Content uploaded by Thierry Aubin
Author content
All content in this area was uploaded by Thierry Aubin on Dec 15, 2017
Content may be subject to copyright.
ANIMAL BEHAVIOUR, 1999, 57, 1175–1183
Article No. anbe.1999.1086, available online at http://www.idealibrary.com on
Finding a parent in a king penguin colony: the acoustic system of
individual recognition
PIERRE JOUVENTIN*, THIERRY AUBIN† & THIERRY LENGAGNE*†
*CEBC–CNRS UPR 4701, Station de Chize´, France
†NAM–CNRS URA 1491, Universite´ Paris-Sud
(Received 4 September 1997; initial acceptance 3 December 1997;
final acceptance 13 January 1999; MS. number: 5646R)
To be fed, a king penguin, Aptenodytes patagonicus, chick must identify the call of its parents, in
the continuous background noise of the colony. To study this recognition process, we played back to the
chicks parental calls with acoustic parameters modified in the temporal and frequency domains. The
parental call is composed of syllables (complex sounds with harmonic series) separated by pronounced
amplitude declines. Our experiments with modified signals indicate that the chick’s frequency analysis of
the call is not tuned towards precise peak energy values, the signal being recognized even when the carrier
frequency was shifted 100 Hz down or 75 Hz up. To recognize the adult, chicks used frequency rather
than amplitude modulation, in particular the frequency modulation shape of the syllable. This structure
is repeated through the different syllables of the call giving a distinct vocal signature. Our experiments
also show that the receiver needs to perceive only a small part of the signal: the first half of the syllable
(0.23 s) and the first three harmonics were sufficient to elicit recognition. The small amount of
information necessary to understand the message, the high redundancy in the time and frequency
domains and the almost infinite possibilities of coding provided by the frequency modulation signature
permit the chick to recognize the adult, without the help of a nest site. For these reasons, the code used
in the call of the king penguin can be regarded as a functional code, increasing the possibility of
individual recognition in an acoustically constraining environment.
1999 The Association for the Study of Animal Behaviour
In birds, many vocal exchanges, particularly between
mates and between parents and young, occur at short
range, over distances of a few metres at most (Falls 1982).
At these short distances, the signal is only weakly modi-
fied during propagation, by, for example, the ground
effect, atmospheric absorption or geometric attenuation
(Wiley & Richards 1978;Dabelsteen 1984;Dabelsteen
et al. 1993). Nevertheless, even at short range, communi-
cation between individuals may sometimes be difficult,
for example in noisy environments, such as dense
colonies of birds. In these colonies, a continuous back-
ground noise is generated by sounds used for communi-
cation and by other sounds such as wind, waves, beak
clapping and wing flapping. The level of ambient noise is
high (more than 70 dB: Robisson 1991;Mathevon 1996)
and consequently the value of the signal-to-noise ratio is
low. In addition, the numerous vocalizations generate
jamming in both frequency and temporal domains. Thus,
it is difficult for individuals in the colony to extract
information from the background noise.
Seabird colonies are particularly crowded and noisy.
Breeding on land and feeding at sea, mates are separated
for days or weeks during the breeding season, but are
faithful to each other and to their offspring (see Jouventin
1982 for penguins). The ability to recognize mates,
parents or chicks is particularly important in seabird
colonies, where nest sites are densely packed, increasing
the possibility of confusion (Hutchison et al. 1968). To
find the egg(s) or chick(s), nesting birds also use land-
marks, so to isolate vocal recognition in this study, we
used a non-nesting species, the king penguin, Aptenodytes
patagonicus.
King penguins breed on flat areas in homogeneous and
dense monospecific colonies numbering thousands of
birds (1.6 breeders/m
2
;Guinet et al. 1995). There are no
nest sites: each bird carries its egg and then its small chick
on its feet. At the end of the breeding cycle, as in our
study, each individual is identified by its chick only by
vocal cues (the ‘long call’ in the behavioural repertoire
described by Stonehouse 1960) and a few landmarks
(Derenne et al. 1979 for the king penguin; Jouventin
Correspondence and present address: P. Jouventin, CEFE–CNRS UPR
9056, 1919 Route de Mende, F-34213 Montpellier Cedex 5, France
(email: jouventin@cefe.cnrs-mop.fr). T. Aubin is at NAM–CNRS URA
1491, Universite´ Paris-Sud, F-91400 Orsay, France.
0003–3472/99/061175+09 $30.00/0 1999 The Association for the Study of Animal Behaviour1175
1982 for the penguin family). Observations of banded
birds and playback experiments show that only the mate
responds to its partner coming back. Similarly, chicks
respond only to their parents (or their calls played back).
For the chick, recognition of the parental call is important
for survival as parents usually only feed a chick they have
identified by its call (Jouventin 1982). Nevertheless, some
adoptions do occur (Stonehouse 1960) by failed breeders
(Jouventin & Mauget 1996).
That birds can recognize one another by voice alone
has been demonstrated repeatedly. In numerous studies
(for reviews see Falls 1982;Dhondt & Lambrechts 1992),
the structure of the signal has been analysed, to
determine which parameters encode acoustic identity
and allow recognition between individuals. An ideal
signal for individual recognition would be highly
stereotyped within each individual but would differ
noticeably between individuals. A method of quantifying
this is to define a ratio of acoustic parameters such
as Cvb/Cvi, where Cvb is the coefficient of varia-
tion between individuals and Cvi the coefficient of
variation within individuals (Hutchison et al. 1968; for
a review see Scherrer 1984). The Cvb/Cvi ratio and
coefficients of correlation have been calculated for the
king penguin call (Jouventin 1982;Robisson 1992;
Lengagne et al. 1997): the between-individual variation
was less stereotyped and on average three times greater
than the within-individual variation for time parameters
and four times greater for frequency parameters. Thus,
individuals encode their call by making it highly
stereotyped.
Knowing that it is possible to distinguish the signals of
individuals statistically does not, however, tell us how
birds manage do it. Playback experiments are needed to
investigate this. Our previous observations (Jouventin
1982) and experiments (Aubin & Jouventin 1998) show
that king penguin chicks are able to decode a precise
acoustic signal despite extreme jamming and high-level
background noise generated by the colony (called the
‘cocktail-party effect’ in humans by Cherry 1966), sug-
gesting that a variety of fine details are fitted into the
code. How can penguins appreciate subtle qualities of
sounds in such a noisy and jamming environment? How
can acoustic individuality be encoded when faced by such
strong acoustical constraints?
Detection of signals in noise is a complex issue.
The most detailed knowledge of the mechanisms
involved comes from psychoacoustic experiments
performed in controlled laboratory environments (Klump
1996). Few field studies have studied experimentally
how signals of animals are detected amidst background
noise in the natural environment. The capacity of
the king penguin chick to recognize the parental
call requires peculiar strategies of coding/decoding.
Our aim in this study was to determine to which
acoustic parameters the chick is tuned in order to
extract the information from the background noise. For
this purpose, we broadcast different parental calls modi-
fied in the frequency and temporal domains, to
chicks waiting in the colony for their parents to return
from the sea.
METHODS
Subjects and Location
We studied king penguins at Ile Possession, Crozet
archipelago (4625S, 5145E) from early December 1995
to mid-January 1996. The study was conducted at Baie du
Marin, in a large colony containing about 40 000 pairs
of adults and 1500 chicks. The chicks we tested were
between 10 and 12 months old. At this stage of its life, the
chick is pushed away by new breeders from where the egg
was laid but is entirely dependent on its parents for food.
To identify them, we banded tested chicks on a flipper
with a temporary plastic band.
Recording Procedure
To record parental calls of king penguins (the ‘long
call’ of Stonehouse 1960) we used an omnidirectional
Sennheiser MD211 microphone (frequency response
150–18 000 Hz 1 dB) mounted on a 2.5-m perch and
connected to a Sony TCD10 Pro II digital audiotape
recorder (sampling frequency: 44.6 kHz, frequency
response flat within the range 20–20 000 Hz1 dB). The
distance between the beak of the recorded bird and the
microphone was approximately 1 m.
Sound Synthesis and Analysis
Signals were digitized with a 16-bit Oros Au21 acquisi-
tion card (with a 120-dB/octave antialiasing filter) at a
sampling frequency of 16 kHz and stored on the hard disk
of a PC computer. They were then examined and modi-
fied with the Syntana analytical package (Aubin 1994).
Sound pressure level measurements (SPL in dB) were
measured with a Brue¨l & Kjaer Sound Level Meter type
2235 (linear scale, slow setting) equipped with a 1-inch
condenser microphone type 4176.
Playback Procedure
For playback experiments we used a 4200 Uher tape-
recorder (tape speed 19 cm/s) connected to a 50-W Audix
PH3 self-powered loudspeaker (frequency response 100–
5600 Hz2 dB). Signals were played at a natural sound
pressure level (Robisson 1993a), of approximately 95 dB,
measured 1 m from the loudspeaker.
We conducted tests between 1000 and 1700 hours,
during clear and dry weather, with a wind speed of less
than 4 m/s. The chick was generally resting in the feeding
area, preening itself. The distance between the loud-
speaker and the bird was ca. 7 m, this corresponding to a
natural calling distance of an adult (Robisson 1993b). At
first, chicks were tested with the natural call of one of
their parents. A few chicks (1/20) that did not respond
because they had just been fed were tested the next day.
Only chicks whose intensity of response was ranked 4 (see
criteria of responses below) were kept for further tests
with modified signals. Thus, the population of chicks
tested was homogeneous in motivation to detect the
1176 ANIMAL BEHAVIOUR, 57, 6
parental call. All these chicks reacted without ambiguity.
To minimize habituation, we tested chicks with the
experimental signals 2 or 3 days after testing them with a
natural call, while their parents were absent. We broad-
cast two identical experimental signals, separated by an
interval of 5 s, to a chick in a feeding area with a normal
density of birds; then 15 min later, we broadcast another
series of two identical signals. The 15-min period between
broadcasts allowed the chick to recover its natural
activity. To prevent habituation, a maximum of three
series of signals a day was broadcast to any one chick.
Each chick was tested with all the different types of
signal. The order of presentation of the signals was
randomized for the different chicks tested. In the same
way, the order of presentation of experimental signals
from day to day was not the same for each chick. Hence,
the observed responses for the whole group of chicks
tested were neither a result of cumulative excitation nor
dependent on playback order. To avoid a possible mask-
ing effect not studied here (see Aubin & Jouventin 1998),
experimental signals were broadcast only during relative
periods of silence, that is, when birds in the vicinity of the
tested chick remained silent.
Criteria of Responses
In natural conditions, when the parents are absent, the
chick remains silent, lying quietly. The adult, returning
from the sea to feed its chick, makes its way to the area of
the colony where the chick is usually located (rendezvous
site) and calls at regular intervals. The chick in the flock
holds up its head, calls in reply and moves towards the
parent, often running (Stonehouse 1960;Jouventin
1982). The other chicks in the vicinity, resting or preen-
ing themselves, never react to the extraneous calls and
their behaviour does not change.
To evaluate the intensity of response of tested chicks to
playback signals, we used a five-point scale, as follows: 0
(none): no reaction; 1 (weak): head turning, agitation; 2
(medium): head turning and calls after the second broad-
cast; 3 (strong): head turning, calls after the first broad-
cast; and 4 (very strong): head turning, calls after the first
broadcast, approaches in the direction of the loudspeaker
and stops in the vicinity (less than 3 m). This behavioural
scale is similar to those used in previous studies on the
king penguin (see Derenne et al. 1979;Jouventin et al.
1979;Robisson 1990).
Statistical Analysis
Statistics and interpretations of results were based on
the analysis of the distribution of the observed values
within the five response classes. To compare paired
samples in more than two categories, we used the mar-
ginal homogeneity test (Agresti 1990) together with exact
two-sided Pvalues. When the same marginal distribution
was used through several comparisons, we used the
Bonferroni-corrected Pvalues to assess the final signifi-
cance of the test. Computations of exact two-sided P
values were carried out with StatXact software (Cytel
1995). We used a significance level of P≤0.05.
Experimental Signals
We tested 17 experimental signals. These consisted of
natural calls modified in the frequency and temporal
domains.
The original signal
This signal was the natural parental call specific to
the chick being tested. So, for each chick tested there
was an original signal corresponding to the call of one
of the parents (male or female) and a series of experi-
mental signals. Figures 1 and 2show an example of a
natural parental call. The call corresponds to a series
of sound components, termed syllables by Jouventin
(1982), separated by strong amplitude declines which
coincide with falls in frequency. The call duration
varied from 3 to 6 s (XSD= 4.451.16 s, N=66), the
first syllable generally being the longest. The spectral
composition of syllables corresponds to harmonic
series. Most of the energy is concentrated between
500 and 2500 Hz, with a maximum level corresponding
most often to the harmonic (twice the fundamental
frequency, F
o
).
Signals with a modified harmonic structure
We modified the parental calls in two ways.
(1) The parental call was filtered by low-pass or high-
pass digital filters (Fig. 1) by applying optimal filtering
with a fast Fourier transform (FFT; Press et al. 1988;
Mbu-Nyamsi et al. 1994). The window size of the FFT was
4096 Hz (precision in frequency: ÄF=4 Hz). Four signals
were constructed (Fig. 1): with the fundamental fre-
quency alone; with the fundamental (F
o
) and the first
harmonic (2F
o
); with the lower part of the spectrum
(F
o
+2F
o
+3F
o
); and with the upper part of the spectrum
(between 2000 and 8000 Hz).
(2) The parental call was shifted up or down in fre-
quency. This was done by picking a data record through a
square window, applying short-term overlapping (50%)
FFT, followed by a linear shift (+ or ) of each spectrum
and by a short-term inverse fast Fourier transform
(FFT
1
;Randall & Tech 1987). The window size was
4096 Hz (ÄF=4 Hz). The linear shifts of the spectra were
+100, + 75, +50, 50, 75 and 100 Hz. These values
were chosen on the basis of the natural distribution of
fundamental frequency values. Except for these modifica-
tions of the pitch of the carrier frequency, temporal
and amplitude parameters of the parental call were
unchanged.
Signals with a modified temporal pattern
We modified both frequency and amplitude modula-
tions (FM and AM) and the syllable duration of natural
calls.
(1) For FM and AM modifications, two signals were
constructed. (a) A natural AM was applied to a carrier
frequency without FM (Fig. 2). The carrier frequency was
a harmonic series and the value of the fundamental
corresponded to the mean value of the fundamental
frequency of the parental call. We applied to this carrier
1177JOUVENTIN ET AL.: KING PENGUIN CALL RECOGNITION
frequency the natural AM (the envelope) that was
extracted from the call of the parent, using the Hilbert
transform calculation (Seggie 1987;Bre´mond & Aubin
1992;Mbu-Nyamsi et al. 1994). In these conditions,
because of the application of the envelope, the temporal
succession of syllables of the parental call was main-
tained. The only difference between this call and the
natural call was the lack of FM. (b) The AM of the parental
call was removed without modification of the natural FM
and the natural carrier frequency. To do this, we used
analytical signal analysis (Mbu-Nyamsi et al. 1994). The
result was a signal with a normal FM and duration, but
without any AM.
(2) To modify syllable duration, we truncated the sylla-
ble. To prevent spectral artefacts arising from an abrupt
gap in amplitude, an envelope was applied (by multipli-
cation) to the data set in the time domain so as to smooth
all the edges. As previously, the Hilbert transform calcu-
lation was used to build the envelope. Previous studies
(Derenne et al. 1979;Jouventin 1982) have shown that
the broadcast of a part of the parental call is sufficient to
elicit the chick’s response. In the present study, we
Figure 1. Sound spectrograms of a king penguin parental call and of the four experimental signals corresponding to the same call modified
in the frequency domain.
1178 ANIMAL BEHAVIOUR, 57, 6
broadcast signals where only the first or a part of the first
syllable of the parental call was present. We constructed
five signals: with the first syllable (Sy1); with a syllable
belonging to the middle of the call (Sym); with the
first half of the first syllable (H1Sy1); with the second
half of the first syllable (H2Sy1); and with the first quarter
of the first syllable (Q1Sy1). The mean durationSD
of the syllables of the calls tested was 0.4620.011 s
(N=17).
RESULTS
To simplify the results, we grouped response classes 0 and
1 together as negative responses, and classes 2, 3 and 4 as
positive responses. Effectively, it was only for classes 2, 3
and 4 that recognition appeared clearly, with a call in
reply to the signal broadcast. In contrast, for class 1
responses, chicks only looked towards the loudspeaker,
not particularly because they recognized the parental call
but more probably because they were surprised by an
unusual (nonspecific?) signal. As a general rule, the chicks
did not recognize the manipulated signals as well as
they did the original parental call (Agresti’s marginal
homogeneity test: P<0.05).
Harmonic Structure
Table 1 shows the results. There was no significant
difference between responses to F
o
and to F
o
+2F
o
. For
both signals a majority of positive responses was
observed. In contrast, there was a significant difference
between the low-pass and the high-pass signals, the latter
eliciting a majority of negative responses.
There were significant differences between some
frequency-shifted signals, but they elicited a majority of
positive responses (Table 1) except for the +100-Hz
signal, which differed significantly from all other signals
(marginal homogeneity tests: + 100 Hz versus: +75 Hz:
S=25; +50 Hz: S=29; 50 Hz: S=29; 75 Hz: S= 28;
100 Hz: S=18; P<0.001 in all cases except 100 Hz
where P=0.005).
Temporal Pattern
Table 2 shows the results. Signals with a natural AM
and without FM did not trigger a response. In every case
(except one instance of head turning), chicks remained
stationary, resting or preening themselves, as before the
broadcast. In contrast, signals with a natural FM and
without AM triggered recognition. These signals differed
significantly.
The broadcast of one syllable was sufficient to elicit
recognition (majority of positive responses). Recognition
was not linked to a particular syllable since it was
observed for both the first syllable and one from the
middle of the call (no significant difference between Sy1
and Sym). The majority of chicks recognized just the first
half of the first syllable (mean durationSD =0.231
0.005 s, N=17; no significant difference between Sy1 and
H1Sy1). Recognition did not occur when only the first
quarter or the second half of the first syllable was broad-
cast (significant difference between Sy1 on the one hand
and Q1Sy1 and H2Sy1 on the other).
DISCUSSION
Which Parameters Encode Acoustic Identity?
That the chicks did not recognize the manipulated
signals as well as they did the original signal, that is, the
6Parental call (CS)
0
1 s
With AM: without FM
Frequency (kHz)
6
0
With FM: without AM
6
0
Figure 2. Sound spectrograms of the parental call shown in Fig. 1
and of the two experimental signals corresponding to the same call
but with either the amplitude modulation (AM) or the frequency
modulation (FM) removed.
1179JOUVENTIN ET AL.: KING PENGUIN CALL RECOGNITION
parental call, is not surprising. The majority of birds use a
complex of differentially weighted parameters, rather
than any simple feature, to recognize their signals. This
has been shown for songs (Weary 1990) and for calls
(Gaioni & Evans 1986;Dooling et al. 1987). The king
penguin is similar in this respect. Thus, the lack of some
parameters even weakly important for individual recog-
nition in our manipulated calls would explain the
different level of response to the manipulated signal.
Our experiments on harmonic structure show that
chicks pay attention to the low part of the spectrum of
their parents’ call, not to the higher part. Only the
low-pass calls were recognized by the chicks. Even a signal
with only the F
o
and 2F
o
retained was still recognized.
Nevertheless, a parental call with only the fundamental
frequency kept was not recognized. A pure tone, such as a
signal with the fundamental alone, is therefore not suffi-
cient and the addition of the first harmonic is necessary
to trigger a response. Thus we conclude that chicks pay
attention to the width of the spectrum of their parental
calls (two frequencies at least are necessary), the import-
ant part being the lower frequencies. They may use low
frequencies because high frequencies cannot be transmit-
ted far in the atmosphere without strong attenuation
(Wiley & Richards 1978) and the background noise is so
loud that a call cannot be heard more than 16–18 m away
(Jouventin 1982;Aubin & Jouventin 1998). Moreover,
high frequencies cannot be propagated through penguin
bodies (Robisson 1991).
Our results also show that chicks are sensible to the
effects of shifting frequency: significant differences were
obtained with changes of as little as 25 Hz (for example
between + 75- and +50-Hz signals and +100- and +75-Hz
signals). Nevertheless, signals shifted 100 Hz down or
75 Hz up still elicited a majority of positive responses
(ranked 2–4 in intensity).
Our experiments showed that AM alone was not
sufficient to elicit recognition, even though such a
Table 1. Responses of king penguin chicks to parental calls with modified frequency parameters
Experiments
Responses
N
Marginal
homogeneity
tests
0123 4 SP
Filtration
F
o
5325 116
}
F
o
+2F
o
2016 312 24 0.351
Lowpass 0008 816
}
Highpass 5531 216 62 0.0004
Linear shift (Hz)
+100 8314 016
}
+75 0227 516
}
25 0.0002
+50 00061016
}
26 0.016
−50 00041216
}
8 1.000
−75 1116 7 16
}
28 0.013
−100 3147 116 56 0.039
For the linear shift series, Pvalues are Bonferroni corrected.
Table 2. Responses of king penguin chicks to parental calls with modified temporal parameters
Experiments
Responses
N
Marginal
homogeneity
tests
0123 4 SP
Modulation
Without FM 15 1 0 0 0 16
}
Without AM 0 0 0 7 10 17 16 <0.001
Syllable duration
1st syllable (Sy1) 1 2 1 11 2 17
}
}
}
}
Middle syllable (Sym) 1 1 2 4 2 12 29 3.375
1st half of 1st syllable (H1Sy1) 4 1 1 8 1 15 19 0.750
2nd half of 1st syllable (H2Sy1) 10 1 1 0 0 12 41 0.004
1st quarter of 1st syllable (Q1Sy1) 12 1 1 0 0 14 53 <0.001
For the syllable duration series, Pvalues are Bonferroni corrected.
1180 ANIMAL BEHAVIOUR, 57, 6
signal respects the harmonic structure and the temporal
succession of syllables of the parental call. In contrast, a
signal in which the parental AM was suppressed was
recognized without ambiguity by all the chicks tested.
These experiments prove that FM is a key feature for the
recognition of the parental call, whereas AM is not used
for individual recognition. The information content of
the FM is concentrated at the level of the syllable since
recognition occurred when only one syllable of the
parental call was broadcast (signals Sy1 and Sym), what-
ever the syllable (the first or a syllable in the middle of the
call). In fact, it seems that the first half of the syllable was
sufficient for recognition. In contrast, the second part of
the syllable was not recognized by the chicks. With regard
to the FM structure of a syllable of a parental call, it
appears that the basic shape is always the same: an
increase followed by a decrease in frequency. For all the
parental calls that we have analysed, the inflexion point
between the increase and decrease in frequency was
always contained in the first half of the syllable. It seems
that this inflexion point is necessary to elicit recognition.
When the inflexion point was lacking, as with the first
quarter or the second half of the syllable (signals Q1Sy1
and H2Sy1), the signal was not recognized. The shape of
the increasing and then decreasing frequency of the first
half of the syllable should correspond to the vocal signa-
ture of the call. These results with chicks parallel previous
findings with adult king penguin calls. Derenne et al.
(1979) and Robisson (1992) performed field playback
experiments in which they tested the ability of adults to
recognize their mates. They played back reversed calls or
signals whose pitch and duration of syllables had been
more or less modified and in all cases individual recogni-
tion failed. In these experiments, the FM signature of the
syllable was modified. As for recognition between mates,
the recognition of the parent by the chick seems to be
linked to the FM shape of the syllable, which is repeated
as a vocal signature on the different syllables of the call.
Adaptation to the Noisy Environment of the
Colony
Problems of animal communication in noise can be
analysed in engineering terms (Okanoya & Dooling
1991). Two methods used in communication engineering
are used to study sensory filtering in animals from the
receiver’s point of view: the frequency-based filter model
and the matched filter model (Hopkins 1983). In a
frequency-based filter the output signal corresponds to
the correlation between the spectrum of the input signal
and the gain of the filter. Such examples of correlations
between the acoustic signal and behavioural and auditory
thresholds are numerous in birds (Konishi 1970;Dooling
et al. 1971;Dooling & Saunders 1975;Okanoya &
Dooling 1988). In the matched filter model, the output of
the filter corresponds to the cross-correlation between the
received and expected signals (an internal template). The
existence of this model is suggested by some studies (for
example Gaioni & Evans’s 1986 study of the distress call
of mallard ducklings, Anas platyrhynchos). It is likely that
king penguin chicks use matched filtering to detect the
parental call embedded in the masking noise of the
colony. Effectively, signals with reversed syllables or with-
out FM or without the inflexion point (in the FM struc-
ture) did not elicit recognition. On the other hand,
signals strongly shifted up or down in frequency did elicit
positive responses, and this fails to support the notion
that chicks were using frequency-based filters. Cross-
correlation detection by matching filtering is known to
be the most sensitive method for detecting a signal in
noise (Lee 1960). The use of this method of detection
should explain why chicks are able to extract the parental
call even when call intensity is well below that of the
noise of simultaneous calls produced by other adults in
the colony (Aubin & Jouventin 1998). In humans, this
process of acoustic recognition against a background
noise has been called the cocktail-party effect (Cherry
1966) and several authors have suggested its occurrence
in animals (Busnel 1977;Wiley & Richards 1982).
Another important point concerns the redundancy of
the information. The parental call corresponds to a suc-
cession of syllables with a broad frequency band. Our
experiments showed that the receiver needs only a small
part of the information to recognize the call: chicks
identified the parental call with only the first two har-
monics, with one syllable and even with the first half of a
syllable, a time period less than 6% of the total signal
duration (230 ms for a signal of 4–5 s mean duration).
Thus the king penguin call is highly redundant in time
and frequency. Why such a high redundancy? The back-
ground noise of the colony is almost continuous and
windows of silence are scarce and unpredictable, for
example during a 4-min recording made in a feeding area
(unpublished data), only 15% of the time corresponded
to relatively quiet periods and the mean duration of these
periods of silence was 0.6 s. The adult cannot predict
when and for how long it can be heard without jamming.
To enhance the chance of feeding its chick, the adult
repeats the same information many times and, therefore,
has the opportunity to find a window of silence. In terms
of harmonic structure, some parts of the spectrum are
more or less modifed, depending on the conditions of
propagation of the signal, for example, distance, the
screening effect of penguin bodies. In this case, if import-
ant parameters, such as the frequency modulation, are
distributed over many harmonics, these redundant ele-
ments of information are more likely to be transmitted
without destruction. According to the theory of informa-
tion (Schannon & Weaver 1949), such high redundancy
improves the probability of receiving a message in a noisy
channel.
Lastly, the parental call of the king penguin comprises
three to seven syllables (Jouventin 1982), each syllable
being separated by deep declines in amplitude which
appear as silences. These sharp amplitude variations,
corresponding to a succession of syllables, are easily
distinguished from continuous noises such as wind or sea
waves. They also have the advantage of increased locat-
ability (Wiley & Richards 1982) which in turn helps the
receiver to detect the signal in the background noise. This
is important for king penguins, which do not have a nest
site. All the chicks responded with a score of 4 to the
1181JOUVENTIN ET AL.: KING PENGUIN CALL RECOGNITION
natural parental call and with a score of 2–4 for modified
signals such as low pass, without AM or one-syllable
signals. These results do not necessarily imply that these
modified signals are recognized less; instead they could be
more difficult to locate (class 4 response corresponding to
an approach oriented towards the loudspeaker).
With an FM signature highly redundant in time and
frequency, the code of the king penguin call can thus be
regarded as a functional code increasing the possibility of
individual recognition in an acoustically constraining
environment.
Biological Significance of the Code
Parent–offspring recognition by acoustic signals has
been amply documented for birds, and particularly for
birds with nests (Falls 1982;Beecher 1989). Often, discus-
sions have emphasized the parents’ need to recognize
offspring but have neglected pressures for young to rec-
ognize parents (Beecher et al. 1985), especially in species
without a nest site, such as king penguins. Derenne et al.
(1979) showed that parents were able to recognize the call
of their young and we have now shown that chicks also
recognize the FM signature of the parental call. This
recognition process supposes that three conditions are
fulfilled.
(1) The chick must learn the parental call. As a result,
the call of the parent forms a template (Slater 1989) that,
according to the matched filter model described above,
the chick correlates (compares) with each received call. In
numerous birds, parent–young recognition is effective
when young become mobile (Falls 1982;Williams 1982;
Bo¨hner 1990), for example the Adelie penguin, Pygoscelis
adeliae (Jouventin & Roux 1979). The young king pen-
guin has the opportunity to memorize the call of its
parent during its first 5 weeks when it remains on the
parents’ feet, and maybe before, during hatching. Each
time a parent returns from foraging to meet its mate, it
gives a call for mutual identification, and this call, heard
by the chick, is identical to the one the chick later uses to
identify its parent (Jouventin 1982).
(2) The FM signature must offer a sufficient number of
variations to ensure recognition of each individual and
avoid confusion with others. Beecher (1982,1989) devel-
oped a quantitative method for measuring the informa-
tion capacity needed in a signature system to identify
each member of a population with a small probability of
confusion. In our case, it was difficult to calculate the
theoretical number of possible features of FM signatures.
Effectively, an FM signature allows an almost infinite
number of combinations between temporal and fre-
quency parameters, far more than that necessary under
natural conditions to ensure individual recognition in the
colony. In addition, there could be fewer FM signatures
than adults in the colony since chicks meet their parents
on a rendezvous site, so limiting the probability of
confusion.
(3) The FM signature of the parents must be sufficiently
stable during the year. The chick must learn at least one
FM signature contained in the call of each parent and
must be able to distinguish these parental signatures from
others. This is a complex learning process. A parental
signature system that continually changed would be very
confusing for the chick. Previous findings indicate that
king penguins do not change their calls: the temporal and
frequency features of the calls remain remarkably con-
stant during the year (Jouventin 1982) and even from
year to year (unpublished data). Nevertheless, in some
species of seabirds, the calls vary continuously. Learning
of parental calls by chicks appears to be a countinuing
process in the laughing gull, Larus atricilla (Beer 1979), for
example; this species has nest sites, however, so land-
marks limit the possibility of confusion and facilitate
individual recognition.
The evolution of coding and decoding sounds may be
determined by phylogeny or ecology. The penguin family
uses the largest range of breeding sites known for colonial
breeding species: (1) burrows as in the little penguin,
Eudyptula minor; (2) nests made with stones, grasses or
branches as in the majority of penguins (and birds); (3)
without a nest but on a site during the brooding phase as
in the king penguin; and (4) without a nest and without
a site such as the emperor penguin, Aptenodytes forsteri,
moving with its egg on its feet. Since the two non-nesting
species need only vocal cues to identify themselves
among several hundred breeders (Jouventin 1982), they
constitute good models for individual recognition. In
future studies, we intend to compare the acoustic mecha-
nisms of identification of the nesting and non-nesting
species to determine if coding-decoding processes are
related to the breeding constraints of these four groups of
birds.
Acknowledgments
The study was supported in the field by the Institut
Franc¸ais pour la Recherche et la Technologie Polaires
(I.F.R.T.P.) and in the laboratory by the Centre National
de la Recherche Scientifique (C.N.R.S.). We are grateful to
Jean Claude Bre´mond, Marcel Lambrechts, Jacques Lauga,
Stephen Hall and an anonymous referee for comments
and Antoine Catard for help in the field. Mary-Anne Lea
improved the English.
References
Agresti, A. 1990. Categorical Data Analysis. New York: J. Wiley.
Aubin, T. 1994. Syntana: a software for the synthesis and analysis of
animal sounds. Bioacoustics,6, 80–81.
Aubin, T. & Jouventin, P. 1998. Cocktail-party effect in king
penguin colonies. Proceedings of the Royal Society of London, Series
B,265, 1665–1673.
Beecher, M. D. 1982. Signature systems and kin recognition.
American Zoologist,22, 477–490.
Beecher, M. D. 1989. Signalling systems for individual recognition:
an information theory approach. Animal Behaviour,38, 248–261.
Beecher, M. D., Stoddard, P. K. & Loesche, P. 1985. Recognition
of parent’s voices by young cliff swallows. Auk,102, 600–605.
Beer, C. G. 1979. Vocal communication between laughing gull
parents and chicks. Behaviour,70, 118–146.
Bo¨ hner, J. 1990. Early acquisition of song in the zebra finch,
Taeniopygia guttata.Animal Behaviour,39, 369–374.
1182 ANIMAL BEHAVIOUR, 57, 6
Bre´mond, J. C. & Aubin, T. 1992. The role of amplitude modulation
in distress-call recognition by the black-headed gull (Larus argen-
tatus). Ethology, Ecology and Evolution,4, 187–191.
Busnel, R. G. 1977. Acoustic communication. In: How Animals
Communicate (Ed. by T. A. Sebeok), pp. 233–251. Bloomington:
Indiana University Press.
Cherry, C. 1966. On Human Communication. 2nd edn. Cambridge,
Massachusetts: MIT Press.
Cytel 1995. StatXact 3 User’s Manual. Cambridge: Cytel Software
Corporation.
Dabelsteen, T. 1984. An analysis of the full song of the blackbird
Turdus merula with respect to message coding and adaptations for
acoustic communication. Ornis Scandinavica (Journal of Avian
Biology),15, 227–239.
Dabelsteen, T., Larsen, O. N. & Pedersen, S. B. 1993. Habitat-
induced degradation of sound signals: quantifying the effects of
communication sounds and bird location on blur ratio, excess
attenuation, and signal-to-noise ratio in blackbird song. Journal of
the Acoustic Society of America,93, 2206–2220.
Derenne, P., Jouventin, P. & Mougin, J. L. 1979. Le chant du
manchot royal (Aptenodytes patagonicus) et sa signification e´ volu-
tive. Le Gerfaut,69, 211–224.
Dhondt, A. A. & Lambrechts, M. M. 1992. Individual recognition in
birds. Trends in Ecology and Evolution,7, 178–179.
Dooling, R. J. & Saunders, J. C. 1975. Hearing in the parakeet
(Melopsittacus undulatus), absolute thresholds, critical ratios, fre-
quency difference limens, and vocalizations. Jounal of Comparative
Psychology,88, 1–20.
Dooling, R. J., Mulligan, J. M. & Miller, J. D. 1971. Auditory
sensitivity and song spectrum of the common canary
(Serinus canarius). Journal of the Acoustic Society of America,50,
700–709.
Dooling, R. J., Brown, D. D., Park, T. J., Okanoya, K. & Soli, S. D.
1987. Perceptual organization of acoustic stimuli by budgerigars
(Melopsittacus undulatus): 1. Pure tones. Journal of Comparative
Psychology,101, 367–381.
Falls, J. B. 1982. Individual recognition by sound in birds. In: Acoustic
Communication in Birds, Vol. 2 (Ed. by D. E. Kroodsma & E. H.
Miller), pp. 237–278. New York: Academic Press.
Gaioni, S. J. & Evans, C. S. 1986. Perception of distress calls in
mallard ducklings (Anas platyrhynchos). Behaviour,99, 250–274.
Guinet, C., Jouventin, P. & Malacamp, J. 1995. Satellite remote
sensing in monitoring change of seabirds. Use of spot image in
king penguin population increase at Ile aux cochons, Crozet
Archipelago. Polar Biology,15, 511–515.
Hopkins, C. P. 1983. Sensory mechanisms in communication. In:
Animal Behaviour 2, Communication (Ed. by T. R. Halliday & P. J. B.
Slater), pp. 123–154. Oxford: Blackwell Scientific.
Hutchison, R. E., Stevenson, J. G. & Thorpe, W. 1968. The basis for
individual recognition by voice in the Sandwich tern (Sterna
sandvicensis). Behaviour,32, 150–157.
Jouventin, P. 1982. Visual and Vocal Signals in Penguins, their
Evolution and Adaptive Characters. Berlin: Verlag Paul Parey.
Jouventin, P. & Mauget, R. 1996. The endocrine basis of the
reproductive cycle in the king penguin (Aptenodytes patagonicus).
Journal of Zoology,238, 665–678.
Jouventin, P. & Roux, P. 1979. Le chant du Manchot Ade´ lie. Roˆle
dans la reconnaissance individuelle et comparaison avec le chant
d’un oiseau non-territorial, le Manchot empereur. L’Oiseau et
R.F.O.,49, 31–37.
Jouventin, P., Guillotin, M. & Cornet, A. 1979. Le chant du
manchot empereur et sa signification adaptative. Behaviour,70,
231–250.
Klump, G. M. 1996. Bird communication in the noisy world. In:
Ecology and Evolution of Acoustic Communication in Birds (Ed. by
D. E. Kroodsma & E. H. Miller), pp. 321–338. Ithaca: Cornell
University Press.
Konishi, M. 1970. Comparative neurophysiological studies of hear-
ing and vocalizations in songbirds. Zeitschrift fu¨ r Vergleichende
Physiologie,66, 257–272.
Lee, Y. M. 1960. Statistical Theory of Communication. New York: J.
Wiley.
Lengagne, T., Lauga, J. & Jouventin, P. 1997. A method of
independent time and frequency decomposition of bioacoustic
signals: inter-individual recognition in four species of penguins.
Comptes Rendus de l’ Acade´ mie des Sciences,320, 885–891.
Mathevon, N. 1996. What parameters can be used for individual
acoustic recognition by the greater flamingo? Comptes Rendus de
l’ Acade´mie des Sciences,319, 29–32.
Mbu-Nyamsi, R. G., Aubin, T. & Bre´mond, J. C. 1994. On the
extraction of some time dependent parameters of an acoustic
signal by means of the analytic signal concept. Its application to
animal sound study. Bioacoustics,5, 187–203.
Okanoya, K. & Dooling, R. J. 1988. Hearing in the swamp sparrow,
Melospiza georgiana, and the song sparrow, Melospiza melodia.
Animal Behaviour,36, 726–732.
Okanoya, K. & Dooling, R. J. 1991. Detection of species-specific
calls in noise by zebra finches Poephila guttata and budgerigars
Melopsittacus undulatus: time or frequency domain? Bioacoustics,
3, 163–172.
Press, W. H., Flannery, B. P., Teukolsky, S. A. & Vetterling, W. T.
1988. Numerical Recipes in C. The Art of Scientific Computing. New
York: Cambridge University Press.
Randall, R. B. & Tech, B. 1987. Frequency Analysis. Naerum: Bru¨el &
Kjaer.
Robisson, P. 1990. The importance of the temporal pattern of
syllables and the syllable structure of display calls for individual
recognition in the genus Aptenodytes.Behavioural Processes,22,
157–163.
Robisson, P. 1991. Broadcast distance of the mutual display call in
the emperor penguin. Behaviour,119, 302–316.
Robisson, P. 1992. Roles of pitch and duration in the discrimination
of the mate’s call in the king penguin Aptenodytes patagonicus.
Bioacoustics,4, 25–36.
Robisson, P. 1993a. Adaptation du transfert de l’information
individuelle au milieu colonial chez les manchots. Revue
d’Ecologie—Terre et Vie,48, 133–141.
Robisson, P. 1993b. La reconnaissance individuelle chez deux
espe`ces jumelles. Le manchot empereur Aptenodytes forsteri,etle
manchot royal Aptenodytes patagonicus. PhD thesis, University of
Rennes.
Schannon, C. E. & Weaver, W. 1949. The Mathematical Theory of
Communication. Urbana: University of Illinois.
Scherrer, B. 1984. Biostatistique. Que´bec: Gae¨tan Morin.
Seggie, D. 1987. The application of analytic signal analysis in speech
processing. Processing Institute of Acoustics,8, 82–85.
Slater, P. J. B. 1989. Bird song learning: causes and consequences.
Ethology, Ecology and Evolution,1, 19–46.
Stonehouse, B. 1960. The king penguin Aptenodytes patagonicus of
South Georgia. 1. Breeding behaviour and development. Falkand
Island Dependencies Survey Scientific Reports,23, 1–81.
Weary, D. M. 1990. Categorization of song notes in great tits: which
acoustic features are used and why? Animal Behaviour,39, 450–
457.
Wiley, R. H. & Richards, D. B. 1978. Physical constraints on acoustic
communication in the atmosphere: implication for the evolution
of animal vocalizations. Behavioral Ecology and Sociobiology,3,
69–94.
Wiley, R. H. & Richards, D. G. 1982. Adaptations for acoustic
communication in birds: transmission and signal detection. In:
Acoustic Communication in Birds, Vol. I (Ed. by D. E. Kroodsma &
E. H. Miller), pp. 131–181. New York: Academic Press.
Williams, H. 1982. Models for song learning in the zebra finch:
fathers or others? Animal Behaviour,39, 745–757.
1183JOUVENTIN ET AL.: KING PENGUIN CALL RECOGNITION
