0990
Fischer, J., Noser, R., & Hammerschmidt, K. (2013). Bioacoustic field research: A
primer to acoustic analyses and playback experiments with primates. American
Journal of Primatology, 75(7), 643e663. https://doi.org/10.1002/ajp.22153
Fischer, J., Wadewitz, P., & Hammerschmidt, K. (2016). Structural variability and
communicative complexity in acoustic communication. Animal Behaviour, 134,
229e237. https://doi.org/10.1016/j.anbehav.2016.06.012
Fitch, W. T., & Hauser, M. D. (2006). Unpacking ‘honesty’: Vertebrate vocal pro-
duction and the evolution of acoustic signals. In A. M. Simmons, R. R. Fay, &
A. N. Popper (Eds.), Acoustic communication (pp. 65e137). New York: Springer.
https://doi.org/10.1007/0-387-22762-8_3.
Fitch, W. T., Neubauer, J., & Herzel, H. (2002). Calls out of chaos: The adaptive sig-
nificance of nonlinear phenomena in mammalian vocal production. Animal
Behaviour, 63(3), 407e418. https://doi.org/10.1006/anbe.2001.1912
Friel, M., Kunc, H. P., Griffin, K., Asher, L., & Collins, L. M. (2019). Positive and
negative contexts predict duration of pig vocalisations. Scientific Reports, 9(1),
2062. https://doi.org/10.1038/s41598-019-38514-w
Gogoleva, S. S., V olodina, E. V. , Volodin, I. A., Kharlamova, A. V., & Trut, L. N. (2010).
The gradual vocal responses to human-provoked discomfort in farmed silver
foxes. Acta Ethologica, 13(2), 75e85. https://doi.org/10.1007/s10211-010-
0076-3
Gottlieb, D. H., & Capitanio, J. P. (2013). Latent variables affecting behavioral
response to the human intruder test in infant rhesus macaques (Macaca
mulatta). American Journal of Primatology, 75(4), 314e323. https://doi.org/
10.1002/ajp.22107
Gouzoules, H. (2005). Monkeying around with symbolism: Are vocalizations simple
symbols…or more like cymbals? In L. Namy (Ed.), Symbol use and symbolic
representation: Developmental and comparative perspectives (pp. 245e265).
Mahwah, NJ: Lawrence Erlbaum.
Gouzoules, H., & Gouzoules, S. (1989). Design-features and developmental modifi-
cation of pigtail macaque, Macaca nemestrina, agonistic screams. Animal
Behaviour, 37, 383e401.
Gouzoules, H., & Gouzoules, S. (1990). Body size effects on the acoustic structure of
pigtail macaque (Macaca nemestrina) screams. Ethology, 85(4), 324e334.
https://doi.org/10.1111/j.1439-0310.1990.tb00411.x
Gouzoules, H., & Gouzoules, S. (1995). Recruitment screams of pigtail monkeys
(Macaca nemestrina): Ontogenetic perspectives. Behaviour, 132(5), 431e450.
https://doi.org/10.1163/156853995X00658
Gouzoules, H., & Gouzoules, S. (2000). Agonistic screams differ among four species
of macaques: The significance of motivation-structural rules. Animal Behaviour,
59(3), 501e512. https://doi.org/10.1006/anbe.1999.1318
Gouzoules, S., Gouzoules, H., & Marler, P. (1984). Rhesus monkey (Macaca mulatta)
screams: Representational signalling in the recruitment of agonistic aid. Animal
Behaviour, 32(1), 182e193. https://doi.org/10.1016/S0003-3472(84)80336-X
Gouzoules, H., Gouzoules, S., & Marler, P. (1985). External reference and affective
signaling in mammalian vocal communication. In G. Zivin (Ed.), The develop-
ment of expressive behavior: Biology environment interactions (pp. 77e101).
Orlando, FL: Academic Press.
Gouzoules, H., Gouzoules, S., & Tomaszycki, M. (1998). Agonistic screams and the
classification of dominance relationships: Are monkeys fuzzy logicians? Animal
Behaviour, 55(1), 51e60. https://doi.org/10.1006/anbe.1997.0583
Hamel, A. F., Lutz, C. K., Coleman, K., Worlein, J. M., Peterson, E. J., Rosenberg, K. L.,
Novak, M. A., & Meyer, J. S. (2017). Responses to the human intruder test are
related to hair cortisol phenotype and sex in rhesus macaques (Macaca
mulatta). American Journal of Primatology, 79(1), 1e10. https://doi.org/10.1002/
ajp.22526
Hammerschmidt, K., Newman, J. D., & Champoux, M. (2000). Changes in rhesus
macaque ‘coo’vocalizations during early development. Ethology, 886,873e886.
Hansen, E. W. (1976). Selective responding by recently separated juvenile rhesus
monkeys to the calls of their mothers. Developmental Psychobiology, 9(1),
83e88. https://doi.org/10.1002/dev.420090111
Hauser, M. D., & Marler, P. (1993). Food-associated calls in rhesus macaques (Macaca
mulatta): I. Socioecological factors. Behavioral Ecology, 4(3), 194e205. https://
doi.org/10.1093/beheco/4.3.194
Howell, B. R., Godfrey, J., Gutman, D. A., Michopoulos, V., Zhang, X., Nair, G., Hu, X.,
Wilson, M. E., & Sanchez, M. M. (2014). Social subordination stress and sero-
tonin transporter polymorphisms: Associations with brain white matter tract
integrity and behavior in juvenile female macaques. Cerebral Cortex, 24(12),
3334e3349. https://doi.org/10.1093/cercor/bht187
Kalin, N. H., Larson, C., Shelton, S. E., & Davidson, R. J. (1998). Asymmetric frontal
brain activity, cortisol, and behavior associated with fearful temperament in
rhesus monkeys. Behavioral Neuroscience, 112(2), 286e292. https://doi.org/
10.1037/0735-7044.112.2.286
Kalin, N. H., & Shelton, S. E. (1989). Defensive behaviors in infant rhesus monkeys:
Environmental cues and neurochemical regulation. Science, 243(4899),
1718 e1721. https://doi.org/10.1126/science.2564702
Kalin, N. H., Shelton, S. E., Rickman, M., & Davidson, R. J. (1998). Individual differ-
ences in freezing and cortisol in infant and mother rhesus monkeys. Behavioral
Neuroscience, 112(1), 251e254. https://doi.org/10.1037//0735-7044.112.1.251
Kuznetsova, A., Brockhoff, P. B., & Christensen, R. H. B. (2017). lmerTest package:
Tests in linear mixed effects models. Journal of Statistical Software, 82(13), 1e26.
https://doi.org/10.18637/jss.v082.i13
Lande, R., & Arnold, S. J. (1983). The measurement of selection on correlated
characters. Evolution, 37(6), 1226. https://doi.org/10.2307/2408842
Lehoczki, F., Szamosv€
olgyi, Z., Mikl
osi,
A., & Farag
o, T. (2019). Dogs' sensitivity to
strange pup separation calls: Pitch instability increases attention regardless of
sex and experience. Animal Behaviour, 153,115e129. https://doi.org/10.1016/
J.ANBEHAV.2019.05.010
J. W. Schwartz et al. / Animal Behaviour 190 (2022) 125e138 137
Lenth, R. V. (2016). Least-squares means: The R package lsmeans. Journal of Sta-
tistical Software, 69(1), 1e33. https://doi.org/10.18637/jss.v069.i01
Liao, D. A., Zhang, Y. S., Cai, L. X., & Ghazanfar, A. A. (2018). Internal states and
extrinsic factors both determine monkey vocal production. Proceedings of the
National Academy of Sciences of the United States of America, 115(15), 3978e3983.
https://doi.org/10.1073/pnas.1722426115
Lingle, S., Wyman, M. T., Kotrba, R., Teichroeb, L. J., & Romanow, C. A. (2012). What
makes a cry a cry? A review of infant distress vocalizations. Current Zoology,
58(5), 698e726. https://doi.org/10.1093/czoolo/58.5.698
Linhart, P., Ratcliffe, V. F., Reby, D., &
Spinka, M. (2015). Expression of emotional
arousal in two different piglet call types. PLoS One, 10(8), 1e13. https://doi.org/
10.1371/journal.pone.0135414
Mayr, E. (1961). Cause and effect in biology: Kinds of causes, predictability, and
teleology are viewed by a practicing biologist. Science, 134(3489), 1501e1506.
https://doi.org/10.1126/science.134.3489.1501
McCormack, K., Newman, T. K., Higley, J. D., Maestripieri, D., & Sanchez, M. M.
(2009). Serotonin transporter gene variation, infant abuse, and responsiveness
to stress in rhesus macaque mothers and infants. Hormones and Behavior, 55(4),
538e547. https://doi.org/10.1016/j.yhbeh.2009.01.009
Mendl, M., Burman, O. H. P., & Paul, E. S. (2010). An integrative and functional
framework for the study of animal emotion and mood. Proceedings of the Royal
Society B: Biological Sciences, 277(1696), 2895e2904. https://doi.org/10.1098/
rspb.2010.0303
Owren, M. J. (2008). GSU Praat tools: Scripts for modifying and analyzing sounds
using Praat acoustics software. Behavior Research Methods, 40(3), 822e829.
https://doi.org/10.3758/BRM.40.3.822
Owren, M. J., Amoss, R. T., & Rendall, D. (2011). Two organizing principles of vocal
production: Implications for nonhuman and human primates. American Journal
of Primatology, 73(6), 530e544. https://doi.org/10.1002/ajp.20913
Patel, E. R., & Owren, M. J. (2007). Acoustics and behavioral contexts of ‘gecker’vo-
calizations in young rhesus macaques (Macaca mulatta). Journal of the Acoustical
Soc iet y of Ame ric a, 121(1), 575e585. https://doi.org/10.1121/1.2390662
Pincus, M. (2018). Impact of stress, pubertal tempo, and diet onneurobehavioral develop-
ment in femalerhesus macaques (Ph.D.dissertation). Atlanta, GA: Emory University.
Retrieve d from: https://etd.library.emory.edu/concern/etds/8c97kq41f?locale¼en.
Price, T., Wadewitz, P., Cheney, D., Seyfarth, R., Hammerschmidt, K., & Fischer, J.
(2015). Vervets revisited: A quantitative analysis of alarm call structure and
context specificity. Scientific Reports, 5,1e11. https://doi.org/10.1038/srep13220
Puppe, B., Sch€
on, P. C., Tuchscherer, A., & Manteuffel, G. (2005). Castration-induced
vocalisation in domestic piglets, Sus scrofa: Complex and specific alterations of
the vocal quality. Applied Animal Behaviour Science, 95(1e2), 67e78. https://
doi.org/10.1016/j.applanim.2005.05.001
Ralph, C. R., & Tilbrook, A. J. (2016). Invited review: The usefulness of measuring
glucocorticoids for assessing animal welfare. Journal of Animal Science, 94(2),
457e470. https://doi.org/10.2527/jas2015-9645
Rendall, D. (2003). Acoustic correlates of caller identity and affect intensity in the
vowel-like grunt vocalizations of baboons. Journal of the Acoustical Society of
America, 113(6), 3390e3402. https://doi.org/10.1121/1.1568942
Romero, L. M. (2004). Physiological stress in ecology: Lessons from biomedical
research. Trends in Ecology &Evolution, 19(5), 249e255. https://doi.org/10.1016/
J.TREE.2004.03.008
Sanchez, M. M., Mccormack, K., Grand, A. P., Fulks, R., Graff, A., & Maestripieri, D.
(2010). Effects of sex and early maternal abuse on adrenocorticotropin hormone
and cortisol responses to the corticotropin-releasing hormone challenge during
the first 3 years of life in group-living rhesus monkeys. Development and Psy-
chopathology, 22(1), 45e53. https://doi.org/10.1017/S0954579409990253
Sanchez, M. M., McCormack, K. M., & Howell, B. R. (2015). Social buffering of stress
responses in nonhuman primates: Maternal regulation of the development of
emotional regulatory brain circuits. Social Neuroscience, 10(5), 512e526. https://
doi.org/10.1080/17470919.2015.1087426
Sapolsky, R. M. (1982). The endocrine stress-response and social status in the wild
baboon. Hormones and Behavior, 16(3), 279e292. https://doi.org/10.1016/0018-
506X(82)90027-7
Schamberg, I., Wittig, R. M., & Crockford, C. (2018). Call type signals caller goal: A
new take on ultimate and proximate influences in vocal production. Biological
Reviews, 93(4), 2071e2082. https://doi.org/10.1111/brv.12437
Scherer, K. R. (1986). Vocal affect expression: A review and a model for future
research. Psychological Bulletin, 99(2), 143e165. https://doi.org/10.1037/0033-
2909.99.2.143
Schielzeth, H. (2010). Simple means to improve the interpretability of regression
coefficients. Methods in Ecology and Evolution, 1(2), 103e113. https://doi.org/
10.1111/j.2041-210X.2010.00012.x
Schielzeth, H., & Forstmeier, W. (2009). Conclusions beyond support: Overconfident
estimates in mixed models. Behavioral Ecology, 20(2), 416e420. https://doi.org/
10.1093/beheco/arn145
Schrader, L., & Todt, D. (1998). Vocal quality is correlated with levels of stress
hormones in domestic pigs. Ethology, 104(10), 859e876. https://doi.org/10.1111/
j.1439-0310.1998.tb00036.x
Schwartz, J. W., Engelberg, J. W. M., & Gouzoules, H. (2020). Evolving views on
cognition in animal vocal communication: Contributions from scream research.
Animal Behavior and Cognition, 7(2), 192e213. https://doi.org/10.26451/
abc.07.02.12.2020
S
ebe, F., Poindron, P., Andanson, S., Chandeze, H., Delval, E., Despres, G.,
Toporenko, G., Bickell, S., & Boissy, A. (2012). Bio-acoustic analyses to assess
emotion in animals: Acoustic patterns are linked to behavioural, cardiac and
hormonal responses of ewes to the separation from their lambs. Bioacoustics,
21(1), 54.
Seyfarth, R. M., & Cheney, D. L. (2003). Meaning and emotion in animal vocaliza-
tions. Annals of the New York Academy of Sciences, 1000,32e55. https://doi.org/
10.1196/annals.1280.004
Siebert, K., Langbein, J., Sch€
on, P. C., Tuchscherer, A., & Puppe, B. (2011). Degree of
social isolation affects behavioural and vocal response patterns in dwarf goats
(Capra hircus). Applied Animal Behaviour Science, 131(1e2), 53e62. https://
doi.org/10.1016/j.applanim.2011.01.003
Sugiura, H. (2007). Effects of proximity and behavioral context on acoustic variation
in the coo calls of Japanese macaques. American Journal of Primatology, 69,
1412e142 4. https://doi.org/10.1002/ajp.20447
Szipl, G., Ringler, E., Spreafico, M., & Bugnyar, T. (2017). Calls during agonistic in-
teractions vary with arousal and raise audience attention in ravens. Frontiers in
Zoology, 14(1), 1e13. https://doi.org/10.1186/s12983-017-0244-7
Thierry, B. (2010). The macaques: A double-layered social organization. In
C. J. Campbell, A. Fuentes, K. C. MacKinnon, M. Panger, & S. K. Bearder (Eds.),
Primates in perspective (pp. 229e241). Oxford University Press.
Titze, I. R. (1994). Principles of voice production. Englewood Cliffs, NJ: Prentice Hall.
Wadewitz, P., Hammerschmidt, K., Battaglia, D., Witt, A., Wolf, F., & Fischer, J. (2015).
Characterizing vocal repertoires eHard vs. soft classification approaches. PLoS
One, 10(4), Article e0125785. https://doi.org/10.1371/journal.pone.0125785
Wheeler, B. C., & Fischer, J. (2012). Functionally referential signals: A promising
paradigm whose time has passed. Evolutionary Anthropology, 21(5), 195e205.
https://doi.org/10.1002/evan.21319
Wilson, M. E., Bounar, S., Godfrey, J., Michopoulos, V., Higgins, M., & Sanchez, M.
(2013). Social and emotional predictors of the tempo of puberty in female
rhesus monkeys. Psychoneuroendocrinology, 38(1), 67e83. https://doi.org/
10.1016/j.psyneuen.2012.04.021
Yamaguchi, C., Izumi, A., & Nakamura, K. (2010). Time course of vocal modulation
during isolation in common marmosets (Callithrix jacchus). American Journal of
Primatology, 72(8), 681e688. https://doi.org/10.1002/ajp.20824
Yeon, S. C., Kim, Y. K., Park, S. J., Lee, S. S., Lee, S. Y., Suh, E. H., Houpt, K. A.,
Chang, H. H., Lee, H. C., Yang, B. G., & Lee, H. J. (2011). Differences between
vocalization evoked by social stimuli in feral cats and house cats. Behavioural
Processes, 87(2), 183e189. https://doi.org/10.1016/j.beproc.2011.03.003
Zei Pollermann, B., & Archinard, M. (2002). Acoustic patterns of emotions. In
E. Keller, G. Bailly, A. Monaghan, J. Terken, & M. Huckvale (Eds.), Improvements in
speech synthesis. Cost 258: The naturalness of synthetic speech (pp. 237e245).
Hoboken, NJ: John Wiley.
Zimmermann, E., Leliveld, L. M. C., & Schehka, S. (2013). Toward the evolutionary
roots of affective prosody in human acoustic communication: A comparative
approach to mammalian voices. In E. Altenmüller, S. Schmidt, & E. Zimmermann
(Eds.), Evolution of emotional communication: From sounds in nonhuman mam-
mals to speech and music in men (pp. 117e132). Oxford University Press. https://
doi.org/10.1093/acprof:oso/9780199583560.001.0001.
J. W. Schwartz et al. / Animal Behaviour 190 (2022) 125e138138
