Vocal Communication of Wild Crested Macaque (Macaca nigra).

VOCAL COMMUNICATION OF WILD CRESTED
MACAQUES (Macaca nigra)

MARIA ROSDALIMA PANGGUR

GRADUATE SCHOOL
INSTITUT PERTANIAN BOGOR
BOGOR
2013

STATEMENT LETTER
I hereby declare that this thesis entitled Vocal Communication of Macaca
nigra is the original result of my own research under supervision of an advisory
committee and has never been submitted in any form to any other institution
before. All information from other authors cited here are mentioned in the text and
listed in the reference list at the end of the thesis.

Bogor, 4 February 2014

Maria Rosdalima Panggur
Student ID G352110191

ABSTRACT
MARIA ROSDALIMA PANGGUR. Vocal Communication of Wild Crested
Macaque (Macaca nigra). Supervised by RR DYAH PERWITASARI and
ANTJE ENGELHARDT.
Macaca nigra is one of seven Sulawesi macaques’ species, endemic to
northern part of Sulawesi. The species is one representative of highly tolerant
species of macaques that has received less attention than low tolerance species
(e.g. rhesus macaque, M. mulatta). This study presents an overview over the vocal
repertoire of wild crested macaques (Macaca nigra) in order to get a
comprehensive picture of their communication. The calls were categorized
according to the social context in which they were uttered. Each call types within
context are presented in the form of spectograms along with descriptive statistics
of the different acoustic parameters measured. M. nigra issued 11 call types which
were distributed in five distinct social contexts and one loud call. The five social
contexts in which the calls were uttered were affiliation, cohesion (including
group movement), agonism, predation, and mating contexts. Loud calls were
emitted only by males and in various contexts including non-social context. Calls
were highly graded, meaning that measures of acoustic parameters overlapped
between different contexts. The behavioural contexts and the use of other

communication mode can be considered as a source of variation.
Key words: Crested macaque, Macaca nigra, vocal repertoire, social context

ABSTRAK
MARIA ROSDALIMA PANGGUR. Komunikasi Suara pada Macaca nigra.
Dibimbing oleh RR DYAH PERWITASARI and ANTJE ENGELHARDT.
Macaca nigra merupakan salah satu spesies dari genus Macaca endemik
Sulawesi yang menghuni pulau Sulawesi bagian utara. Spesies ini merupakan
perwakilan dari kelompok spesies dengan tingkat toleransi sosial tinggi, yang
belum banyak diteliti jika dibandingkan dengan spesies dari kelompok dengan
toleransi rendah (misalnya monyet rhesus, M. mulatta). Studi ini menyediakan
data awal tentang komunikasi suara pada monyet hitam Sulawesi di alam untuk
mendapatkan gambaran yang komprehensif tentang komunikasi sosialnya yang
kompleks. Suara-suara yang dikeluarkan monyet tersebut dikelompokkan
berdasarkan konteks sosial saat suara tersebut dikeluarkan. Setiap tipe suara dalam
masing-masing konteks digambarkan dalam spektogram yang disertai dengan
hasil analisis dekripsi dari parameter akustik yang diukur. M. nigra mengeluarkan
11 tipe suara yang berbeda yang terbagi dalam lima konteks sosial yaitu afiliasi,
kohesi kelompok (termasuk pergerakan kelompok), agonistik, predasi, suara

terkait perilaku kawin, dan satu tipe suara loud call. Tipe suara loud call
merupakan suara yang dikeluarkan dalam beberapa konteks yang berbeda. Suara
M. nigra tergolong graded, yang berarti bahwa nilai parameter akustik suara-suara
saling tumpang tindih antara konteks. Perbedaan konteks sosial dan penggunaan
komunikasi multimodal dapat menjadi penyebab variasi suara-suara tersebut.
Kata kunci: Crested macaque, konteks sosial, spektogram, vokal repertoire,
graded

SUMMARY
MARIA ROSDALIMA PANGGUR. Vocal Communication of Wild Crested
Macaque (Macaca nigra). Supervised by RR DYAH PERWITASARI and
ANTJE ENGELHARDT.
Communication is an important aspect in the life of primates. In general,
communication in primates involves the visual, auditory, and olfactory system.
However, the vocal communication is an effective mode of communication since
it may reach further than any other mode, especially for primates living in the
forest habitat. Communication has important roles for the social life of primates
such as individual or group identity, providing information of the presence of food
and predators, and facilitating social interactions in primate social groups.
This research was carried out on wild Macaca nigra, which is one of the

seven macaque species endemic to Sulawesi. The species is characterised by a
high level of social tolerance. Vocal communication studies on tolerant species
are very limited compared to other, more despotic macaque species. The purpose
of this study was to determine the variation of vocalizations of M. nigra and the
social contexts in which the calls are issued. This study aims at providing a
comprehensive overview of the complex social communication of these macaques
and indirectly, contributes to the conservation and management of it, both in the
wild and in captivity.
A total of 1369 call units were generated from 328 sequences of calls. The
calls were categorized based on their acoustic structure and the contexts in which
they were uttered. Descriptive statistics of the acoustic parameter measurements
are presented.
This study found 11 call types emitted by M. nigra. Based on the physical
structure of acoustics, the calls are categorized as tonal calls, non-tonal calls and
complex calls (combination of tonal and non-tonal parts). M. nigra emitted the
calls in all five main social contexts: affiliation, group cohesion (incl. group
movement), agonism, predation and mating behaviour, whereas the loud call was
uttered in several different contexts including non-social one. The calls were
emitted by all individual, except the calls during mating behaviour and loud calls.
Females and males had distinct call structure and calling time during mating

behaviour. The loud call was issued only by adult males.
One of the vocalizations can be considered as discrete signals while the
others calls show high degrees of grading into each other. The discrete signals
include calls in the group cohesion. The graded call signals were issued in the
context of affiliation, agonism, the presence of predators and calls related to
sexual behaviour. The ability of receiver to assess various clues and the usage of
multimodal by the sender would influence the individual’s response to the graded
signals.
As a tolerant species, M. nigra showed high variation of calls compared to
other monkey species. The call variation differences were likely influenced by
various factors such as the character of species, behaviour and habitats. Vocal
communication data on the species M. nigra can complement the data supporting
the taxonomic determination and conservation activities of the species. The calls

might give information regarding the emotional state of animals, which is
important in ecotourism activities and captive animal management.
Key words: Macaca nigra, vocalization, graded, social context, conservation,
species management

RINGKASAN

MARIA ROSDALIMA PANGGUR. Komunikasi Suara Pada Monyet Hitam
Sulawesi (Macaca nigra). Dibimbing oleh RR DYAH PERWITASARI dan
ANTJE ENGELHARDT.
Komunikasi merupakan salah satu aspek penting dalam kehidupan primata.
Secara umum komunikasi pada primata melibatkan sistem visual, auditori, dan
olfaktori. Namun, komunikasi dengan suara merupakan moda komunikasi yang
efektif terutama bagi primata yang hidup di habitat hutan. Komunikasi memiliki
fungsi penting bagi kehidupan sosial primata seperti penanda identitas individu
dan kelompok, informasi keberadaan makanan dan predator, dan memfasilitasi
interaksi sosial dalam kelompok sosial primata.
Penelitian ini dilakukan pada Macaca nigra liar yang merupakan salah satu
primata endemik Pulau Sulawesi. Spesies ini termasuk dalam kelompok dengan
tingkat toleransi sosial yang tinggi. Belum banyak penelitian tentang ukuran
variasi vokalisasi pada spesies toleran jika dibandingkan dengan spesies Macaca
dengan toleransi sosial yang rendah.
Tujuan penelitian ini adalah untuk mengetahui variasi vokalisasi M. nigra
dan konteks sosial saat suara tersebut dikeluarkan. Penelitian ini diharapkan dapat
memberi gambaran yang komprehensif tentang komunikasi sosial M. nigra yang
kompleks dan secara tidak langsung berkontribusi pada konservasi dan
managemen spesies baik di alam maupun di penangkaran.

Jumlah unit suara yang diperoleh dalam penelitian ini adalah 1369 unit yang
dihasilkan dari 328 sekuens suara. Suara dibedakan satu sama lain berdasarkan
struktur akustik yang dapat terlihat pada spektogram. Hasil pengukuran parameter
akustik disajikan dalam analisis statistik deskriptif.
Hasil penelitian ini menemukan 11 tipe suara yang dikeluarkan oleh M.
nigra. Berdasarkan struktur fisik akustik, suara M. nigra dibedakan atas suara
tonal, non-tonal dan suara kompleks (gabungan antara tonal dan non-tonal). M.
nigra bersuara pada lima konteks utama yaitu konteks afiliasi, kohesi (termasuk
pergerakan kelompok), agonistik, predasi dan perilaku kawin. Sedangkan loud
call merupakan suara yang dikeluarkan dalam beberapa konteks yang berbeda.
Semua suara yang diteliti dikeluarkan oleh semua individu pada semua kelompok
jenis kelamin kecuali suara pada perilaku kawin dan loud call. Jantan dan betina
mengeluarkan suara terkait perilaku kawin yang berbeda secara struktur akustik
dan waktu bersuara. Loud call hanya dikeluarkan oleh jantan dewasa.
Suara yang dikeluarkan terbagi atas sinyal diskret dan graded. Kelompok
suara diskret sinyal meliputi suara dalam konteks kohesi kelompok (termasuk
pergerakan kelompok). Sinyal suara yang termasuk kelompok graded meliputi
suara-suara yang dikeluarkan dalam konteks afiliasi, agonistik, kehadiran predator
dan suara terkait perilaku kawin. Spesies ini mampu membedakan suara graded
dan bereaksi sesuai dengan konteks interaksi. Hal ini dipengaruhi kemampuan

penerima sinyal untuk memahami berbagai petunjuk sinyal yang dikirimkan oleh
pemberi sinyal. Selain itu, pemberi sinyal menggunakan lebih dari satu moda
komunikasi untuk memperkuat maksud sinyal yang disampaikan.

Sebagai anggota kelompok monyet toleran, M. nigra memiliki variasi suara
yang lebih beragam dibandingkan dengan spesies monyet lain. Perbedaan variasi
suara dengan spesies dari genus Macaca yang lain kemungkinan dipengaruhi
berbagai faktor antara lain karakter sosial spesies, perilaku dan habitat.
Data vokalisasi M. nigra dapat melengkapi data penunjang taksonomi dan
kegiatan konservasi. Suara yang juga merupakan gambaran kondisi emosi satwa
penting dalam kegiatan ekowisata dan manajemen di penangkaran yang berbasis
pada kenyamanan satwa.
Kata kunci: Macaca nigra, vokalisasi, graded, konteks sosial, konservasi,
manajemen satwa

Copyright © 2013 Bogor Agriculture University
Copyright are protected by law,

It is prohibited to cite all or part of this thesis or dissertation without referring to
and mentioning the sources. Citation only permitted for the sake of education,

research, scientific writing, report writing, critical writing or reviewing scientific
problem. Citation does not inflict the name and honour of Bogor Agriculture
University.
It is prohibited to publish or to reproduce all or part of thesis/ dissertation
without the written permission from Bogor Agriculture University.

VOCAL COMMUNICATION OF WILD CRESTED
MACAQUES (Macaca nigra)

MARIA ROSDALIMA PANGGUR

Thesis
as a partial fulfilment of the requirements for a Master
Degree in Animal Bioscience Master Program of Graduate
School of Bogor Agriculture University

GRADUATE SCHOOL
INSTITUT PERTANIAN BOGOR
BOGOR
2013

Examiner: Dr Ir Entang Iskandar, MSi

Title
Name
NIM

: Vocal Communication of Wild Crested Macaques (Macaca nigra)
: Maria Rosdalima Panggur
: G35100191

Certified by,
Advisory Committee

Dr Ir RR Dyah Perwitasari, MSc
Chairman

Dr Antje Engelhardt
Member

Acknowledged by,

Coordinator Major of
Animal Bioscience

Dr Ir RR Dyah Perwitasari, MSc

Examination date:

Dean of Graduate School

Dr Ir Dahrul Syah, MSc. Agr

Graduation date:

PREFACE
I praise and grateful to Almighty God over the completion of this scientific
work. This study is part of research collaboration between Macaca Nigra Project,
German Primate Center and Bogor Agricultural University. The study was
conducted from April 2010 to April 2011. This research topic is part of the study
on Social Communication of Macaca nigra by Dr. Jerome Micheletta.
I thank profusely to Dr. Ir. Dyah Perwitasari, MSc and Dr. Antje Engelhardt
as the supervising committee, which have supervised and put so much trust to me
during my master study and the completion of the thesis. Study funding, research
and acoustic training in Germany (April-June 2012) was provided by Dr. Antje
Engelhardt from Macaca Nigra Project.
I sincerely thank to Dr. Ir. Entang Iskandar MSi as the thesis defence
examiner and to Dr. Rika Raffiudin from Biology Department for insightful
comments and correction of the thesis.
I specially thank to Dr. Jerome Micheletta for the invaluable assistance
during field data collection and writing of the thesis. Thank you for the patience,
knowledge and experience that you shared, as well as the ideas and inspiration
that had raised my interest on animal communication. I thank Dr. Kurt
Hammerschmidt for the help and guidance in the analysis of acoustic and
statistical analysis. I knew nothing when I first met you. And thank to other
scientists Dr. Christof Neuman, Dr. Julie Dubosq, Daphne Kerhoas for numerous
help and inspiration. I also would like to thank to all the lectures in Animal
Bioscience Master Program, who at the end shaping my knowledge up through
their high dedication into the research and teaching.
I also would like to thank BKSDA Manado for allowing me conducted the
study and to Dr. Muhammad Agil, Edith Sabara, Siti Aminah (IPB) for all the
help. This study would not have succeeded without the technical assistance and
moral support by the staff of Macaca Nigra Project (Stephen M. Lentey, the late
Benediktus Giyarto, Mathilde Chanvin, Meldy Tamengge, Irawan Halir) and all
Macaca Nigra research team. The endless gratitude is addressed to my parents,
brothers and sister who have supported me by their own way during the study.
The least, but very important, I would like to thank to all my friends; class of BSH
2011, BSH students, Andi Darmawan, Rahayu Oktaviani, Andy N Cahyana, Yuni
Fanggi Tasik, Yane Moy, Ican Hakim and Sasi Kirono.
Hopefully this scientific work will be useful.

Bogor, 4 February 2014

Maria Rosdalima Panggur

CONTENT LIST
CONTENT LIST
LIST OF TABLES
LIST OF FIGURES
1 INTRODUCTION
2 LITERATURE REVIEW
Mechanism of Vocal Production of Primates
Sound Transmission
The Constraint of Sound Propagation in the Field
Sound Analysis
3 METHODS
Study Site and Time
Behavioural Observation and Recording Methods
Acoustic Analysis
Statistical Analysis
4 RESULTS AND DISCUSSION
Results
Acoustic Measurement of Call Type
Call Variation per Context
1. Affiliation
2. Group Cohesion
3. Agonism
4. Mating
5. Predation
6. Others Category: Loud call
Discussion
Discrete and Graded Signal
Comparison to Other Macaque Species
Implication to Taxonomy and Conservation Issues
5 CONCLUSIONS
REFERENCES
APPENDICES
AUTHOR’S BIOGRAPHY

v
vi
vi
1
2
2
3
3
4
5
5
5
6
7
8
8
8
9
9
12
14
15
16
16
17
17
17
18
20
20
25
27

LIST OF TABLES
Table 1 Detail description of acoustic parameters measured in the analysis
Table 2 Description of social contexts in which the calls were emitted
Table 4 The comparison of Macaque’s vocal communication.

6
9
19

LIST OF FIGURES
Figure 1 The map of study site of Macaca Nigra Project in TangkokoBatuangus NR and Dua Saudara NR, North Sulawesi, Indonesia
Figure 2 Spectograms and parameters of call
Figure 3 Spectograms of calls used in the affiliation context
Figure 4 Spectograms of calls used in the group cohesion and movement
context
Figure 5 Spectograms of calls used in the agonistic context
Figure 6 Spectograms of calls used in sexual context
Figure 7 Spectograms of calls used in the predation context
Figure 8 Spectograms of loud calls

5
7
13
13
15
16
16
16

LIST OF APPENDICES
Appendix 1 The social tolerance among macaque’s society as adapted from
Thierry (2007)
25
Appendix 2 Classification result of discriminant function analysis; main context
as grouping variable.
26

1

1 INTRODUCTION
Most primates live within complex social systems. Primates mainly use the
visual (gestural and facial) and vocal communication channel in their
communication system. Cercophithecinae uses vocalizations more widely since
most of them live in dense forest habitats that often limit the transmission of
visual signals (Altmann 1967). The calls are given by monkeys in social contexts
such advertising individual identity, alerting con-specific about predators,
maintaining cohesion during group movement, signalling the presence of food and
facilitating social interaction (Cheney and Seyfarth 2010). It can thus be noted that
vocal communication played a key role in the evolution of primate social
behaviour (McComb and Semple 2005).
The crested macaque (Macaca nigra) is one of seven endemic macaque
species that inhabit the island of Sulawesi (Fooden 1969). Similarly to the other
Sulawesi macaques, crested macaques are a socially tolerant species where
interactions among group members are weakly constrained by kinship and
dominance hierarchy (Thierry et al. 2004, Duboscq et al. 2013). This social
tolerance is thought to have led to the evolution of a larger communicative
repertoire than more despotic species. In a meta-analysis carried out on several
macaque species, Dobson (2012) showed that increased social tolerance correlates
with a large repertoire of facial expressions. A similar study carried out by
Maestripieri (2005) additionally suggests that species with increased social
tolerance also have a larger repertoire of communicative gestures. Freeberg et al.
(2012) proposed a hypothesis that social communication complexity increases
alongside social complexity. The social complexity (group size and time spent
grooming) is likely to increase vocal complexity (McComb and Semple 2005).
Since the social tolerance might play role in social complexity, the study therefore
suggests that in a species as highly tolerant as the crested macaques individuals
will also display a high degree of vocal repertoire.
Research on vocalization has been carried out in many species of macaques
such as M. fuscata (Green 1975), M. silenus (Hohmann and Herzog 1985), M.
radiata (Hohmann 1989), M. fascicularis (Palombit 1992), M. sylvanus (Fischer
and Hammerschmidt 2002) and M. cyclopis (Hsu et al. 2005). However, most
available data are from less tolerant species, and those carried out on Sulawesi
macaques (M. nigra: Lewis 1985; M. tonkeana: Masataka and Thierry 1993) were
conducted in captivity where the full extent of a species repertoire might not be
expressed. The only reports on the vocalisation repertoire of wild Sulawesi
macaques are only descriptive and anecdotal (Thierry et al. 2000). No quantitative
descriptions on the vocal patterns of any of the highly tolerant Sulawesi macaques
are so far available. Such a study is therefore urgently needed in order to get a
more comprehensive picture of the relationship between social tolerance and the
complexity of communication in macaques.
Since 2008, IUCN red-list listed crested macaques as Critically Endangered
(Supriatna and Andayani 2008), which means that the species is currently facing
an extremely high risk of extinction. The decline of the population is mainly
caused by human disturbance such as land clearing and hunting for bush meat and
pets (Sugardjito et al. 1989; Rosenbaum et al. 1998; Palacios et al. 2011). This

2
human disturbance in general raises animal anxiety in the presence of humans,
which in turn leads to difficulties in monitoring wild individuals. Since at least
certain vocalisations are species specific (Hartwig 2005, Rupell 2009) and
individual specific (Bouchet et al. 2012, Neumann et al. 2010, Geissman and
Nijman 2006), knowing a species repertoire and an individual’s call pattern will
help population monitoring activities for conservation purposes. It has been
demonstrated by Adi et al. (2009) that a population census in Emberiza hortulana
using vocal recognition can predict the population density and structure simpler
and more accurate as compared to other population monitoring methods.
Furthermore, different types of vocalisations might represent the internal emotion
of animals either positive state (Boissy et al. 2007) or negative state (Dupjan et al.
2008, Fichtel et al. 2001), so the understanding of macaque's vocalization will be
useful for the species’ management and to ensure their welfare.
The aim of this study therefore is to provide quantitative data of crested
macaque vocalizations to get a better understanding about the communication
system of this species. The study also attempt to categorize calls based on the
context in which they were emitted and examine whether acoustic features overlap
within and between calls in different contexts to provide an estimation of the
graded nature of the crested macaque vocal communication system. In addition to
support species management and conservation, this research will provide a key of
future studies on the vocal communication of Sulawesi macaques.

2 LITERATURE REVIEW
Mechanism of Vocal Production of Primates
Vocal production of mammals employs several organs such as the lungs,
larynx, and vocal tract (Fitch and Hauser 1995). Lungs are the source of power;
they convert the steady air into a series of puff air. The larynx then modulates air
pressure to become a sound source through the opening and closing process of the
larynx valve. The larynx valve is under the control of the vocal folds (vocal
cords). There are two major mechanisms of vocal sound production: phonation
and impulsive or noisy sound. Phonation occurs when the vocal folds vibration
through larynx produces repetitive series of puff air in a certain level of
frequency, which is the fundamental frequency (Figure 2). Impulsive (cough and
clicks) or noisy sounds (hisses and whispers) are busted out by the high pressure
of air that forces the opening of the larynx. The variation of primate vocalizations
results from the phonation of the sound source. The sound source then flows
through the supralaryngeal vocal tract on its way out to the lips and nostrils. This
organ includes the pharyngeal, oral, and nasal cavities, which serve to resonate the
sound source. The differences in the length and shape of the human vocal tract are
known to cause the difference of resonance frequency (known as formants in
human, Fitch and Hauser 1995). Non-human primates have a longer and thinner
tongue, larynges that are positioned higher in the neck, and there is a relative lack
of flexible soft tissues in the supralaryngeal cavities compared to humans (Owren

3
and Linker 1995). These differences have presumably limited the ability of vocal
variation in non-human primate.
The air puff from the larynx produces sound, which can be either tonal or
noisy sound. Tonal sound is characterized by one excitation frequency, which is
sometimes followed by harmonics or overtones. The harmonics are integer
multiple of the fundamental frequency. While noisy sound is a periodic signal,
which may also be produced at the vocal folds as a result of air turbulence. Noisy
call can be produce either at the larynx or above the larynx (due to constriction
elsewhere in the vocal tract, Fitch and Hauser 1995). The combination of tonal
and non-tonal call can be called complex calls (Micheletta 2012).
Sound Transmission
The sound waves are only able to propagate through a medium. The sound
waves that travel through gasses and liquid are called longitudinal waves while
sound waves that transmit through solid medium are called transversal waves
(Bradbury and Vehrencamp 1998). Hence the vocal exchange in primate is a
longitudinal sound wave that is transmitted through air. The sound wave
transmission is simply described as the following explanation; when there is no
vibration from any sound source, the air pressure is called ambient pressure. The
vibration of the sound source increases the collision of molecules in the air. When
the molecules move and stay away due to the collision, the air pressure is below
ambient pressure and increase again when the next tremor occurs. The state when
the molecule assembled is called compression while the state when the molecule
expanded is called rarefaction. The temporal pattern of rise and fall in signal
amplitude is called the waveform of the signal. The largest distance of air pressure
toward ambient pressure is called amplitude. The distance between two successive
peaks of high pressure called wavelength. The number of cycles (one rise and fall
of the signal) per second is called frequency. The unit of frequency, cycles/second
is called Hertz. The time is needed by a cycle is called period (Bradbury and
Vehrencamp 1998).
The Constraint of Sound Propagation in the Field
Acoustic transmission can be constrained by two fundamental problems:
attenuation and degradation, which affect signal quality. Attenuation can be
caused by the atmospheric absorption, ground attenuation, scattering of a sound
beam, and deflection of sound by stratified media (Wiley and Richards 1978).
Signal attenuation in the dense forest is relatively severer than in open habitat. It
also occurs over long distance and low transmission height (Maciej et al. 2011).
Sound degradation results from irregular amplitude fluctuation (e.g. atmospheric
turbulence of the wind) and reverberation (e.g. vegetation). In an open habitat,
irregular amplitude fluctuation happens often and masks the low frequency of
amplitude modulation. In the forest habitat, reverberation is more severe than in
open habitat and masks the transmission of high frequencies. Both types of
degradation affect more to the temporal pattern of amplitude or intensity
modulation than frequency modulation (Wiley and Richards 1978)

4
The constraints of sound wave transmission in the field can affect the
quality of sound recordings (Maciej et al. 2011). Signal quality can also be
masked by background noise such as the sound of other animals or observer.
Large distances between sound emitter and observer will potentially enhance
sound degradation. Some simple efforts such as pointing the microphone towards
to the sound source within close proximity, can overcome the problem.
Depending on the call, e.g. low amplitude calls like grunts, good sound quality
may be obtained from 3-5 m distance while high amplitude signals might be
recorded at 15 m (Fischer et al. 2013). Nevertheless, for further analysis using
computers, preliminary assessment of the reliability of recordings needs to be
done for any signals recorded in large distance (Maciej et al. 2011).
Sound Analysis
Fischer et al. (2013) provided a detailed explanation of the study of
bioacoustic in primates along with a guideline for playback experiments. Data
collected in the field is transferred into a computer for analysis. Computer
program analyses that are widely used to extract acoustic features of sound signal
are Avisoft SAS Lab (R. Specht, Berlin), RAVEN (Cornell Lab of Ornithology),
PRAAT (Institute of Phonetic Science, http://www.praat.org) or Signal
(Engineering Design, Belmont, MA). The computer-aided analysis still faces
problems related to the complexity of signal, high variation of vocalization,
environmental effect and distance and spatial orientation of the animal towards the
microphone. Using a multi-parameter approach, as provided by the software LMA
(Schrader and Hammerschmidt 1997), can reduce the importance of these
problems.
Sound is a complex signal, mostly made up of non-pure sinusoidal
waveforms. To deal with this acoustic complexity, the Fourier transformation
technique is required. The main principle of this technique is to transform any
continuous waveform into a pure sine wave. Sound signals in pure sinusoidal
form are easy to compare or to analyse their frequency, amplitude and relative
phase. In a computer program, the Fast Fourier Transform (FFT) breaks down the
signal into smaller unit samples and records the amplitude at each point. The
advantage of this technique is that sound signals can be analysed rapidly. The
disadvantage is that when the number of signal points is too small, the computer
will have difficulties to determine the frequency of the signal. This problem can
be solved by applying the Nyquist-Shanon equation in which the sampling rate of
recording equipment is set at twice the frequency of the signal frequency
(Bradburry and Vehrencamp 1998). The sampling rate for most mammal vocal
recordings can be set at 40 kHz, which is twice the maximum capability of human
hearing, 20 kHz. Yet, the sampling rate that is commonly used is 44.1 kHz,
parallel to the sampling rate of a signal in compact disc storage.
The acoustic signal can also be plotted in spectrograms. A spectrogram is a
graphical illustration of the distribution of signal energy along a time axis. In a
two-dimensional spectrogram, the vertical axis refers to frequency value, while
the horizontal axis refers to time (Bradburry and Vehrencamp 1998, Figure 2),
while the shading represents amplitude. Spectrograms with good resolution are
modified by the choice of frequency range, FFT-size, length of time segment, and

5
of the transformation window, which is chosen with respect to the quality of the
signal (Schrader and Hammerschmidt 1997).

3 METHODS
Study Site and Time
The research was conducted on two wild groups of crested macaques living
in the Tangkoko Nature Reserve (NR), North Sulawesi (Figure 1). The research
was carried out under the Macaca Nigra Project (research collaboration between
Bogor Agricultural University, the University Sam Ratulangi and the German
Primate Center), which has been running since 2006 (www.macaca-nigra.org).
Monkeys are well habituated to the presence of researchers. The two groups of
crested macaques, which were the focus of research, are group Rambo 1 (R1) and
group Pantai Batu (PB), each consisting of 65-75 individuals. The field data
collection was conducted from September 2010 until March 2011.

Figure 1 The map of study site of Macaca Nigra Project in Tangkoko-Batuangus
NR and Dua Saudara NR, North Sulawesi, Indonesia

Behavioural Observation and Recording Methods
During the study, the groups were followed for full days (one group per
day) from dawn, when the macaques started being active (usually at 0530-0600 h)
to dusk, when the macaques reached their sleeping site (usually at 1700-1800 h).
The data were obtained from 28 adult individuals from R1 (20 females and 8
males) and 23 individual adults from PB (18 females and 5 males). Additional
recordings were obtained from other studies: loud calls (Neumann et al. 2010) and
female copulation call of eight females (Macaca Nigra Project, unpublished data,

6
2006-2012). Ad-libitum and 30 min long focal animal sampling were used to note
the behaviour associated with call production (Altmann 1974). The focal
individual was chosen according to a pseudo-randomised order so that all adults
were followed every 2 days. The behavioural data of each focal individual were
obtained equally between three different time periods: morning (0600-1000),
midday (1000-1400) and afternoon (1400-1800). The observational data was
recorded using a Psion handheld computer with Windows mobile and Excel
(Microsoft). All social interactions were videotaped with a Panasonic HDCSD700 HD camcorder that was connected to a Sennheiser K6-ME66 directional
microphone. The sampling rate was 44.1 kHz. In this way, the video and sound
were simultaneously obtained.
Acoustic Analysis
For the acoustic analysis, each sequence of calls was inspected using
Avisoft SAS Lab Pro 5.2 (Avisoft Bioacoustics, Berlin, Germany). Call sequences
were cut off per unit call for subsequent analysis. Each selected unit calls were
free from unwanted signals like bird song, insect or other sound. Call units were
down sampled from 44025 to 11025 Hz for better frequency resolution. Fast
Fourier transform (1024-pt FFT; frame size 100%; window: Hamming; overlap
93.73%) were conducted for all the calls. The resulting time resolution was 5.8 ms
at 11 Hz of frequency resolution. All the call units were saved in the ASCII
format files for further analysis using LMA, a software tool to extract different
measures of acoustic parameters from acoustic signals (Schrader and
Hammerschmidt 1997).
An interactive macro was applied to calculate acoustic parameter of each
call unit. The cut-off frequency was set at 200 Hz to reduce the noise signal. In
this way, all frequencies ranging below 200 Hz were excluded from the analysis.
The start and end threshold was set at 20 %, meaning that all time segments in the
beginning and the end of calls that were considered for analysis were more than
20 % of maximal amplitude. Detailed information of acoustic parameter measured
in the study was presented in Table 1 and Figure 2.
Table 1 Detailed description of acoustic parameters measured in the analysis
Abbreviation
Parameters
Dur
Duration
F0 max/min

F0 mean
Pf max/min

Pf mean

Fundamental
frequency
maximum/minimum
Fundamental
frequency mean
Peak frequency
maximum/minimum
Peak frequency
mean

Unit Description
ms Time between start and the end of the
call
Hz Maximum/minimum value of the
fundamental frequency in all time
segments
Hz Average value of the fundamental
frequency in all time segments
Hz Maximum/minimum value across all
time segments of the frequency with
the highest energy
Hz Mean value across all time segments of
the frequency with the highest energy

Frequency (kHz)

7

Figure 2 Spectograms and parameters of call (a: call start, b: call end, c: duration
(ms), Fmax: Maximum frequency, Pf: Peak frequency, F0: Fundamental
frequency).
Statistical Analysis
A step-wise discriminant function analysis (DFA) was used to assign the
calls to each context by determining which acoustic variables would maximize the
differences between contexts. First, the assignment test was applied to determine
whether the calls would be reliably classified to the main contexts; affiliation,
cohesion, agonism, predation, mating and loud calls. Since there were at least two
different calls within the main contexts (named as sub-context calls), the second
test was applied to check the assignment of sub-context calls within each main
contexts. The sub-context calls refer as call types except for predation calls where
sub-context calls differentiated by the predator type. Those contexts of interaction
and call types were used as grouping variables and all acoustic parameters as
predictors. The leave-one-out procedure of cross validation was applied to check
the validation of the discrimination result by which each case is classified by the
function derived from all cases other than that case. The assignment of calls into
graded and discrete was determined by 70 % threshold of correctly assignment
value.
To test whether the acoustic structure of call parameters (duration, F0
max/min/mean, Pf max/min/mean) varied related to the context and sex, a linear
mixed model was applied. The sex and sub contexts were used as fixed factors
while identity of caller was set as random factors. The test was done to each
acoustic parameter as dependent factors. The Least Significant Differences (LSD)
was used as adjustment for multiple comparisons. For mating context, the
differences between copulation calls of males and females were compared with

8
Anova. Sex was set as factor and the acoustic parameters were used as dependent
variables. All data analysis was carried out with SPSS 20 (IBM).

4 RESULTS AND DISCUSSION
Results
A total 1369 call units were generated from 328 sequences of calls. The
analyses resulted in 11 call types distributed in five social contexts. Another
category of call, the loud calls was used in several different contexts (Table 2).
The general physical structures of calls were tonal, atonal and complex. All
individuals issued all types of calls except sexual related call and loud call.
Acoustic Measurement of Call Type
The assignment into main context categories resulted in a percentage of
correctly classified calls of 47.8 %. The calls were discriminated based on the sub
context within each main context. For example, the affiliation context consists of
two sub-contexts, which was discriminated by 61.3 % after cross validation. The
variables that significantly discriminated the two groups of sub contexts were Pf
mean, duration, F0 max, F0 min (Table 3). Descriptive statistics of the acoustic
parameter for each call type per each context is presented in the Table 4.
Linear mixed model analysis showed that call types and sex have no
significant influence on the complexity of calls in the affiliation context. While, in
group cohesion contexts, both call types and sex have significant contribution to
the complexity of calls (call types: duration, F0 max, F0 min, F0 mean; sex:
duration, F0 max, F0 mean and Pf max). The harsh coo has longer duration and
higher frequency than the low coo. In agonism context, the complexity of calls is
significantly influenced by the call types in all variables except duration. The
more intense the aggression, the higher frequency calls would be emitted. While
sex factor has significant influence only in F0 max and F0 mean. Sub contexts in
predation contributed significantly to the call complexity in all variables except Pf
mean and Pf max. The warning calls toward the python and model python were
more salient than toward human or dog. While sex has significant influences to
the calls complexity in duration, Pf mean and Pf max (Table 5). The two
copulation calls of male and female are significantly different in all variables
except Pf min, in which the male has longer duration and higher frequencies than
female copulation calls (Table 6).

9
Table 2 Description of social contexts in which the calls were emitted
No
1

Main
Contexts
Affiliation

2

Group
cohesion

3

Agonism

4

Mating

5

Predation

Other call type
6
Loud calls

Details
When group members were involve in
social positive interactions or nonaggression behaviours, either dyadic or
polyadic. The behaviour types were
approach, lip smack, expressive run which
sometimes leads to grooming behaviour.
The context was characterized by relaxing
and calm situations. The individuals gave
calls to inform their position to other
individuals and to maintain group
cohesion
The context was characterised by the
aggressive behaviour either dyadic or
polyadic. The aggressive behaviours were
attack, bite, scream, hit, grab and chase.
The aggression rates were classified into
three levels due to the aggressive
intensity, which were low agonistic rate,
medium agonistic rate and high agonistic
rate.
When individuals performed sexual
behaviour involving female parade, male
approaching and mounting.
When an individual detected the presence
of a reticulated python (Python
reticulatus), dogs (Canis familiaris), birds
of prey and humans.

Call types
1. Affiliation
calls
2. Soft grunt

3. Coo
4. Harsh/lost coo

5. High threat
6. Medium threat
7. High threat

8. Male
copulation call
9. Female
copulation call
10. Alarm call

The calls are emitted only by males, in 11. Loud call
various contexts (Neumann et al. 2010)

Call Variation per Context
1. Affiliation
Call types in this context were affiliation calls and soft grunts (see also
Thierry et al. 2000). These calls were mostly characterised by a non-tonal
structure (Figure 3a, 3b). Affiliation calls were uttered repeatedly to initiate
positive interactions and reconciliation after conflicts. Soft grunt calls have a nasal
component, and were mostly produced in affiliative interactions which involved
physical contact.

10

Table 3 The classification of calls based on the discriminant function analysis (DFA). Discriminant analysis to the loud call provided by
other study (Neumann et al. 2010)
Main Context

Call types (N)

Significant
parameter

Test
of
Function (s)

Wilks’
Lambda

Chisquare

df

P

Original
Classification

Cross-validated
Classification

1

Affiliation

Affiliation call (231)
Soft grunt (78)

Pf mean

1

0.906

10.228

1

0.001

61.3 %

61.3 %

2

Group cohesion

Coo (161)
Harsh coo (14)

F0 max

1

0.897

18.687

1

0.000

80 %

80 %

1 through 2

0.403

291.362

8

0.000

No

3

Low Threat (210)
Middle rate (52)
High rate agonistic (53)
Real python (99)
Model python a) (120)
Dog (8)
Other b) (9)

Agonism

4

5

Predation

Mating

Female copulation call (116)
Male copulation call (167)

Pf mean
F0 max
Pf min
F0 min

69.5 %
2

0.981

6.018

3

0.111

1 through 2

0.830

43.175

6

0.000

2

0.986

3.388

2

0.184

F0 max
Pf max
Pf min

1

0.820

55.514

3

0.000

Duration

1 through 4

0.436

962.736

20

0.000

F0 min

2 through 4

0.784

281.702

12

0.000

Pf max

3 through 4

0.922

94.618

6

0.000

F0 max

4

0.998

2.116

2

0.347

Pf min
duration

67.70%

54.7 %

53.4%

65.7%

65.4 %

48.5 %

47.8 %

Total

Among
contexts

main

Affiliation (309)
Group Cohesion (175)
Agonism (325)
Predation (236)
Mating (283)
Loud call (41)

Description: N= number of unit calls measured, a) The alarm calls were obtained by presenting a life-size picture of a python in a realistic setting (Micheletta et al. 2012), b) Warning call toward human and
unknown object. The significant value is p10m from the main group. By ear, the low coo and
clear coo were not easily distinguished each other. While harsh coo has higher
frequency and amplitude than the other two, which is easily recognise by
observer.
a)

b)

0

Frequency (kHz)
(kHz)
Frequency

Frequency (kHz)

10
8

8

6

6

4

4

2

2

1

2

3

4

1

5

2

Time (s)
Frequency (kHz)

10

8

8

6

6

4

4

2

2

1

2

Time (s)

4

3

d)
Frequency (kHz)

c)

0

3

Time (s)

1

2

3

4

5

6

7

8

9 10 11 12 13 14 15

Time (s)

Figure 4 Spectograms of calls used in the group cohesion and movement context;
a) low coo exchanges between two females at a distance, b) clear coo of a
female who was far from the group, c) Harsh call emitted by adult female
who was separated from the group, d) female contact call for her baby.

14
3. Agonism
There were several different types of calls in this context, which seemed to
correspond to the intensity of the aggressive interactions: low threat (Figure 5a),
medium threat (Figure 5b), and high threat (Figure 5 c-g). Due to the high
variability of acoustic features of agonistic calls, the calls fell into three main call
groups, namely pulsed calls, tonal screams and complex scream (Micheletta
2012). One aggressive interaction could comprise a sequence of many basic
patterns of calls.
Low threat calls were generally issued when initiating aggression. Mild
threat were frequently accompanied by the half-open mouth display and seemed to
prevent intervention or as a protest to avoid interaction initiated by other
individuals. The open mouth bared-teeth face accompanied high threat calls. High
threat calls were issued to initiate aggression, to counter attack and during
aggressive interventions.
a)

5

15

0

10

5

5
1

2

Time (s)1

Time (s)

20
Frequency (kHz)

b)

0

Frequency (kHz)

Frequency (kHz)

20

c)

5
0

5
1

2

3

4

Time (s)

d)

Frequency (kHz)

20
5
0
5
1

2

3

4

Time (s)

5

e)

Frequency (kHz)

20
15
10
5
1

2

3

4

Time (s)

5

6

7

15
f)
Frequency (kHz)

20
15
10
5
1

2

3

4

Time (s)

g)

Frequency (kHz)

20
5
0
5
1

2

3

Time (s)

4

5

Figure 5 Spectograms of calls used in the agonistic context; a) hard grunt in low
intensity agonistic interaction to prevent individual starting an interaction,
b) female is giving mild threat to juvenile. Spectograms of calls used in high
intensity agonistic interactions: c) complex call of a female during a
counter-attack against a male, d) complex call of a female after she received
aggression from another female, e) a female tonal call given as a counter
attack in an agonistic context with a male. Females also seek support from
other male, f) the complex call uttered by a female as a response to the
threat of a male. The caller performed a counter attack and sought support,
g) a complex call of a female during a counter attack after an aggression
initiated by a juvenile
4. Mating
Sexually-related calls consist of male copulation calls and female copulation
calls. Male copulation calls were characterized by a series of similar repetitive
squeaks (Figure 6a). Female copulation call consisted of repetitive short-pulsed
calls (Fig