ADVANCESIN THE STUDY OF BEHAVIOR, VOL. 31 How to Vocally Identify Kin in a Crowd: The Penguin Model T H I E R R Y A U B I N 1 AND P I E R R E JOUVENTIN 2 1NAM-CNRS U R A 1491 UNIVERSITE PARIS-XI F-91400 ORSAY, FRANCE 2CEFE-CNRS 34293 MONTPELLIER CEDEX 5, FRANCE I. INTRODUCTION Sociality has several major advantages but also some disadvantages. For example, it is well known that colonial breeding improves care of offspring through communai protection (Alcock, 1972), but it is less well known that the difficulty of finding one's own young is increased. Marine birds and mammals breed on land but have to forage at sea. This fact is indeed the main constraint on their social behavior and their life history: seals have large fat reserves and in some species such as fur seals the adults have to alternate feeding trips at sea, a characteristic particularly striking in seabirds also (see Jouventin and Cornet, 1980, for a comparison between pinnipeds and seabirds; Bried and Jouventin, 2001, for seabirds). A wandering albatross, Diomedea exulans, may fly several thousand kilometers to find food (Jouventin and Weimerskirch, 1990) and a king penguin, Aptenodytes patagonicus, may swim several hundred kilometers on a foraging trip. As a consequence, both sexes have to cooperate to brood and rear one chick, and this heavy breeding cost explains why all of the nearly 200 species of seabirds are monogamous (Lack, 1968). Pelagic birds have few if any predators on land, and they usually breed in colonies numbering several hundreds or thousands of pairs. After fishing for several hours, days, or weeks, one parent comes back to the crowded colony, finds its mate and takes its turn at brooding while the mate forages. Later, both parents have to forage at sea to feed the growing young. The arriving parent has to be recognized at the nest by its mate to be safe and to brood in turn and later by the large chick(s) to feed it. An 243 Copyright © 2002 by Academic Press All rights of reproduction in any form reserved. 0065-3454/02$35.00 244 THIERRY AUBIN AND PIERRE JOUVENTIN alien bird is pecked and harassed away. The ability to invest cooperatively in their own offspring (Trivers, 1972) by identifying the members of the family is thus crucial for marine animals, as it is for many other animal species. Among seabirds where feeding trips are an ever-present feature of the breeding biology, the family of penguins is unique (1) in our knowledge of their individual recognition strategies, and (2) in the fact that they exhibit two different methods of care for eggs and chicks (some species have a nest and others do not). In a comparative study (Jouventin, 1982), we demonstrated experimentally in several species that penguins identify their mates or their chick(s) only by vocal cues, being unable to use visual cues to distinguish family members from surrounding birds: the display call is consequently the only marker used in kin recognition. Penguins are also unique because they include both nesting species as in other seabirds and nonnesting species, which walk with their egg or chick on their feet. In this group of colonial species, in which even topographical cues of nest location can be lacking (Isenmann and Jouventin, 1970), the difficulty of meeting up with mates and offspring is extreme, particularly with the latter, unusual way of breeding. So penguins, which are easy to observe, to record, to manipulate, and to test, constitute extraordinary models for studying individual recognition by acoustic means. There are 15 species of penguins that, as in most seabirds, breed on a nest. This nest can be made with stones, as in the antarctic Ad61ie penguin (Pygoscelis adeliae) or the subantarctic macaroni penguin (Eudyptes chrysolophus). Nests can also be made of grass, as in the gentoo penguin (P. papua), which breeds on the same subantarctic islands as the latter species, but on plateaus. Some species nest in a burrow such as the nocturnal little blue penguin (Eudyptula minor) which breeds along the Australian and New Zealand coasts. Much more exceptional in seabirds, the two large (nearly 1-meter tall) nonnesting Aptenodytes penguins brood their single egg on their feet and usually walk around during brooding. They breed on flat and inhospitable areas such as wet subantarctic beaches in the king penguin (A. patagonicus) and the antarctic sea-ice in the emperor penguin (/t. forsteri). The first of these does not move more than a few meters during the brooding phase but is much more mobile during the rearing phase when the chick waits alone for its parents to feed it (Stonehouse, 1960; Lengagne et al., 1999a). The latter species moves through all of the breeding cycle because when there are blizzards blowing, brooders and chicks have to huddle in tight groups of 10 birds/m 2 to keep warm (Pr6vost and Bourlihre, 1957; Pr6vost, 1961). During storms, the wind may blow at 350 km/h and the temperature drop VOCALLY IDENTIFYING KIN IN A CROWD: THE PENGUIN MODEL 245 to -34°C (meteorological data from Terre Ad61ie, where our studies on antarctic penguins were conducted). The ritualized postures and the various calls of the penguin family have been described in detail by Jouventin (1982). The biological meaning of calls and displays was determined through the observation of marked birds, often over several years, according to the ecological and social context. All genera, and for some genera several species, were studied to describe, understand, and compare their visual and vocal displays and to test by playback their use of calls in individual recognition. Using computers, more sophisticated playback experiments have been conducted to examine the coding-decoding systems of penguins. In this chapter we summarize the acoustic constraints and behavioral adaptations described by Jouventin (1982), and we synthesize playback results to give a comparative account of the acoustic systems of six different species (Aubin and Jouventin, 1998; Jouventin et aL, 1999; Lengagne et al., 1999a,b,c, 2000; Aubin et al., 2000; Jouventin and Aubin, 2000, 2001). II. A. LOCATIONS AND METHODS SUBJECTS AND LOCATION Six species of penguins have been studied in the past five years, at the following locations: • The Ad61ie penguin and the emperor penguin were studied at the Pointe G6ologie Archipelago (66o40, S, 140o01' E), Terre Addlie, Antarctica, • The king, the gentoo, and the macaroni penguins were studied mainly at Possession Island (46025, S, 51°45 , E), Crozet Archipelago, Indian Ocean, • The little blue penguin was studied at Phillip Island (38o31' S, 145008` E), 60 km from Melbourne, Australia. B. RECORDING AND PLAYBACK PROCEDURES Calls were recorded with a Sony TCD Pro II DAT recorder (frequency response flat within the range 20-20,000 Hz) and an omnidirectional Sennheiser MKH 815T microphone (frequency response 100-20,000 Hz ± 1 dB) mounted on a 3 m pole, so that birds could be approached without disturbance. The distance between the beak of the recorded bird and the microphone was approximately 1 m for all species studied. Experimental 246 THIERRY AUBIN AND PIERRE JOUVENTIN signals were broadcast with the same tape recorder connected to a PSP-2 E.A.A. preamplifier and a 20 W self-powered amplifier built in the laboratory, equipped with an Audax loudspeaker (frequency response 100-5600 Hz -4-2 dB). For propagation tests, signals were rerecorded by means of an omnidirectional Sennheiser MKH 815T microphone connected to another Sony TCD10 Pro II DAT. For sound pressure level measurements (SPL in dB), we used a Bru~l & Kjaer Sound Level Meter type 2235 (linear scale, slow setting) equipped with a 1 in. condenser microphone type 4176 (frequency response 2.6-18,500 Hz 4- 2 dB). For propagation experiments, representative calls of each species studied were chosen. These signals were broadcast repetitively (generally 10 calls) through the colony and recorded at distances of 1 m (reference), 7 m (average distance between two birds when the incoming one started the acoustic search of the mate or the chick), and 14 m (maximum distance of recognition observed in most cases). The loudspeaker and the microphone were mounted on a tripod at the height of a penguin head (0.9 m for a king or emperor penguin, 0.7 m for an Ad61ie penguin, and 0.4 m for a little blue penguin). To quantify the screening effect of the bodies of birds, the recordings were compared with propagation records made at the same microphone and loudspeaker height and the same distances, but without any penguins present. The series of recorded calls were then examined in the amplitude-versus-times and the amplitude-versus-frequency domains. In playback experiments, both chicks and adults tested were flipperbanded with a temporary plastic band to identify them. Playback experiments were conducted during clear and dry weather, with a wind speed of less than 4 m/s. The bird tested (adult or chick) was generally quiet, preening itself. The distance between the loudspeaker and the bird was on average 7 m, this corresponding to a natural calling situation between birds. Signals were played at SPL levels equivalent to those produced by the species tested. To prevent habituation, a maximum of three experimental signals per day was broadcast to any one bird. For the different birds tested, the order of presentation of the different experimental signals was randomized. In the same way, the order of presentation of signals tested was not the same for each bird from day to day. Thus, the observed responses for the whole group of individuals tested were neither the result of cumulative excitation nor dependent on playback order. For each species and for each experimental signal, from 8 to 25 individuals were tested. In natural conditions and during the absence of the parents, chicks gathered in flocks, remaining silent and standing or lying quietly. An adult coming from the sea to feed its chick makes its way to the area of the colony where the nest (or the meeting place) is located and calls. Then, the corresponding chick in the flock holds up its head, looks around, calls in reply, and moves VOCALLY IDENTIFYING KIN IN A CROWD: THE PENGUIN MODEL 247 toward the parent, often running (Jouventin, 1982). The other chicks in the vicinity, resting or preening themselves, exhibit no behavioral reaction to the apparently extraneous calls. The recognition process between mates is exactly the same as this except that the brooding bird does not move toward the incoming bird, but only calls in reply until the two partners meet. On the basis of the observation of these natural meeting situations between parents and chicks or between mates a behavioral scale was used to evaluate the intensity of response of tested birds for all the different penguin species studied. The behavioral scale was ranked as follows: - , no response; +, moderate response (visually inspect the environment by head-turning and calling in reply to the signal after a delay); + + , strong response (turn to loudspeaker, immediately call in reply to the signal, and, only for the chicks, move in the direction of the loudspeaker). C. SOUND ANALYSIS AND SYNTHESIS Signals were digitized through 12- or 16-bit acquisition cards equipped with an antialiasing filter (low-pass filter, i 1 2 0 dB/octave) at a sampling rate of 12-48 kHz (depending on the call studied) and stored on the hard disk of a computer. Calls were then analyzed and signals synthesized mainly using Syntana software built in the laboratory (Aubin, 1994). Signals were examined in the amplitude-versus-frequency domain by spectrum analysis (fast Fourier transform (FFT) calculation) and in the amplitude-versus-time domain by envelope analysis (analytic signal calculation). To follow the time evolution of the frequency we used the Hilbert (Papoulis, 1977) or zerocrossing calculations which provide a representation of the instantaneous frequency. Fundamental frequencies were detected and measured using the Cepstrum calculation defined as the power spectrum of the logarithm of the power spectrum (Noll, 1967). Experimental signals were built either by constructive synthesis (i.e., by computer synthesis starting from scratch) or by destructive synthesis (i.e., by modifying natural calls). For the constructive synthesis, the signal to be synthesized at time t, S(t) is obtained by: N(h) S(t) = Z oJi sin[Zrr~o(t)t], i=1 where N(h) is the harmonic number, COi is the relative amplitude of the harmonic i as determined from the power spectrum of a reference signal (a natural call taken as model). For the constructive synthesis, natural calls 248 THIERRY AUBIN AND PIERRE JOUVENTIN were modified in the temporal, frequency, and amplitude domains. Amplitude and frequency modulations (respectively, AM and FM) were modified or removed using the Hilbert transform calculation (Brdmond and Aubin, 1992; Mbu-Nyamsi et el., 1994). For modifications of the harmonic structure, natural calls were filtered by low-, high-, or band-pass digital filters, by applying optimal filtering with FFT (Press et al., 1988; Mbu-Nyamsi et al., 1994). The window size of the F F r was 4096 points. Natural calls were also shifted up or down in frequency by picking a data record (from a natural call) through a square window, applying short-term overlapping (50%) followed by a linear shift up or down of each spectrum, and finally by a short-term overlapping inverse FFT (Randall and Tech, 1987). As before, the window size was 4096 points. To modify call or syllable durations, we truncated the sounds. To prevent spectral artifacts arising from gaps in amplitude, an envelope was applied (by multiplication) to the data set in the time domain to smooth all the edges. III. A. THE CONSTRAINTS THE BIOLOGICAL PROBLEMS Most seabirds breed in dense colonies numbering tens, hundreds, or even thousands of pairs. In penguin species, some colonies can near one or two million pairs. It is impressive to follow a king penguin leaving the sea with its stomach full and coming directly to its brooding mate or its chick on a flat area among many thousands of others in only 1 or 2 min (Lengagne et aL, 1999a). Although olfaction is well developed in petrels, another family of seabirds, penguins are not known for their abilities in olfaction and seem unable to find their partner using smell (Jouventin, 1982; Jouventin and Robin, 1984; Lequette et aL, 1989; Verheyden and Jouventin, 1994). More surprisingly, penguins apparently see well, even on land, but they are unable to identify their mate or chick(s) visually. Jouventin (1982 and unpublished data) confirmed its observations by experiments consisting of obstruction of the auditory ducts or of closing the birds' beaks by means of adhesive tape. In both these situations, partners were unable to recognize each other. The lack of visual recognition is not obvious because, after recognizing their partner vocally, penguins try to follow the bird visually in the crowd. Nevertheless, if the mate or the chick disappears completely in a group, or if the visual link is lost for some minutes, the birds cannot find each other again without calling. In fact, penguins, as with fur seals (Roux and Jouventin, 1987) and many other seabirds (Beer 1970, 1979; White and White, 1970; Charrier, 2001), find their kin using an acoustic signal. Because individual recognition is only V O C A L L Y I D E N T I F Y I N G KIN IN A C R O W D : THE P E N G U I N M O D E L 249 vocal in penguins, and the call concerned is the display call, it was possible to experiment using playback of the partner: the mate for adults or, more often, the parent for chicks. B. ThE AcousTIC PROBLEMS 1. The Background Noise of the Colony In all penguin species, the display call consists of a series of sound components termed syllables, separated by pronounced amplitude declines (see Figs. 1-3). The syllable is a complex signal based on harmonic series from 250 to >5000 Hz. The call is emitted at a relatively high sound pressure level (SPL) ranging from 85-90 dB for small penguins (Jouventin and Aubin, 2001, and manuscript in preparation) to >95 dB for large ones (Robisson, 1993b; Aubin and Jouventin, 1998). As penguins breed in dense colonies, calls emitted by other individuals generate a continuous background noise in the colony. In addition, other signals, such as agonistic calls and chick calls and nonbiologically significant sounds (wind, flipper flap), increase the level of ambient noise so that it is particularly high inside the colony (see Fig. 4). Thus, for a king penguin colony numbering 40,000 pairs, we have measured an average value of 74 dBsPL during a 4-min recording period (Aubin and Jouventin, 1998). In these conditions, we hypothesized that penguins could only establish communication at short range, that is, some meters. Penguins are large animals that are able to call loudly, an ability that is probably useful in overcoming the sound of the sea and the wind, but which cannot prevail against the signals of their equally loud neighbors. The noise generated by birds in the colony is almost continuous, and periods of relative silence are short, infrequent, and unpredictable. Thus, in a king penguin colony, periods of relative silence represent only 15% of the time and are of short (mean: 20 ms) duration (Aubin and Jouventin, 1998). In addition, the only noises that cover exactly the same frequency band as that of a penguin call, and which would thus in theory lead to a masking effect (Scharf, 1970), are the calls produced by conspecifics. The acoustic properties of the masker and those of the signaler are similar. In these conditions of competing noise, the jamming effect is very important, from an amplitude, time, and frequency point of view, and this increases the difficulty for a given bird to extract the information provided by the partner. 2. The Screening Effect of Bodies In breeding areas of a penguin colony, the density of birds is high, for example, 2.2 breeders/m2, measured by Barrat (1976) in a king penguin colony. Many individuals (nonbreeding adults or chicks) also gather in groups and sometimes huddle closely together. In such places, the density of birds can 250 THIERRY AUBIN AND PIERRE JOUVENTIN a 3 ~" 0.5 s I I b t~ Is I ! FrG. 1. Sound spectrograms and oscillograms of (a) an emperor penguin, Aptenodytes forsteri, call; and (b) a king penguin, Aptenodytes patagonicus, call. Calls correspond to a succession of syllables separated by gaps m amplitude. VOCALLY IDENTIFYING KIN IN A CROWD: THE PENGUIN MODEL 251 @.5 S I I tl klgl~ 0 0.5 s FIG. 2. Sound spectrograms and oscillograms of (a) a gentoo penguin, and (b) an Ad~lie penguin, Pygoscelis adeliae, caW. Pygoscelispapua, call; 252 THIERRY AUBIN AND PIERRE JOUVENTIN a 5kI~ . . . . . ! 0 0.5 s I I 10kNz : a .... ;X~: 0 ls I , ! Fro. 3. Sound spectrograms and oscillograms of (a) a little blue penguin, call; and (b) a macaroni penguin, Eudyptes chrysolophus, call. Eudyptula minor, 253 VOCALLY IDENTIFYING KIN IN A CROWD: THE PENGUIN MODEL wind + flipper flaps ~, calls 10000 - (adults+chicks) ~ I 9000 I I 8000 ,; i 7000 ; ; 6000 ; 5000 .~ ,.~ 4000 "~ 2000 ; i -- King -- Addlie 3000 1000 0 1000 2000 3000 4000 5000 6000 7000 8000 Frequency (Hz) Fie. 4. Spectra of a 4-rain recording of the background noise of king and Ad~lie colonies (for each colony, average of 937 successive FFTs, window size = 4096 data points, sampling frequency = 16 kHz, Hamming window). Adult and chick calls represent 60 and 55% of the energy (Welch calculation method) in the spectra of king and Addtie colonies, respectively. become twofold higher than it would be otherwise (Kodyman and Mullins, 1990; Lengagne et al., 1999b) and can even reach 10 birds/m 2 in an emperor penguin colony during a blizzard (Pr6vost, 1961). Then, the bodies of birds constitute an obstructed environment that impairs the propagation of signals. The screening effect of penguin bodies causes absorption and multiplescattering effects. To study the modification of the call during propagation, we broadcast and recorded signals at different distances in penguin habitats. In "open field" conditions, that is, without any birds between the emitter and the receiver, the measured attenuation of the signal fitted well with the amplitude decrease that would be expected from spherical spreading: it diminishes with distance according to the inverse square law ( - 6 dB per doubling of distance). When the same measurements are performed with penguins between the emitter and the receiver, excess attenuation occurs because porous objects such as penguin bodies absorb sounds (Wiley and Richard, 1978; Dabelsteen, 1981; Dabelsteen et aI., 1993). The greater the distance, and the greater the number of bodies in the way, the more the excess attenuation. According to our measurements, amplitude and frequency parameters of the signal showed strong degradation with increasing broadcast 254 THIERRY AUBIN AND PIERRE JOUVENTIN TABLE I PEARSON PRODUCT-MOMENT CORRELATION BETWEEN REFERENCEa AND PROPAGATED SIGNALS FOR AVERAGED ENVELOPES b AND AVERAGED SPECTRA c OF THE DISPLAY CALLS OF KING AND ADI~LIE PENGUINS Kingpenguin Distance of propagation 7m 14m Correlations for averaged envelopes AdGliepenguin Correlations for averaged spectra Correlations for averaged envelopes Correlafionsfor averaged spectra 0.54*** 0.47*** 0.68*** 0.54** 0.05* 0.27* 0.12" 0.39** Note: *p < 0.05, **p < 0.01, ***p < 0.001. aForeach species,the callrecordedat 1 m. bN = 10,63,600data points. CN= 10, FFT = 512 data points. distance (see the results for two penguin species in Table I). Thus, in the penguin colonies studied, after 14 m of propagation, the attenuation is so strong that the amplitude gaps that separate syllables tended to disappear and the amplitude of the signal became equal or inferior to the background noise (signal-to-noise ratio equal to or less than 1). In the frequency domain, peaks above the wavelength corresponding to the body size of a penguin (i.e., above 350 Hz for a 0.9 m high king penguin or emperor penguin and above 500 Hz for a 0.7 m AdGlie penguin) are more severely attenuated and disappear in the background noise after 14 m of propagation. For little blue penguins, the same disappearance is observed, but after only 8 m of propagation, due mainly to vegetation effects (Jouventin and Aubin, 2000). For amplitude parameters, peaks tend to be embedded in the noise and to disappear in the background noise after 14 m of propagation. According to these propagation tests, communication involving individual recognition seems possible only at a short or moderate range (<16 m). Communication in a penguin colony, and generally in seabird colonies, appears poorly adapted to transmission of individual information at long range (Robisson, 1991; Aubin and Lengagne, 1997; Aubin and Jouventin, 1998; Lengagne et al., 1999b). IV. THE SOLUTIONSFOUND A seabird colony, and particularly a penguin colony, is an extreme environment from an acoustic point of view only partly due to the loud background noise. It presents a particularly difficult problem of acoustic communication, due not only to the extraneous noises but also to propagation problems VOCALLY IDENTIFYING KIN IN A CROWD: THE PENGUIN MODEL 255 linked to the distance between partners or between the parent and chick and to the massive screen of birds. Faced with the problem of finding a particular individual in a penguin colony among several hundreds or thousands of conspecifics, the display call alone seems at first glance inadequate to secure communication. Nevertheless, penguins succeed, performing acoustic identification of the partner or chick usually in a few minutes (i.e., a relatively short time). How do they manage this? A. BEHAVIORAL SOLUTIONS TO OPTIMIZE IDENTIFICATION Acoustic recognition is a critical process for the breeding success of seabirds (Brooke, 1978). The incoming bird arriving from the sea has to find its mate or its chick in the colony on the basis of only a few calls. If it does not succeed, the bird returns to the sea. Thus, there is a possibility that the pair will fail to reunite or that the chick will die, so demonstrating the survival value of vocal recognition. To limit the high aggressiveness and the vigorous territorial defense of other breeders (via pecking and flipper flapping), the incoming bird must limit the time it takes in the colony to identify the mate or chick. For this purpose, penguins adopt some special strategies. 1. The Meeting Place Although colonies of penguins often number thousands of birds, an incoming adult does not have to locate its partner among all of these individuals. As described previously, the search is limited to particular meeting places (either nests or previous feeding sites for nonnesting species) that are memorized by adults and chicks. These visual cues may assist in individual recognition, but the mutual display call remains the only way to identify with certainty the partner and offspring. For the nesting species of penguins, such as macaroni, gentoo, or Addlie, the nest is so important that, even when it is abandoned, large chicks stay close to it while waiting at the colony for their parents. The parent comes directly to the location of the old nest and utters the display call which the chick answers immediately. Sometimes several chicks answer and try to be fed but the parent pecks the alien ones, feeding and preening its own. On rare occasions, mates or parents and chicks meet on the nest without calling (Aubin and Jouventin, personal observations with little blue, Ad41ie, and macaroni penguins). This occurs if no confusion is possible because of other behavioral cues and the use of the nest as a meeting place. In nonnesting penguins, the problem is more difficult because the parent has to find its partner with only vocal cues and without any landmarks. In addition, the call decreases in amplitude by half in 9 m and completely in about 15 m in this loud colony noise (Jouventin et aL, 1999; Lengagne et al., 1999b). In the king penguin, during the brooding stage, the problem is not so 256 THIERRY AUBIN AND PIERRE JOUVENTIN different from that of a nesting penguin because the brooder with its egg or small chick on its feet does not move more than 4 m (Lengagne et al., 1999a). Even if they build no nest, breeders lay in a hollow on the ground and they defend this against intruders by vigorous pecks. During the brooding stage, the mate returning from sea needed on average 5 calls and a mean of 2 rain (after the first call was emitted) to meet its moving partner in the colony (data established on the basis of observation of 28 pairs of birds, Lengagne et al., 1999a). When the incoming bird starts the acoustic search for its mate, the distance between the two birds was 8 m and 70% of the incubating birds were able to discriminate the first call emitted by the mate. During the rearing stage, when the chick is too large to fit on the adult's feet and can walk itself, it stays in the same feeding area some 10 m across and waits among many other chicks. Both parents memorize this place, termed rendezvous site by Stonehouse (1960) and attachment zone by Barrat (1976), and come directly to the edge of the chicks' huddles to call. The awaiting chick responds, usually within 15 s of the parent's first call (Stonehouse, 1960). Consequently, even without a nesting place, king penguins have a rendezvous site that assists their meeting considerably. In the emperor penguin, there is no nest, feeding place, or landmark and therefore the difficulty of finding the partner or the chick is greater. Birds seem to explore the colony at random, starting more often from the center of the colony and progressively expanding their search to the edge. In this species, the time necessary for an adult to find the chick can reach more than 2 h (Robisson, 1993a). Moreover a special behavioral adaptation exists to prevent the jamming of calls: when a bird calls, a neighbor that was about to call (as we can detect by its low head position) stops and waits for the end to utter its own call. In fact, to eliminate signal jamming, two individuals less than 7 m apart abstain from calling together. This "courtesy rule" of not interrupting was proved experimentally by playing a call from a loudspeaker positioned at different distances from a calling bird (Jouventin et al., 1979; Jouventin, 1982). 2. The Signaling Posture For efficient communication, the sender must maximize the propagation of the signal and the listener must optimize reception. Thus, in calling, penguins adopt particular body positions or "signaling postures." King penguins, and in general all penguin species except emperor penguins, raise their beaks slowly to a vertical position, stretch their necks to their fullest extent, and call (Jouventin, 1982). This signaling posture limits signal-to-noise ratio reduction caused by the screening effect of the bodies of the birds gathered in dense flocks. The macaroni penguin holds its head up, raising the beak, and then swings its head from side to side, beaming the signal in different VOCALLY IDENTIFYING KIN IN A CROWD: THE PENGUIN MODEL 257 directions in the colony. These calling attitudes, with the beak from 0.4 to I m above the ground, maximize the signal transmission distance in the colony (Robisson, 1991, 1993; Lengagne et al., 1999b). Finally, the emperor penguin remarkably directs its beak downward during calling probably to facilitate the control of the temporal pattern of this call (Jouventin, 1982). According to the measures of Robisson (1993a), another complementary explanation is that this unusual posture directs the signal forward, so that sound energy is beamed in the direction that the bird moves during the search for its partner or chick. Concerning the receiver, different postures are observed in the colony: rising to the feet or crouching, with back bowed and head lowered between the shoulders, are characteristic of incubating birds. Finally, especially during hot weather, birds can be observed lying on the ground. This last position is not an efficient one for receipt of acoustic information: we have shown in king penguin colonies (Lengagne et al., 1999b) that, when the receiver was located 10 cm above the ground, degradation of the signal was much more pronounced in both frequency and amplitude domains than when the receiver was located either 45 or 90 cm above ground. The characteristic listening postures (with head either 90 or 45 cm up) enhance the probability that the receiver will hear the call. Birds lying on the ground are generally found to be sleeping, so acoustic communication would seem to be of no importance to them. Conversely, when the incubating bird hears the call of its mate for the firs/; time, it leaves its incubating posture, rises to its feet, and assumes what we have shown to be the best position for signal reception. 3. The Searching Strategy Our signal propagation tests in colonies indicated that the communication system involving the mutual recognition between either mates or parents and chicks could be established in penguin species only at short range, as predicted by Falls (1982) for seabird colonies. Our playback tests with banded chicks or adults confirms Falls' hypothesis. To estimate the maximum distance at which a call can be recognized by the partner, natural calls were broadcast to the corresponding bird at different distances in the colony (Fig. 5). The experiment started at a distance from the bird to the loudspeaker of 8 to 20 m, depending on the species studied. The call was played back and the behavior of the bird was observed. Then, the distance was reduced by moving the loudspeaker closer. After a pause of 6 rain, the call was played back again, until a detection process was observed. Birds detect, recognize, and localize the natural display call at a relatively short distance: an average distance of 11 m for the king penguin (number of birds tested: N = 12; Aubin and Jouventin, 1998) and 4 m for the little blue penguin ( N = 7, Jouventin and Aubin, 2000). 258 THIERRY AUBIN AND PIERRE JOUVENTIN ,~' 1O0 E 75 m 50 o gh .o 25 t1# ; 0 20 19 18 . "ql T T 17 15 14 16 r/ f , , , , 13 12 11 10 I 9 I 8 I 7 Distance loudspeaker-bird (in In) FIG. 5. Maximal distance discrimination of the parental display call by the chick in king penguins. During experiments (12 chicks tested), the chick-to-loudspeaker distance, 20 m at the beginning, was progressively reduced by approaching the loudspeaker in steps of i m (data from Aubin and Jouventin, 1998, and Lengagne, 1999). When a bird comes from the sea and makes its way into the colony to find its partner, it calls regularly at different distances from the receiving bird. The farther from the receiving bird the acoustic search was initiated, the more time was necessary to complete the search and the greater the number of calls that were emitted by the incoming bird. On the other hand, we have shown in the king penguin that 70% of the birds started the acoustic search for their mate when the distance was less than or equal to the discrimination range (Lengagne et al., 1999b). So, the calling strategy adopted for finding the partner appears particularly efficient in penguins. B. SPECIES RECOGNITION Although we have not systematically studied species recognition in penguins, but have considered them rather a unique model for individual recognition, Jouventin (1982) did compare territorial behavior between species. We found that the display call has several biological meanings, being used by a single bird or by a pair throughout the breeding cycle to indicate both the species, the sex, and the individual. The nesting penguins are the more territorial species, particularly the burrowing genera (Spheniscus and Eudyptula). The latter is nocturnal and is strongly aggressive against intruders even if it cannot see them (Waas, 1988, 1991). Consequently this species, the little blue penguin, constitutes the best model for specific recognition in penguins. We compared it with a burrowing petrel (Jouventin and VOCALLY IDENTIFYING KIN 1N A CROWD: THE PENGUIN MODEL 259 Aubin, 2000), and tested their responses to playback experiments. The response is not tuned to a precise frequency analysis since entire calls that have been shifted strongly (200 Hz) up or down still elicit territorial responses. In fact, it appears that the little blue penguin pays attention to the lower sounds, the presence of high frequencies being unnecessary to elicit territorial responses. It appears also that both exhaled and inhaled temporal syllables provide territorial information. With respect to the territorial function of the calls, the coding process appears very simple: a territorial call consists of a rhythmic succession of two sounds having a particular pattern of frequency modulation. The parameters which encode territorial information are those that are resistant to degradation (low frequencies, slow FM, gaps in frequency and amplitude). To communicate the territorial message, the information represented by the patterned arrangement of the two "binary" units of sounds is highly redundant (Brackenbury, 1978), as it is based on the repetition of identical units of information at two levels. Nevertheless, it would be enough for a territorial function where the breeder has only to know that a conspecific is approaching and merely has to reply: "Keep out, this burrow is occupied!" C. A W E L L - M A T C H E D ACOUSTIC C O D E FOR IDENTIFICATION 1. The Cocktail-Party Effect As mentioned earlier, the display call is transmitted in a context involving the background noise generated by the colony plus the screening effect of the birds' bodies, both reducing the signal-to-noise ratio. In addition, the signal is masked by background noise with similar spectral and temporal characteristics. To estimate the minimal discrimination threshold of the display call in a jamming situation, a series of mixed signals was broadcast to penguin chicks (Fig. 6). The parental call was combined with five extraneous adult calls with different emergence levels, the tested signal increasing in energy ratio among the extraneous noise. The superimposition results in a mixed signal with a total lack of silences and with numerous frequencies overlapping. This jamming mimics a situation frequently observed in a penguin colony. Chicks of three species were tested with signals of different emergence levels (i.e., the difference between the energy level of the tested call and that of the five extraneous calls) at a distance of 7 m. The emergence level was defined as E = 20 log Ap/Ae, where E represents the emergence level in dB of the parental call of the chick tested, Ap is the absolute amplitude of the parental call, and Ae is the absolute amplitude of the mixed extraneous 260 THIERRY AUBIN AND PIERRE JOUVENTIN 4 ~3 [] Emperor I Ad~lie 2 .~ [] King j, 0 -9 -6 -3 0 3 6 Emergence level (dB) Fro. 6. Test ofdetectionoftheparentalcallbythe chick in three penguin species in ajamming situation. The parental display call is mixed with five extraneous display calls to yield different emergence levels. calls. Our experiments indicate that the chick can detect its parental call in an extreme jamming situation. With king penguin chicks, detection is possible even when the parental call intensity is well below ( - 6 dB) that of the noise of simultaneous calls produced by other adults (Aubin and Jouventin, 1998). At the same distance, the Ad61ie and emperor penguins are not as good at detecting Calls as the king penguin chick. Nevertheless, Ad61ie and emperor penguin chicks have a good ability to recognize the parental call even embedded in the noise of the colony (0 dB of emergence, Jouventin and Aubin, 2001). This capacity tO perceive and extract the information from an ambient noise with similar acoustic properties to that of the signal, termed the "cocktail-party effect" in speech intelligibility tests, enhances the chick's ability to find its parents. This process of perception must be linked to an acoustic coding system adapted to the constraints of colonial life. 2. The Vocal Signature By playing back natural calls, it has been demonstrated that individual recognition by voice exists in all the species of penguins that have been studied (Derenne et aL, 1979; Jouventin et aL, 1979; Jouventin, 1982; Waas, 1988; Speirs and Davis, 1991). Previous analysis of temporal and frequency parameters have shown effectively that an individual signature can also be found in the display calls of all of these species (Jouventin and Roux, 1979; Robisson et aL, 1989; Brdmond et aL, 1990; Robisson, 1992a; Robisson et al., 1993; Lengagne et aL, 1997; Jouventin and Aubin, 2000). This has mainly been done by comparing, for a given parameter, the betweenindividual variation and the within-individual variation (Jouventin, 1982; 261 V O C A L L Y I D E N T I F Y I N G KIN IN A C R O W D : T H E P E N G U I N M O D E L Jouventin and Aubin, 2000). By comparing the temporal and frequency patterns in the calls of two species of penguins with nests (macaroni and Ad61ie) with those of two species without a nests (king and emperor), a direct relationship was shown between the potential of individual coding and the difficulty of finding the partner in the colony (Lengagne et al., 1997). However, knowing that it is possible to distinguish the signals of individuals by analysis does not tell us whether the birds do it. The only way to investigate this process was to test the birds by playing back different kinds of experimental signals. These signals corresponded to natural display calls modified in different ways: modifications of amplitude and frequency modulations (AM and FM), modifications of frequency and temporal parameters, and modification of the two-voice system (Table II). T A B L E II RESPONSES OF PENGUINS TO EXPERIMENTAL SIGNALS CORRESPONDING TO NATURAL CALLS MODIFIED Species Signals EP KP MP AP GP -÷ + ÷÷ --- -÷ +4- -4++ -+ ÷+4- Fo Low Pass High Pass S h i f t 4- 25 S h i f t 4- 50 nt ÷÷ ÷ nt +4- + ÷+ -nt ÷÷ nt ++ ÷÷ ÷÷ ++ -++ -÷ -- ---++ -- Shift ± 75 Shift 4 - 1 0 0 +÷ -- 4-- +-t+ -nt -nt ÷ -nt nt +÷ +÷ ++ -- +4+ -nt ÷÷ +÷ -nt nt ++ -nt -- -- nt nt nt Modulation Without AM Without FM Reversed Frequency domain Temporal domain Half Call OneSyl HalfSyl QuarterSyl Two-voices phenomenon One voice suppressed Note: + ÷ , s t r o n g r e s p o n s e ; + , m o d e r a t e r e s p o n s e ; - - , n o r e s p o n s e ; n t , n o t t e s t e d . ER KE ME AP, GP: respectively, emperor, king, macaroni, Addlie, and gentoo penguins. ( R e s u l t s f r o m : J o u v e n t i n , 1982, A u b i n et al., 2 0 0 0 , H i l d e b r a n d , u n p u b l i s h e d d a t a , f o r t h e e m p e r o r p e n g u i n ; J o u v e n t i n et al., 1999, L e n g a g n e et aL, 2000, f o r t h e k i n g p e n g u i n ; J o u v e n t i n a n d A u b i n , 2001, f o r t h e Ad61ie a n d g e n t o o p e n g u i n s ; A u b i n and Jouventin, unpublished data, for the macaroni penguin.) 262 THIERRY AUBIN AND PIERRE JOUVENTIN a. A M and FM. Concerning the modulations, only the king penguin among the different species studied, continues to respond after the total elimination of amplitude modulation (AM) in its natural display call. This does not imply that the AM structure is without use in this species. AM is strongly degraded during propagation through the colony (Aubin and Mathevon, 1995, Aubin and Jouventin, 1998; Lengagne et al., 1999b) and this is unlikely to carry a message at distance. Nevertheless, we cannot rule out the possibility that AM might transmit information which would allow estimating the distance of the emitter (Naguib, 1996) or locating the acoustic source (Konishi, 1973; Wiley and Richard, 1982), two helpful qualities for locating an individual in a crowd. Concerning frequency modulation (FM), all the species studied except the king penguin are able to identify the mate or parental call even if the natural FM is lacking. It is the same for signals where each syllable (and at the same time the FM) is reversed (Fig. 7). Thus FM appears to be a key parameter in individual recognition for the king penguin but not for the other species, where amplitude modulation is important. b. The Frequency Domain. In the frequency domain, all of the species studied prefer the lower part of the sound spectrum to the higher one, probably due to the fact that high frequencies attenuate strongly during propagation in the colony (Aubin and Jouventin, 1998; Lengagne et al., 1999b). Nevertheless, some species (Ad61ie and gentoo) pay greater attention to the spectral profile of the call than others (emperor, king, macaroni). For example, recognition occurs for low-pass calls in emperor, king, and macaroni penguins but not for Ad61ie and gentoo penguins. Pitch parameters reveal the same pattern of responses. Emperor, king, and macaroni penguins tolerate more change in frequency than do Ad61ie and gentoo penguins. To identify its partner or chick, the latter species need the right frequency values for the harmonics of the call (with an accuracy of 25 Hz), whereas the former species tolerate errors of 75-100 Hz (Fig. 8). Thus, for Ad61ie and gentoo penguins, the coding of identity depends on the analysis of the spectral profile and of the precise frequency values of the harmonics, that is, mainly on timbre analysis. c. The Temporal Domain. In the great majority of species studied, the broadcast of only one syllable is sufficient to elicit recognition, and this recognition is not linked to any particular syllable in the call (Jouventin and Aubin, 2001; Jouventin et al., 1999; Lengagne et al., 2000, and unpublished data). For the king penguin, even the first half of a syllable is sufficient to elicit recognition. In this species, the basic FM structure of the syllable is always the same: an increase followed by a decrease in frequency, with the inflection point always in the first half of the syllable. This inflection point is necessary for the recognition of the signal. The small amount of information (about 200 ms) provided by the frequency modulation shape of this half of VOCALLY iDENTIFYING KIN IN A CROWD: THE PENGUIN MODEL ® ® '3klh '3kFlz EmperorPenguin 263 Is ® ® f KingPenguin Fro. 7. Natural and corresponding "syllable reversed" calls for two species of penguins that do not have a nest• Only the king penguin signal with syllable reversed does not elicit positive responses during individual recognition tests• the syllable is sufficient to provide individual information (Jouventin et aL, 1999; Lengagne et al., 2000). At the opposite extreme, the display call of the emperor penguin is the only one where several successive syllables (at least three; Hildebrand, personal communication) are needed to elicit recognition. Effectively, to identify the signal, emperor penguins examine the temporal succession of syllables (Jouventin, 1972, 1982; Jouventin et al., 1979). Display calls of penguins (emperor penguin calls included) appear to be highly redundant, consisting of more or less identical successive syllables with a repetition of the same information many times. This redundancy 0 C~ I IIHillnl III I c~ + © © ~© ~.~ 265 VOCALLY IDENTIFYING KIN IN A CROWD: THE PENGUIN MODEL 14- ,0@ 12- 10 8 e~ . . . . . . . O Z 4 0 4 i i i i e i t 5 6 7 8 9 10 11 Wind speed (m/s) FIG. 9. Enhancement of the number of syllables per call emitted by king penguins as wind speed increases: in windy conditions, birds maintain the efficiency of their communication by increasing the redundancy of the signal. The circle size is proportional to the number of data (ranging from 3 to 6) and the dashed line represents least-squares piecewise linear models (data replotted from Lengagne et al., 1999b). enhances the opportunity to find a quieter window in the continuous noisy environment of a seabird colony. In addition, in penguin colonies circumpolar winds blow strongly, generating a high level of background noise and consequently lowering the signal-to-noise ratio in the colony. In windy conditions, we have observed that birds in king penguin colonies (Lengagne et al., 1999c) maintain the efficiency of their communication by increasing the number of syllables per call, leading to an enhancement of the duration of the calls. From a wind speed of 8 m/s, the duration of the calls increases linearly as the wind speed increases (Fig. 9). Thus, penguins seem to take into account the constraints imposed by wind on their communication. This result conforms with predictions from information theory (Shannon and Weaver, 1949) that increased redundancy in signal improves the probability of receiving a message in a noisy channel. d. The Two-Voice System. The sound-producing structure in birds is the syrinx, usually a two-part organ located at the junction of the bronchi. As each branch of the syrinx produces sound independently, many birds have two acoustic sources. The use of the two voices was first documented in the song of the brown thrasher, Toxostoma rufum (Potter et aI., 1947), and has been actively investigated for the past fifty years. Anatomical, physiological, and acoustical evidence existed for this two-voice phenomenon 266 THIERRY AUBIN AND PIERRE JOUVENTIN (Greenewalt, 1968) but no function for it was known (Sturdy and Mooney, 2000). In songbirds, these two voices with their respective harmonics are often not activated simultaneously but two voices are obvious in the large penguins and generate a beat pattern that varies between individuals. Among the 17 species of penguins, only the Aptenodytes genus employs two frequency bands (Robisson, 1992b). Both Aptenodytes species, the emperor and the king penguins, produce a signal consisting of two simultaneous series of harmonically related bands of slightly differing frequencies (on average 65 Hz for the emperor, Aubin et al., 2000, and 25 Hz for the king, Robisson, 1992b), resulting in a two-voice call that produces audible "beats." The two-voice system appears well suited for the environment in which it is used. We have done experiments with two voice signals broadcast through the colony and recorded at different distances (1, 8, 16 m). The recorded signals were then analyzed and the amount of degradation between the amplitude modulation of the beats and the true modulation existing in the call itself were compared (Fig. 10). We found that, although the true amplitude beats (a) 800 (b) 300 I I I I instantaneous frequency (I-filbert calculation) 100 m s Fr~. 10. King penguin display call: analysis of beats generated by the "two-voice system." Only the fundamentals of one syllable are analyzed here. The periodic amplitude fluctuations [(a) oscillographic representation] coincide with the periodic frequency discontinuities [(b) spectographic + instantaneous frequency representation] and allow quantification of the period of the beats generated by the two sources. VOCALLYIDENTIFYINGKIN IN A CROWD:THE PENGUINMODEL 267 m o d u l a t i o n of the call was severely c o m p r o m i s e d by p r o p a g a t i o n , the amplitude m o d u l a t i o n of the beats p r o d u c e d by the two voices r e m a i n e d largely unchanged. O u r experiments d e m o n s t r a t e that the beats g e n e r a t e d by the interaction of these two f r e q u e n c y bands p r o p a g a t e well t h r o u g h obstacles, being robust to s o u n d d e g r a d a t i o n t h r o u g h the m e d i u m of bodies in a p e n g u i n colony, but a b o v e all that t h e y c o n v e y i n f o r m a t i o n a b o u t individual identity. To test the hypothesis that the two voices m a y play a key role in individual recognition in the Aptenodytes genus we designed a series of p l a y b a c k experiments a) b) 1.5 kHz 3 kHz 0 0 I,,,,i 1.5 kHz 0 lu ill ,ll i | ~. i,i 0.Ss 0.Ss c) d) 3 kHz 0 Fro. 11. Sound spectrograms (1024 points window size) of syllables of emperor (left) and king (right) penguins. (a) and (b): low-pass filtered natural syllable, with only the fundamentals and the first harmonics kept; (c): one voice removed by filtration in the emperor penguin signal; (d): synthetic signal built on the model of the low-pass filtered king penguin signal, but with only one voice. The low-pass signals with two voices are recognized but not the one-voice signals. 268 THIERRY AUBIN AND PIERRE JOUVENTIN [ "'Without'Nest' [ [ _iWith Nest".~ King penguin Addlie penguin L1 Signal detection: - 6 dB 0 dB ~1Signal location: High High ~l SifnM redundant: High High ,i J FM & Beats (Temporal pattern analysis) l:l Sienal coding-decoding: Complex Timbre (Speetralanalysis) Simple FIG. 12. Signal adaptation to recognition in the noise (>70 dB) in two categories of penguins: without nest (e.g., king penguin) and with nest (e.g., Ad61ie penguin) species. with signals where only one voice was kept (Aubin et al., 2000, for the emperor penguin, Lengagne et al., 2001, for the king penguin). Unfortunately, the upper frequency bands of natural calls were not spaced sufficiently to allow for removal, by simple filtration, of one of the two voices. Therefore, a preliminary playback test was performed with the two Aptenodytes species using only the lower frequency component of the calls as stimulus (Fig. 11). VOCALLY IDENTIFYING KIN IN A CROWD: THE PENGUIN MODEL 269 When presented with the lower frequency component of the calls, positive responses were induced in chicks and adults for both species. When one voice in the call is experimentally suppressed, either by filtration (Aubin et al., 2000, for the emperor penguin) or by synthesizing a one-voice signal (Lengagne et al., 2001, for the king penguin), no response is observed for adults or chicks. Clearly, the acoustic contribution from two voices is required for call recognition in the genus Aptenodytes. This coding process, increasing the call complexity and resisting sound degradation, appears to have evolved in parallel with the loss of territoriality. e. Acoustic Solutions to Ecological Problems. If we summarize the results obtained in our playback experiments with modified display calls, two acoustic code categories emerge in the penguin family (Fig. 12). The first one is elementary and concerns the nesting penguins, such as the Ad61ie and gentoo. These species identify the partner by analyzing the spectral profile and pitch of the call (timbre analysis). The second one is complex and concerns the two nonnesting penguins, the emperor and the king. Both these species use a vocal signature in the time domain based on an amplitudetime (AM) analysis for the emperor penguin and on a frequency-time (FM) analysis for the king penguin. This temporal analysis is complemented by another sophisticated system: the beats generated by the two-voice system. The temporal pattern of syllables associated with the two-voice system creates a huge variety of vocal signatures. This is necessary to distinguish between several thousand birds breeding without nests, that is, without visual cues. At the opposite extreme, identification by a one-dimensional parameter such as the timbre does not offer such an impressive variety of vocal signatures, and, thus, the possibility of confusion should exist with this system. Nevertheless, in nesting penguins, the nest is used as a meeting place, even when chicks have fledged, so that the probability that a bird emits the right call at the wrong place is weak. This strongly limits the possibilities for confusion. V. A. PERSPECTIVES FUNCTIONALS The main perspective from these results is functional because these vocal signatures represent external markers of identity allowing the penguins to find their chicks or their parents in a crowd, that is, to find the kin-related birds with a single sound system when olfactory and, more surprisingly, visual cues are not used. During the brooding phase, the parent coming back from the sea has first to find its mate by its vocal signature if it is to find the egg 270 THIERRY AUBIN AND PIERRE JOUVENTIN or the small chick. Then the detection of the parent by the chick has an obvious survival value because usually parents feed only their own chick(s). The identification of the chick by the parent allows it to invest in its own chick and consequently to propagate its genes. But the general pattern is not symmetrical because, although a chick may gain from getting extra food from birds other than its parents, parents have to be cautious to give food only to their own chick(s). Consequently we have observed chicks trying to obtain "extra-feeding" from adults calling differently from their parents, sometimes with success when confusion occurs, for example, after a storm or when parents cannot find their own young (see Jouventin et al., 1995). We have also observed parents chasing away these "robbers" after hearing their call accurately. B. AcousTIcs According to the theory, to extract a signal from the background noise penguins analyze either frequency bands or temporal patterns (AM, FM, beats) of the call. The first coding-decoding system is used by nesting penguins such as the Ad61ie, the gentoo, and, in the first part of its call, the macaroni, and the second one is used by nonnesting penguins. The two codes seem not to be equivalent in efficiency. Frequency analysis is known to be particularly slow in a physical sense (Pimonow, 1962; Beecher, 1988) as well as physiologically (Bregman, 1978): when the duration of the analysis decreases, the uncertainty in the measurement of frequency increases. Accurate analysis of frequency is more time consuming than analysis in the time domain. The two codes seem also not to be equivalent in terms of production. Modulation in time is difficult to produce (Gaunt et al., 1973; Brackenbury, 1982): frequency or amplitude modulated calls require a high degree of control of the two sound sources of a bird. Nevertheless, this acoustic signal is particularly efficient in allowing an animal to locate the partner immediately in a noisy crowd on the move. Briefly, frequency analysis (easy to produce but costly in terms of analysis time) is enough to solve the relatively simple problem of individual recognition in nesting birds, whereas the complex temporal analysis of modulations used by the two nonnesting penguins (quick to analyze but Costly to produce) appears as an adaptation to extreme acoustic and breeding conditions. C. COMPARATIVE APPROACH The case of the macaroni penguin, a nesting penguin, is perhaps intermediate between the two systems just described. Effectively, according to our first playback experiments, this species seems to identify the partner VOCALLY IDENTIFYING KIN IN A CROWD: THE PENGUIN MODEL 271 on the basis of a simultaneous analysis of the spectral profile and of the temporal succession of syllables. However, this last parameter, again for a nesting species, seems simple involving only a rhythm fixed (= a number), whereas the emperor penguin has a more complex system using a succession of syllable-silence timings (= a "bar code"). Complementary experiments will be necessary to fully understand the coding-decoding process of the macaroni penguin and to compare with other species in this newly studied genus (Eudyptes). Vl. CONCLUSION Highly vocal and colonial, penguins communicate in a particularly constraining acoustic environment to find their mate, parent, or chick. The screening effect of penguin bodies and the loud background noise prevent long-range communication and explain why some sophisticated adaptations have appeared, such as the highly developed "cocktail party effect" in the king penguin or the "courtesy rule" preventing jamming in the emperor penguin. These constraints explain the existence of complementary and visual behaviors optimizing vocal identification, such as the "head up" position when hearing or calling, particularly in the king penguin, and the memory of a feeding place in nesting penguins and in king penguins. They explain also the physical properties of the calls. For species recognition, it seems that the penguins are not more sophisticated than other birds (see Becker, 1982, for a review). Thus, our playback experiments with the little blue penguin show that the acoustic code used for species identification is elementary. But, for individual recognition, it is surprising to find animals that are able to see nevertheless identifying their parents only by voice, in a noisy crowd, and even for some species without the help of a nest as a landmark. In the same family, natural selection has resulted in various acoustic adaptations which are all the more sophisticated given that the problem posed was difficult to solve. In the nesting penguins where visual cues help vocal recognition, we found two main identification systems based on a single parameter, frequency, in the two species of Pygoscelis and on two elementary parameters, temporal and frequency patterns, in one species of Eudyptes. The vocal solutions found in closely related species are really different and demonstrate once more that behavior can evolve fast. It will be interesting to know what systems occur in other species of nesting penguins, not yet studied, to see if the number of acoustic identification systems is even larger. In the nonnesting penguins, once again two closely related species have two different vocal identification systems. This dual system is much more 272 THIERRY AUBIN AND PIERRE JOUVENTIN complex than in nesting species using topographical cues. The first identification system is based mainly on the variation of frequency in time for the king penguin and on the temporal syllable pattern in the emperor penguin. The second system relies on the two voices. It is common to the genus Aptenodytes, and this double-voice system is completely new because, although we have known for 30 years of the existence of this anatomical, physiological, and physical phenomenon, no function for it was known. In fact, these new findings on a single family, although exceptional by the variety of its breeding habits, are all the more interesting since we have poor knowledge of the systems of individual identification in birds, particularly when they are not songbirds. Individual recognition playbacks are particularly difficult to carry out because to test a bird we have to present a particular call to each, whereas we need only one call type in the study of species recognition. It remains to be seen whether we can extrapolate our results to other penguins and then more widely among seabirds. We can certainly learn from this penguin study that nonsongbirds can be sophisticated in their vocal adaptations and deserve more attention, being currently relatively little studied. Enlarging upon the scope of the acoustic systems, it is also possible to study individual recognition beyond birds, for example, in mammals, where identification strategies by voice have been found in species such as fur seals (Roux and Jouventin, 1987; Charrier et al., manuscript in preparation). Further studies may reveal interesting parallels in the acoustic identification systems between birds and mammals. VII. SUMMARY In penguins, individual recognition is observed between mates and between parents and chick(s). During the past five years, their particular strategies of coding-decoding have been tested by playing back modified display calls to six species, in Australia (little penguin, Eudyptula minor), in Antarctica (Addlie penguin, Pygoscelis adeliae; emperor penguin, Aptenodytes forsteri), and in subantarctic islands (king penguin, Aptenodytes patagonicus; macaroni penguin, Eudyptes chrysolophus; gentoo penguin, Pygoscelis papua). All species use only vocal cues to identify their partner, but in territorial species the nest is used as a meeting point. In large species, such as the king and the emperor penguins, which do not have a nest, the brooder carries the egg or the small chick on the feet, while the mate, and then the chick, has to be located in the noisy colony without any topographical cue. According to theory, to extract a signal from background calls, animals analyze either frequency bands or modulations (amplitude and frequency VOCALLY IDENTIFYING KIN IN A CROWD: THE PENGUIN MODEL 273 modulations) of the partner's call. The first coding-decoding system, used by nesting penguins, is easy to produce but costly in terms of analysis time. The second one, used by nonnesting penguins, is a vocal signature which is fast to analyze but costly to produce. This acoustic signal is particularly efficient as a means to locate immediately the partner on the move in a noisy crowd. Briefly, frequency analysis is enough to solve the relatively easy problem of individual recognition in nesting birds, while the complex temporal analysis of modulations of the two nonnesting penguins is an adaptation to extreme acoustic and breeding conditions. The macaroni penguin, which we have begun to test, seems to use both a frequency code similar to that of the other nesting species and a temporal code close to the one of a nonnesting penguin species, but much simpler. Acknowledgments This study was funded in antarctic and subantarctic areas by the Institut Fran~ais pour la Recherche et la Technologie Polaires (Program N ° 109 then N ° 354). For the study of the little penguin in Australia, 'we thank the Committee of Management of the Penguin Reserve and particularly Peter Dann. We thank also Stephen Dobson, Tim Roper, Peter Slater, and Charles T. Snowdon for helpful comments and English improvements. References Alcock, J. (1972). "Animal Behavior, an Evolutionary Approach." Sinauer, Sunderland, MA. Aubin, T. (1994). Syntana: A software for the synthesis and analysis of animal sounds. Bioacoustics 6, 80-81. Aubin, T., and Jouventin, R (1998). Cocktail-party effect in king penguin colonies. Proc. R. Soc. London B 265, 1665-1673. Aubin, T., and Lengagne, T. (1997). Reconnaissance du cri du parent par le poussin du manchot royal Aptenodytes patagonicus dans le milieu bruyant de la colonie. Bull Soc. ZooL Fr. 123, 267-277. Aubin, T., and Mathevon, N. (1995). Adaptation to severe conditions of propagation: Long-distance distress calls and courtship calls of a colonial seabird. Bioaeoustics 6, 153161. Aubin, T., Jouventin, E, and Hildebrand, C. (2000). Penguins use the two-voice system to recognise each other. Proc. R. Soc. London B 267, 1081-1087. Barrat, A. (1976). Quelques aspects de la biologie et de l'6cologie du manchot royal Aptenodytes patagonicus des ~les Crozet. Com. Nat. Fr. Rech. Ant. 40, 107-147. Becker, E H. (1982). The coding of species-specific characteristics in bird sounds. In "Acoustic Communication in Birds" (D. E. Kroodsma and E. H. Miller, Eds.), Vol. I, pp. 131-181. Academic Press, New York. Beecher, M. D. (1988). Spectrographic analysis of animal vocalizations. Implications of the "uncertainty principle." Bioacoustics 1, 187-208. 274 THIERRY AUBIN AND PIERRE JOUVENTIN Beer, C. G. (1970). Individual recognition of voice in the social behaviour of birds. Adv. Study Behav. 3, 27-74. Beer, C. G. (1979). Vocal communication between laughing gull parents and chicks. Behaviour 70, 118-146. Brackenbury, J. H. (1978). A comparison of the origin and temporal arrangement of pulsed sounds in the songs of the Grasshopper and Sedge warblers, Locustella naevia and Acrocephalus schoenobaenus. J. Zool. London 184, 187-206. Brackenbury, J. H. (1982). The structural basis of voice production and its relationship to sound characteristics. In "Acoustic Communication in Birds" (D. E. Kroodsma and E. H. Miller, Eds.), Vol. 1, pp. 53-73. Academic Press, New York. Bregman, A. S. (1978). The formation of auditory streams. In "Attention and Performance" (J. Reguin, Ed.). Erlbaum, Hillsdale, NJ. Br6mond, J. C., and Aubin, T. (1992). The role of amplitude modulation in distress call recognition by the black-headed gull (Larus argentatus). Ethol. Ecol. Evol. 4, 187-191. Br6mond, J. C., Aubin, T., Nyamsi, R. M., and Robisson, P. (1990). Le chant du manchot empereur (Aptenodytesforsteri): Recherche des parambtres utilisables pour la reconnaissance individuelle. C R. Acad. Sci. 311, 31-35. Bried, J., and Jouventin, P. (2001). Site and mate-choice in seabirds: An evolutionary approach to the nesting ecology. In "Biology of Marine Birds" (J. Sehreiber and R. Burger, Eds.) CRC Press, Boca Raton, FL. In press. Brooke, M. L. (1978). Sexual differences in the voice and individual vocal recognition in the manx shearwater (Puffinus puffinus). Anita. Behav. 26, 622-629. Charrier, I., Jouventin, P., Mathevon, N., and Aubin, T. (2001). Acoustic communication in a black-headed gull colony: How to identify their parents? Ethology In press. Dabelsteen, T. (1981). The sound pressure level in the dawn song of the blackbird Turdus merula and a method for adjusting the level in experimental song to the level in natural song. Z. Tierpsychol. 56, 137-149. Dabelsteen, T., Larsen, O. N., and Pedersen, S. B. (1993). Habitat-induced degradation of sounds signals: Quantifying the effects of communication sounds and bird location on blur ratio, excess attenuation, and signal-to-noise ratio in blackbird song. J. Acoust. Soc. Am. 93, 2206-2220. Derenne, P., Jouventin, P., and Mougin, J. L. (1979). Le chant du manchot royal (Aptenodytes patagonicus) et sa signification 6volutive. Le Gerfaut 69, 211-224. Falls, J. B. (1982). Individual recognition by sound in birds. In "Acoustic Communication in Birds" (D. E. Kroodsma and E. H. Miller, Eds.), Vol. II, pp. 237-278. Academic Press, New York. Gaunt, A. S., Stein, R. C., and Gaunt, S. L. (1973). Pressure and air flow during distress calls of the starling, Sturnus vulgaris (Ayes: Passeriformes). J. Exp. Zool. 183, 241-262. Greenewalt, C. H. (1968). "Bird Song: Acoustics and Physiology." Random House (Smithsonian Inst. Press), New York. Isenmann, P., and Jouventin, R (1970), Eco-6thologie du Manchot Empereur (Aptenodytes forsteri) et comparaison avec le Manchot Ad61ie (Pygoscelis adeliae) et le Manchot Royal (Aptenodytes patagonica ). L 'Oiseau R.E O. 40, 136-159. Jouventin, P. (1972). Un nouveau syst6me de reconnaissance acoustique chez les oiseaux. Behaviour 43, 176-186. Jouvenfin, P. (1982). "Visual and Vocal Signals in Penguins, Their Evolution and Adaptive Characters" (P. Parey, Ed.). Vol. 24, of the Advances in Ethology series. Jouventin, P., and Aubin, T. (2000). Acoustic convergence between two nocturnal burrowing seabirds: Experiments with a penguin (Eudyptula minor) and a shearwater (Puffinus tenuirostris). Ibis 142, 645-656. VOCALLY IDENTIFYING KIN IN A CROWD: THE PENGUIN MODEL 275 Jouventin, E, and Aubin, T. (2001). Experimental study of the kin-recognition in two penguins (Pygoscelis adeliae andpapua). Acoustic systems in relation with nesting ecologies. J. Avian Biol. Submitted. Jouventin, R, and Cornet, A. (1980). The sociobiology of Pinnipeds. Adv. Study Behav. 11, 121-141, Jouventin, R, and Robin, J. E (1984). Olfaction experiments on some antarctic birds. Emu 84, 46-48. Jouventin, R, and Roux, E (1979). Le chant du Manchot Ad61ie (Pygoscelis adeliae). R~5tedans la reconnaissance individuelle et comparaison avec le Manchot Empereur non-territorial. L'Oiseau R.EO, 49, 31-43. Jouventin, E, and Weimerskirch, H. (1990). Satellite tracking of wandering albatrosses. Nature 343, 746-748. Jouventin, R, Guillotin, M., and Cornet, A. (1979). Le chant du Manchot Empereur et sa signification adaptative. Behaviour 70, 231-250. Jouventin, E, Barbraud, C., and Rubin, M. (1995). Adoption in the emperor penguin Aptenodytes forsteri. Anita. Behav. 50, 1023-1029. Jouventin, R, Aubin, T., and Lengagne, T. (1999). Finding a parent in a king penguin colony: The acoustic system of individual recognition. Anim. Behav. 57, 1175-1183. Kodyman, G. L., and Mullins, L L. (1990). Ross Sea emperor penguin breeding populations estimated by aerial photography. In "Antarctic Ecosystems: Ecological Change and Conservation" (K. R. Kerry and G. Hempel, Eds.), pp. 169-176. Springer-Verlag, Berlin. Konishi, M. (1973). Locatable and nonlocatable acoustic signals for Barn Owls. Am. Nat. 107, 775-785. Lack, D. (1968). "Ecological Adaptations for Breeding in Birds." Methuen, London. Lengagne, T. (1999). Communication acoustique en milieu colonial: Le reconnaissance individuelle chez le manchot royal Aptenodytes Patagonicus. Ph.D. thesis, Universit6 de Toulouse. Lengagne, T., Lauga, J., and Jouventin, E (1997). A method of independent time and frequency decomposition of bioacoustic signals: Inter-individualrecognition in four species of penguins. C. R. Acad. Sci. Paris, Sciences Vie 320, 885-891. Lengagne, T., Jouventin, R, and Aubin, T. (199%). Finding one's mate in a king penguin colony: Efficiency of acoustic communication. Behaviour 136, 533-546. Lengagne, T., Aubin, T., Jouventin, E, and Lauga, J. (1999b). Acoustic communication in a king penguin's colony: Importance of bird location within the colony and of the body position of the listener. Polar Biol. 21, 262-268. Lengagne, T., Aubin, T., Lauga, J., and Jouventin, R (1999c). How do king penguins Aptenodytes patagonicus apply the Mathematical Theory of Information to communicate in windy conditions? Proc. R. Soc. London Ser. B 266, 1623-1628. Lengagne, T., Aubin, T., Jouventin, E, and Lauga, J. (2000). Perceptual saiience of individually distinctive features in the calls of adult king penguins. J. Acoust. Soc. Am. 107, 508-516. Lengagne, T., Lauga, J., and Aubin, T. (2001). Intra-syllabic acoustic signatures used by the king penguin in parent-chick recognition: An experimental approach. J. Exp. Biol. 204, 663-672. Lequette, B., Verheyden, C., and Jouventin, R (1989). Olfaction in subantarctic seabirds: Its phylogenetics and ecological significance. Condor 91, 732-735. Mbu-Nyamsi, R. G., Aubim T., and Br6mond, 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. Naguib, M. (1996). Ranging by song in Carolina wrens: Effects of environmental acoustics and strength of song degradation. Behaviour 133, 541-559. 276 THIERRY AUBIN AND PIERRE JOUVENTIN Noll, A. M. (1967). Cepstrum pitch determination. J. Acoust. Soc. Am. 41, 293-309. Papoulis, A. (1977). "Signal analysis." McGraw-Hill, New York. Pimonow, L. (1962). "Vibrations en rdgime transitoire: Analyse physique et physiologique." Dunod, Paris. Potter, R. K., Kopp, G. A., and Green, H. C. (1947). "Visible Speech." Van Nostrand, New York. Press, W. H., Flannery, B. E, Teukolsky, S. A., and Vetterling, W. T. (1988). "Numerical Recipes in C. The Art of Scientific Computing." Cambridge Univ. Press, Cambridge, UK. Prdvost, J. (1961). "Ecologic du Manehot Empereur." Hermann, Paris. Prdvost, J., and Bourli6re, E (1957). Vie sociale et thermor6gulation chez le manchot empereur. Alauda 25, 167-173. Randall, R. B., and Tech, B. (1987). "Frequency Analysis." Bruel & Kjaer, Naerum. Robisson, E (1991). Broadcast distance of the mutual display call in the emperor penguin. Behaviour 119, 302-316. Robisson, P. (1992a). Roles of pitch and duration in the discrimination of the mate's cali in the king penguin Aptenodytes patagonicus. Bioacoustics 4, 25-36. Robisson, R (1992b). Vocalizations in Aptenodytes penguins: Application of the two-voice theory. Auk 109, 654-658. Robisson, E (1993a). La reconnaissance individuelle chez deux esp6ces jumelles, le manchot empereur Aptenodytes forsteri et le manchot royal Aptenodytes patagonicus. Ph.D. thesis, Universit6 de Rennes. Robisson, R (1993b). Adaptation du transfert de l'information individuelle en milieu colonial chez les manchots. Terre Vie 48, 133-141. Robisson, E, Aubin, T., and Br6mond, J. C. (1989). La reconnaissance individuelle chez le manchot empereur (Aptenodytes forsteri): R61es respectifs du ddcoupage temporel et de la structure syllabique du chant de cour. C. R. Acad. Sci. 309, 383-388. Robisson, E, Aubin, T., and Br6mond, J. C. (1993). Individuality in the voice of the emperor penguin Aptenodytes forsteri: Adaptation to a noisy environment. Ethology 94, 27% 290. Roux, J. E, and Jouventin, E (1987). Behavioral cues to individualrecognition in the subantarctic fur seal, Arctocephalus tropicalis. N.O.A.A.N.M.ES. 51, 95-102. Scharf, B. (1970). Critical bands. In "Foundations of Modern Auditory Theory" (J. V. Tobias, Ed.), pp. 159-202. Academic Press, New York. Shannon, C. E., and Weaver, W. (1949). "The Mathematical Theory of Communication." Univ. Illinois Press, Urbana. Speirs, E. A. H., and Davis, L. S. (1991). Discrimination by Addlie penguins between the loud mutual calls of mates, neighbours and strangers. Anita. Behav. 41, 937-944. Stonehouse, B. (1960). The king penguinAptenodytespatagonicusof South Georgia. 1. Breeding behaviour and development. Falk. Is. Dep. Surv. Sci. Rep. 23. Sturdy, C. B., and Mooney, R. (2000). Bird communication: Two voices are better than one. Curr. Biol. 10, 634-636. Trivets, R. L. (1972). Parentalinvestment and sexual selection. In "Sexual Selection and Descent of Man" (B. Campbell, Ed.), pp. 136-179. Aldine, Chicago. Verheyden, C., and Jouventin, E (1994). Olfactory behavior of foraging procellariiforms. Auk 11I, 285-291. Waas, J. R. (1988). Acoustic displays facilitate courtship in little blue penguin Eudyptula minor. Anita. Behav. 36, 366-371. Waas, J. R. (1991). Do little blue penguins signal their intentions during aggressive interactions with strangers? Anita. Behav. 41, 375-382. VOCALLY IDENTIFYING KIN IN A CROWD: THE PENGUIN MODEL 277 White, S. J., and White, R. E. C. (t970). Individual voice production in gannets. Behaviour 37, 4t-54. Wiley, R. H., and Richards, D. B. (1978). Physical constraints on acoustic communication in the atmosphere: Implication for the evolution of animal vocalizations. Behav. EcoL Sociobiol. 3, 69-94. Wiley, R. H., and Richards, D. G. (1982). Adaptations for acoustic communication in birds: Transmission and signal detection. In "Acoustic Communication in Birds" (D. E. Kroodsma and E. H. Miller, Eds.), Vol. I, pp. 131-181. Academic Press, New York.