Introduction
Vocalizations that are individually distinctive, i.e. potentially characteristic of individual animals, have been described in many avian species (reviewed by Dhondt and Lambrechts Reference Dhondt and Lambrechts1992, Stoddard Reference Stoddard, Kroodsma and Miller1996). Recognition based on individually distinctive vocalizations is a prominent and functionally important aspect of signalling among animals in several contexts (Bradbury and Vehrencamp Reference Bradbury and Vehrencamp1998). Previous research has largely focused on differences in the acoustic structure of vocalizations and vocal recognition in parent-offspring interactions (e.g. Baker Reference Baker1982, Aubin and Jouventin Reference Aubin and Jouventin1998, Mathevon et al. Reference Mathevon, Charrier and Jouventin2003), territory defence (e.g. Galeotti and Pavan Reference Galeotti and Pavan1991, Farquhar Reference Farquhar1993, Aubin et al. Reference Aubin, Mathevon, Da Silva, Vielliard and Sebe2004, Yorzinski et al. Reference Yorzinski, Vehrencamp, Clark and McGowan2006) and sexual interactions (e.g. Jouventin Reference Jouventin1982, Charrier et al. Reference Charrier, Jouventin, Mathevon and Aubin2001, Sung and Miller Reference Sung and Miller2007). Individual variation in acoustic signals may be adaptive for several reasons. It may permit individual recognition that could help in coordinating group movement and distinguishing members of neighbouring groups from strangers (Falls Reference Falls, Kroodsma and Miller1982, Ydenberg et al. Reference Ydenberg, Giraldeau and Falls1988, Chapman and Lefebvre Reference Chapman and Lefebvre1990, Wich et al. Reference Wich, Assink, Becher and Sterck2002, McComb et al. Reference McComb, Reby, Baker, Moss and Sayialel2003). Individual variation of loud calls could also indicate male quality and influence female choice (Thomas Langurs Presbytis thomasi; Steenbeek and Assink Reference Steenbeek and Assink1998) and male-male competition (Red Junglefowl Gallus gallus; Furlow et al. Reference Furlow, Kimball and Marshall1998; Thomas's Langur; Steenbeek et al. Reference Steenbeek, Assink and Wich1999; Common Loon Gavia immer; Mager et al. Reference Mager, Walcott and Piper2007).
In the present study we focused on vocalizations of two species of hornbills in the West Visayas in the Philippines: the Visayan Hornbill Penelopides panini panini and the Rufous-headed Hornbill Aceros waldeni. Both species are endemic to this area and are known to occur on the islands of Panay and Negros. The remaining populations of these species are extremely small and fragmented. The Visayan Hornbill is listed as ‘Endangered’ (BirdLife International 2007a) while the Rufous-headed Hornbill is listed as ‘Critically Endangered’ (BirdLife International 2007b). One subspecies of the Visayan Hornbill (P. p. ticaensis) from Ticao Island, an island off Masbate is feared to be already extinct (Collar et al. Reference Collar, Mallari and Tabaranza1999). Members of the genus Penelopides are probably group-territorial; based on aggression shown towards other hornbills in captivity (Kemp Reference Kemp1995) they defend a territory or merely an area around a nest hole. The Rufous-headed Hornbill is also a territorial bird at least during the breeding season (Kauth et al. Reference Kauth, Engel, Lastimoza and Curio1998).
Although hornbills are amongst the noisiest birds (Kemp Reference Kemp1995), very few studies have been devoted to hornbill bioacoustics. Rainey and Zuberbühler (Reference Rainey and Zuberbühler2007) used acoustic recordings collected during studies of primates to detect seasonal variation in the abundance of forest hornbills over 10 years. Individual distinctiveness of vocalizations has been recorded in the Helmeted Hornbill Rhinoplax vigil (Haimoff Reference Haimoff1987). Until now, no bioacoustic study has been conducted on any of the Philippine hornbill species and only verbal descriptions of vocalizations have been documented (Ripley and Rabor Reference Ripley and Rabor1956, Rabor Reference Rabor1977, Kemp Reference Kemp1995, Brooks et al. Reference Brooks, Evans, Dutson, Anderson, Asane, Timmins and Toledo1992, Kauth et al. Reference Kauth, Engel, Lastimoza and Curio1998, Kennedy et al. Reference Kennedy, Gonzales, Dickenson, Miranda and Fisher2000, Kemp Reference Kemp, del Hoyo, Elliott and Sargatal2001). Visayan hornbills' calls are described as a series of rapidly following notes (Ripley and Rabor Reference Ripley and Rabor1956) or noisy, keeping up incessant notes (Rabor Reference Rabor1977) and nasal high-pitched notes resembling the sound of a toy trumpet (Brooks et al. Reference Brooks, Evans, Dutson, Anderson, Asane, Timmins and Toledo1992, Kennedy et al. Reference Kennedy, Gonzales, Dickenson, Miranda and Fisher2000). Males also utter soft squeaking calls sounding like ta-rik-tik, whilst feeding the female, providing the onomatopoeic common name of Visayan Hornbills – Tarictic - in the local Tagalog language (Kemp Reference Kemp1995, Kennedy et al. Reference Kennedy, Gonzales, Dickenson, Miranda and Fisher2000). Kauth et al. (Reference Kauth, Engel, Lastimoza and Curio1998) identified six different types of vocalization in the Rufous-headed Hornbill. Out of these, four types of male vocalization were described – territorial call, loud and far carrying, sounding very much like the bleating of a lamb; threat call, uttered repeatedly when the male is scared or engaged in an agonistic encounter; soft croaking and babbling contact call, and a monosyllabic “krook”.
Loud calls of hornbills are useful in communication in dense habitats to maintain contact, to attract the attention of flying birds, to proclaim possession of a defended area (Kemp Reference Kemp, del Hoyo, Elliott and Sargatal2001) and to signal to a predator that it has been detected (Rainey et al. Reference Rainey, Zuberbühler and Slater2004a,b). The aim of this study was to analyze the loud calls of the Rufous-headed Hornbill and the Visayan Hornbill to assess their potential for individual identification and to determine which combination of acoustic variables could be employed to distinguish between individual birds.
Methods
Study sites and subjects
The male vocalizations of Rufous-headed Hornbill and Visayan Hornbill were recorded at two sites in the Western Visayas, Philippines. Both sites were breeding centres for endangered animals of the Philippines, particularly those that are endemic to the West-Central Visayas faunal region. The first site was the Biodiversity Conservation Centre of the Negros Forests and Ecological Foundation, Inc. (NFEFI-BCC) in Bacolod City, which is situated on the Island. This centre holds Visayan Hornbills of Negros origin but no Rufous-headed Hornbills. The second site, the Mari-it Conservation Park is situated in the foothills of Mt. Baloy, the third highest peak on the island of Panay. In the Mari-it Conservation Park, Rufous-headed Hornbills have bred successfully in captivity for the first time. At this centre we recorded vocalizations of both species of Visayan hornbills - each originating from Panay.
All subjects were kept in captivity and were housed either in pairs or individually. Vocalizations were recorded during the months of February and March 2007. This period coincided with the time of year when courtship and breeding behaviour was observed in these captive birds (Klop et al. Reference Klop, Curio and Lastimoza2000). Some birds at the Mari-it centre were nesting at this time.
Data collection
Vocalizations were recorded with a Marantz 671 digital recorder and Sennheiser ME 67 directional microphone (frequency response 50–20,000 Hz; 2.5 dB), with K6 powering module (sampling rate 44.1 kHz, sample size 16 bit). The distance between the subject and the microphone ranged from 2 to 8 m. We obtained calls from nine adult male Visayan Hornbills, five at NFEFI, Bacolod and four at Mari-it, and five adult male Rufous-headed Hornbills at Mari-it.
Data analyses
The recordings were analyzed using Avisoft SASLab Pro 4.38 (Specht Reference Specht2006) software. For the detailed analysis, calls that had the lowest background noise among all the recordings available for the particular individual were selected. Only the recordings of non-overlapping calls which had a good signal to noise ratio and only one-element calls (in the case of Visayan Hornbill) were considered in the analysis.
Twelve parameters in Visayan Hornbill and 13 parameters in Rufous-headed Hornbill were measured with a combination of manual and automatic procedures. Single calls were separated manually with the help of the envelope curve and the spectrogram of the following parameters: hamming window, FFT-length 1024, frame size 100% and overlap 88%. This setting provided the frequency resolution of 22 Hz, the time resolution 5.8 ms and the bandwidth 28 Hz. Consequently, temporal parameters such as duration and time distance from the start to maximum amplitude (location of the maximum amplitude) were computed automatically. One-dimensional function Amplitude spectrum (linear) was used for spectral measurements in order to describe the energy spectrum of the call. Maximum frequency, minimum frequency, bandwidth, frequency of maximal amplitude (max frequency peak measured at the mean spectrum of the entire spectrogram), fundamental frequency at the highest point of its frequency modulation (and its first harmonic frequency-measured only in Rufous-headed Hornbill), 25%, 50% and 75% quartile (below this frequency is 25%, 50% and 75% of the total energy) were measured using the function Spectral Characteristics. Linear Prediction Coding procedure (LPC) was applied for the identification of the two major energy peaks of the smooth spectral envelope: the frequency of the first amplitude peak (LPC 1) and the frequency of the second amplitude peak (LPC 2). The LPC algorithm was based on the least-square estimation technique that uses autocorrelation (Specht Reference Specht2006).
Stepwise discriminant analyses (DFA) were used to reduce the number of variables that were highly correlated in order to examine differences between individuals. Firstly, the data were log transformed to improve the normality of the distribution. The following variables were used as source data for these multivariate procedures: duration, time to maximum amplitude in relation to the total call duration (location of the maximum amplitude), fundamental frequency, sum of the fundamental and first harmonic frequency (in Rufous-headed Hornbill only), frequency of maximum amplitude (peak frequency), 25%, 50% and 75% quartile, maximum and minimum frequency, bandwidth, inter-quartile range (= 75% quartile – 25% quartile), first amplitude peak (LPC 1) and second amplitude peak (LPC 2). Only the bandwidth variable did not pass the tolerance criterion of 0.01 for DFA. From the analysis, we excluded highly mutually correlated variables (when r > 0.8) (see Mitchell et al. Reference Mitchell, Makagon, Jaeger and Barrett2006). The number of variables was less than 0.33 times the number of ob-servations, which met the criteria set by Kazial et al. (Reference Kazial, Burnett and Masters2001). A priori probabilities of classification were set proportionally to the group sizes.
Calls of both species were randomly split half-and-half, in order to validate results of discriminant analysis. This procedure provided a training set and a test set for each species. The following classification of one half of the dataset was made, with the discriminant function derived from the other half (see Klecka Reference Klecka1980).
All analyses were done using STATISTICA Analysis System (Release 6.0) and considered significant when P < 0.05. We used the Bonferroni correction factor at an alpha of 0.05 to account for the number of pairwise comparisons made to reduce the chance of type I errors. The values reported in the results represent means ± SD.
Results
From the recordings obtained, we analyzed 127 calls from nine male Visayan Hornbills, 10–15 calls from each individual, and 198 calls from five male Rufous-headed Hornbills, 19–88 from each individual.
The DFA resulted in eight variables in the case of Visayan Hornbill and six variables in the case of Rufous-headed Hornbill (Table 1). These were used in the DFA to test the accuracy of individual identification. In the Visayan Hornbill, two temporal variables were investigated (duration and relative location of the maximum amplitude) and six spectral variables (second amplitude peak, inter-quartile range, fundamental frequency, 50% quartile, maximum and minimum frequency). In the Rufous-headed Hornbill, one temporal variable (duration) and five spectral variables (75% quartile, sum of the fundamental and first harmonic frequencies, inter-quartile range, second amplitude peak and minimum frequency) were investigated.
Ranking of the variables that contributed most to DFA on individual identity of Visayan Hornbill (VisH) and Rufous-headed Hornbill (RufH), ranked from highest contribution to the lowest contribution and the correlations of individual factors with significant canonical roots.
Code: 2LPC (second amplitude peak), duration (call duration; element duration in Rufous-headed Hornbill), location peak (location of the maximum amplitude), 75–25 quart (inter quartile range), F0 (fundamental frequency), quartile 50% (below this frequency is 50% of the total energy), quartile 75% (below this frequency is 75% of the total energy), F max (maximum frequency), F min (minimum frequency), F0 + H1 (sum of the fundamental and first harmonic frequency).
The loud calls of Visayan Hornbills were noisy with harmonic structure, and both frequency and amplitude modulation (Figure 1). Calls were usually formed by a single element. Some calls contained two to three elements (Figure 2). Such calls were recorded only in a small number of individuals, so they were not included in our analysis. Loud calls of Visayan Hornbills were short, with duration of 26–140 ms (70 ± 23 ms; mean ± SD). The peak frequency (frequency with the maximum amplitude) ranged between 540 Hz and 8,090 Hz (4573.7 ± 1344.9 Hz). The bandwidth lay between 7,190 and 17,160 Hz (4,490.2 ± 1,449.1 Hz) with the minimum frequency 290– 600 Hz (470 ± 55.6 Hz) and the maximum frequency 7,660–17,650 Hz (13,525.3 ± 2,155.0 Hz). The loud calls of Rufous-headed Hornbills (Figure 3) were noisy broadband sounds, with harmonic structure and both frequency and amplitude modulation. Calls can be formed by 2–5 prominent amplitude peaks (mean 3.2). In some cases, these peaks were separated into single elements (Figure 4). Calls were uttered singly (n = 11) or in sequences of 2–19 (mean = 6.4) with an interval of 0.7–2.9 s (1.2 ± 0.3 s) (Figure 4). The duration of the calls ranged from 200 to 540 ms (343 ± 64 ms). The peak frequency was between 613 Hz and 4,831 Hz (1,539.4 ± 1,341.9 Hz). The bandwidth lay between 4,336 and 13,299 Hz (8,561.3 ± 2,004.3 Hz) with the minimum frequency 188–357 Hz (288.3 ± 37.1 Hz) and the maximum frequency 4,600–13,638 Hz (8,850.2 ± 1,999.2 Hz).
Spectrogram of Visayan Hornbill calls from four individual males, three calls from each of the four individuals (the intervals between calls were modified).
Visayan Hornbill calls which contained two or three elements.
Spectrograms of the Rufous-headed Hornbill calls from four individual males.
Spectrogram of a typical sequence of Rufous-headed Hornbill calls.
Most of the frequency parameters of Rufous-headed Hornbill (fundamental frequency, frequency of maximum amplitude, 25%, 50% and 75% quartile, minimum and maximum frequency, bandwidth) were significantly lower than those in Visayan Hornbill (Mann-Whitney U test: P < 0.001). In the case of temporal parameters, Rufous-headed Hornbill calls were significantly longer (Mann-Whitney U: test: P < 0.001) than those of Visayan Hornbill and the location of the maximum amplitude did not differ between these species. In the case of Visayan Hornbill, the DFA correctly classified more than 90% of all calls (Wilks’ lambda = 0.0012) and validation procedure assigned 80% correctly. Five significant canonical functions described more than 97% of the variation. The first four canonical functions had an eigenvalue > 1 and described more than 93% of the variation. The first two functions describing 78% of variation were plotted against each other in Figure 5.
Location of Visayan Hornbill calls from nine males on the first two canonical functions.
In Rufous-headed Hornbill, DFA correctly classified more than 89% of all calls (Wilks’ lambda = 0.0285) and validation yielded an average correct assignment of 85%. The analysis generated three significant canonical functions with eigenvalue > 1, explaining more than 99% of the variation. The first two functions describing 79% of the variation were plotted in Figure 6.
Location of all Rufous-headed Hornbill calls from five males on the first two canonical functions.
For distinguishing among individuals of Visayan Hornbill, the most useful acoustic parameters were second amplitude peak (r = −0.79), time to maximum amplitude in relation to the total call duration (r = −0.58), and call duration (r = 0.46) (Table 1). The Kruskal-Wallis ANOVA analysis of the first and second canonical root scores yielded highly significant differences between the calls of individual Visayan Hornbill (Root 1: H = 99.4, P < 0.001; Root 2: H = 86, P < 0.001).
For distinguishing between individuals of Rufous-headed Hornbill, the sum of the fundamental and first harmonic frequency (r = 0.65) and the upper quartile (r = 0.71) were useful (Table 1). The Kruskal-Wallis ANOVA analysis of the first and second canonical root scores provided highly significant differences between the loud calls of individual Rufous-headed Hornbill (Root 1: H = 149, P < 0.001; Root 2: H = 111, P < 0.001).
Discussion
The present research is the first bioacoustic study of any Philippine hornbill species and the first multivariate analysis of hornbill vocalizations.
In broadband acoustic signals with noisy and atonal structure, where energy is spread over a wide frequency range, it is difficult to decide which parameters should be measured to characterize the properties of a signal (Schrader and Hammerschmidt Reference Schrader and Hammerschmidt1997). Recognition between the animals must be based on a multiparametric analysis, taking into account both spectral and temporal features of the calls (Mathevon Reference Mathevon1997). An identification system based on several parameters may better secure vocal signatures and reduce the risk of confusion (Aubin et al. Reference Aubin, Mathevon, Staszewski and Boulinier2007). As such, the multiparametric approach is a very useful technique for analyzing the complex vocalizations of birds (Sparling and Williams Reference Sparling and Williams1978, Martindale Reference Martindale1980, Allenbacher et al. Reference Allenbacher, Böhner and Hammerschmidt1995, Appleby and Redpath Reference Appleby and Redpath1996, Böhner and Hammerschmidt Reference Böhner and Hammerschmidt1996, Lengagne Reference Lengagne2001).
Our analysis revealed that individuals of the two hornbill species studied can be identified on the basis of their loud calls. This means that hornbill calls contain information about the caller's identity and these findings will also have value for studies of behavioural ecology. Results of correct classification revealed 89% success in Rufous-headed Hornbill and 90% success in Visayan Hornbills. These outcomes are comparable with results of other bird species, such as Pygmy Owl Glaucidium passerinum (84%; Galeotti et al. Reference Galeotti, Paladin and Pavan1993), Corncrake Crex crex (100%; Peake et al. Reference Peake, McGregor, Smith, Tyler, Gilbert and Green1998), Christmas Island Hawk-Owl Ninox natalis (91%; Hill and Lill Reference Hill and Lill1998), European Nightjar Caprimulgus europaeus (99%; Rebbeck et al. Reference Rebbeck, Corrick, Eaglestone and Stainton2001), Western Screech-Owl Megascops kennicottii (92%; Tripp and Otter Reference Tripp and Otter2006), European Eagle Owl Bubo bubo (98%; Grava et al. Reference Grava, Mathevon, Place and Balluet2007), and Woodcock Scolopax rusticola (95%; Hoodless et al. Reference Hoodless, Inglis, Doucet and Aebischer2008).
The smaller hornbill species, the Visayan Hornbill, has high-pitched calls that are used as contact calls both between mates and other adults. Signals were uttered mostly as single bouts and thus temporal parameters related to spacing of single calls have not influenced vocal individuality. However, some temporal features within single calls such as the time to maximum amplitude and call duration, which were correlated with the second discriminant function, were useful for distinguishing individual Visayan Hornbills. The second discriminant function was correlated with the second amplitude peak. The parameter, time to maximum amplitude, that can be considered a component of amplitude modulation, was important for individual distinctiveness in Visayan Hornbill. This result contrasts with findings in colonial birds where the amplitude modulation is not often used for individual recognition by the birds (Aubin and Jouventin Reference Aubin and Jouventin2002, Jouventin et al. Reference Jouventin, Aubin and Lengagne1999, Charrier et al. Reference Charrier, Jouventin, Mathevon and Aubin2001, Jouventin and Aubin Reference Jouventin and Aubin2002, Mathevon et al. Reference Mathevon, Charrier and Jouventin2003). In the South Polar Skua Catharacta maccomicki, also, both amplitude and frequency modulation were not good individual markers and individual recognition is based on the spectral profile of their calls (Charrier et al. Reference Charrier, Jouventin, Mathevon and Aubin2001). The importance of second amplitude peak may reflect the influence of supralaryngeal vocal tract filtration, formant structures, on individual distinction, as has been found in the Whooping Crane Grus americana (Fitch and Kelley Reference Fitch and Kelley2000) and the Oilbird Steatornis caripensis (Suthers Reference Suthers1994). Based on our results it seems that loud calls of male Visayan Hornbill are not designed for long-distance communication.
On the contrary, in the case of Rufous-headed Hornbill the first discriminant function was correlated with fundamental and first harmonic frequencies, which were significantly lower then those in Visayan Hornbill. Such frequencies are attenuated less rapidly in all types of habitats (Morton Reference Morton1975, Marten and Marler Reference Marten and Marler1977, Piercy et al. Reference Piercy, Embelton and Sutherland1977). The second discriminant function was correlated with the upper quartile. The importance of this parameter in discrimination of individuals may indicate the role of whole bandwidth for individual recognition as has been found in parental calls of the Gentoo Penguin Pygoscelis papua (Charrier et al. Reference Charrier, Jouventin, Mathevon and Aubin2001). Individual identity in loud calls of Rufous-headed Hornbills seems to be encoded in the frequency structure of each element of a given call series rather than in temporal parameters related to call series such as the repetition of the elements or call duration. In hornbills, individual distinctiveness of vocalizations has been recorded in the Helmeted Hornbill Rhinoplax vigil (Haimoff Reference Haimoff1987) where only the temporal parameter (interval between notes) was observed to be significantly distinct amongst individuals. Individually related differences in spectral parameters are not as commonly recorded as in temporal parameters and they have been found in Whooping Crane (Fitch and Kelley Reference Fitch and Kelley2000), South Polar Skua (Charrier et al. Reference Charrier, Jouventin, Mathevon and Aubin2001), Arctic Fox Alopex lagopus (Frommolt et al. Reference Frommolt, Goltsman and MacDonald2003), Grey Wolf Canis lupus (Tooze et al. Reference Tooze, Harrington and Fentress1990) and Rhesus Macaque Macaca mulatta (Rendall et al. Reference Rendall, Owren and Rodman1998). Frequency parameters in Rufous-headed Hornbill calls were significantly lower than those in Visayan Hornbills.
The differences between these two hornbill species may be caused by different body (van Zyl and Kemp Reference van Zyl and Kemp1998) and casque size (Alexander et al. Reference Alexander, Houston and Campbell1994) or it could also be the result of selection pressure on loud territory calls, when the higher frequency components are lost at long-distance (Wiley and Richards Reference Wiley, Richards, Kroodsma and Miller1982, review in Naguib and Wiley Reference Naguib and Wiley2001). The Visayan Hornbill, which is the smallest hornbill in the Philippines, has a bill with a low narrow casque (Kennedy et al. Reference Kennedy, Gonzales, Dickenson, Miranda and Fisher2000) and utters high-pitched calls, while the larger Rufous-headed Hornbill, with a more pronounced casque, has territorial calls with a significantly lower fundamental frequency. Alexander et al. (Reference Alexander, Houston and Campbell1994) found a correlation between the casque resonance frequencies and the fundamental frequency. The loud calls of the bigger hornbills indicate existence of amplification and many of them also have a large casque (Kemp Reference Kemp1995). According to our results, in the Rufous-headed Hornbill, which has a more pronounced casque, the fundamental frequency allows discrimination between individuals. In the Visayan Hornbill, with a less pronounced casque, the fundamental frequency did not contribute to individual discrimination and was replaced by other spectral parameters (second amplitude peak) and two temporal parameters (relative location of the maximum amplitude and call duration).
Despite the relatively small sample size, our results show that individually distinctive signatures in hornbill acoustic signals differ significantly between individuals and these findings need to be verified in non-captive situations. It must be emphasized that some other studies (Farquhar Reference Farquhar1993, Lengagne Reference Lengagne2001, Delport et al. Reference Delport, Kemp and Ferguson2002, Grava et al. Reference Grava, Mathevon, Place and Balluet2007) have proved bird acoustic individuality based on comparable sample sizes.
Our study demonstrated that the loud calls of hornbills contain some information about the caller. Individual differences in the acoustic structure of the calls are a prerequisite for individual recognition amongst animals (Falls Reference Falls, Kroodsma and Miller1982). However, it is necessary to perform playback experiments to verify that animals are using acoustic differences to recognize individuals (Rendall et al. Reference Rendall, Rodman and Emond1996). There is strong evidence that hornbills have high cognitive abilities (e.g. Rasa Reference Rasa1983, Kemp Reference Kemp1995). Rainey et al. (Reference Rainey, Zuberbühler and Slater2004b) suggested similarities between the social systems of some hornbills and primates that could lead to the development of sophisticated cognitive abilities. These abilities are directly supported by the existence of hornbill social play and also their large brain (Diamond and Bond Reference Diamond and Bond2003). Rainey et al. (Reference Rainey, Zuberbühler and Slater2004a) demonstrated that Yellow-casqued Hornbills Ceratogymna elata are able to distinguish between vocalizations of Leopards Panthera pardus and Crowned Eagles Stephanoaetus coronatus as well as between the predator-specific alarm calls of sympatric Diana Monkeys Cercopithecus diana. Similarly, other hornbill species can distinguish between the calls of the two predators (Hauser and Wrangham Reference Hauser and Wrangham1990, Rainey et al. Reference Rainey, Zuberbühler and Slater2004b). Undoubtedly, hornbills have a complex system of communication which needs further investigation.
The role of using vocal individuality as a tool for studying individual animals has been suggested in several studies (Saunders and Wooller Reference Saunders and Wooller1988, McGregor and Peake Reference McGregor and Peake1992, Darden et al. Reference Darden, Dabelsteen and Pedersen2003). These techniques may provide less biased data than other marking techniques (Terry et al. Reference Terry, Peake and McGregor2005). However, the role of DFA in such research purposes is limited (Terry et al. Reference Terry, McGregor and Peake2001) because of the necessity of knowing the number of individuals, but this limitation may be overcame by using of a non-parametric form of DFA (Terry et al. Reference Terry, Peake and McGregor2005).
The situation of hornbill conservation in the Philippines is especially urgent (Poonswad and Kemp Reference Poonswad and Kemp1993, BirdLife International 2007a, b, Oliver and Wilkinson Reference Oliver, Wilkinson, Kemp and Kemp2007). Captive Philippine hornbills in breeding centres form reserve populations which may in the future enable reintroductions into the wild. Acoustic signals have the potential to be used for vocal tagging of the reintroduced individuals. Use of acoustic monitoring of individuals, as an alternative non-invasive marking technique, could help in monitoring hornbill individual life history and also in collecting many biological data lacking on Philippine hornbills (see Poonswad and Kemp Reference Poonswad and Kemp1993). Individually distinct vocalizations are especially suitable for monitoring individuals in species sensitive to disturbance caused by capturing and handling (including poaching) and for ethical reasons when capturing can affect survival (De Villiers et al. Reference De Villiers, Meltzer, van Heerden, Mills, Richardson and van Jaarsveld1995, Castelli and Trost Reference Castelli and Trost1996) and for birds with cryptic behaviour living in the highest forest canopy. Acoustic identification can improve census data and was successfully applied for population censuses of several bird species such as Tawny Owl Strix aluco (Galeotti and Pavan Reference Galeotti and Pavan1991), Great Bittern Botaurus stellaris (Gilbert et al. Reference Gilbert, Tyler and Smith2002), Corncrake (Terry & McGregor Reference Terry and McGregor2002), Eagle Owl (Grava et al. Reference Grava, Mathevon, Place and Balluet2007), Dupont's Lark Chersophilus duponti (Lailo et al. Reference Lailo, Vögeli, Serrano and Tella2007) and Woodcock (Hoodless et al. Reference Hoodless, Inglis, Doucet and Aebischer2008). In some cases the information gained from acoustical methods increased the census estimates (Peake and McGregor Reference Peake and McGregor2001). Acoustic methods are less time-consuming and have great logistic and welfare advantages over physical capture-recapture methods. But we must keep in mind that acoustic methods are useless for monitoring yearlings in the stage of call development, so each method offers unique information (Lailo et al. Reference Lailo, Vögeli, Serrano and Tella2007). The most effective method for estimating population size or for providing a minimum population size would be capture-mark-recapture technique. This non-invasive method has been used elsewhere by, for example, Eggert et al. (Reference Eggert, Eggert and Woodruff2003) on elephant DNA obtained from dung and by Puechmaille and Petit (Reference Puechmaille and Petit2007) on bat DNA and would be equally appropriate for individually distinct hornbill vocalisations when combined with the DFA. If calls were recorded across the study area at regular intervals, it could be possible relatively quickly estimate population size using this technique. Although there is some probability that unknown individual with similar acoustic parameters will be erroneously identified as another one, it does not preclude reliable use of this method. Probability of erroneous identification increases rapidly with the total number of individuals included. This is not a serious problem in extremely rare species and/or populations where the population size does not exceed dozens. Such errors may decrease, but not increase, the estimated population size. Therefore, they make the estimates more conservative.
The loudness and repetitive character of Rufous-headed Hornbill calls are suitable for vocal tagging of individuals in very small fragmented populations and could be useful for tagging individuals born in captivity and introduced into reserves.
It can be expected that the higher frequency calls of very short duration of Visayan Hornbill may be less effective in long-distance transmission and thus less suitable for vocal tagging in the wild. However, this method could be suitable for monitoring birds at their nesting place. A ‘voice archive’ (see Seymour and Titze Reference Seymour and Titze1989) of calls from known individuals would help to re-identify such individuals that were not identified visually (Hartwig Reference Hartwig2005).
Individual acoustic variability of bird songs affects reproductive success through male-male competition and mate choice (Catchpole and Slater Reference Catchpole and Slater1995). Some frequency parameters in non-passerines may be important for communicating male quality and condition in male-male competition (e.g. Furlow et al. Reference Furlow, Kimball and Marshall1998, Mager et al. Reference Mager, Walcott and Piper2007) and attraction of females (Beani and Dessì-Fulgheri Reference Beani and Dessì-Fulgheri1995, Appleby and Redpath Reference Appleby and Redpath1997, Miyazaki and Waas Reference Miyazaki and Waas2003). So the complexity of acoustic characteristics found in loud calls of hornbills may encode information related to mate quality, fitness or health, as well as the individual identity of the caller. Casque resonance frequencies are correlated with the fundamental frequency (Alexander et al. Reference Alexander, Houston and Campbell1994). The lower frequency calls of well developed healthy hornbills with a larger casque may indicate not only greater size and age, but also a better quality of immune system. Also, parameters related to the harshness quality of broadband sounds of hornbill loud calls could signal health status, as found in the ratio of harshness in distress calls of Lesser Short-toed Lark Calandrella rufescens (Lailo et al. Reference Lailo, Tella, Carrete, Serrano and López2004), in peak amplitude frequency of harsh syllables in Barn Swallow Hirundo rustica (Garamszegi et al. Reference Garamszegi, Heylen, Møller, Eens and de Lope2005) or in the total call duration in Tawny Owls (Appleby and Redpath Reference Appleby and Redpath1997). Some head ornaments can serve as indicators of condition in both studied hornbill species (Curio Reference Curio2004). The relationship of age and parental quality has not been shown for any hornbill species and such ideas await testing for methodological reasons (Curio Reference Curio2004).
Further studies and playback experiments are needed to understand the dimensions of the information potential of hornbill vocalisations. Individual acoustic variation could reveal a hidden complex signal system in hornbills and a comparative study of vocalisations of the various taxa could provide another area of information to assist in a reassessment of the controversial taxonomy of the species and subspecies of the genus Penelopides.
Acknowledgements
The study would not be possible without help of many people. We owe a special thanks to William Oliver for his help and support of our research in the Philippines, also to Marisol Pedregosa and Pavel Hospodařský for their help in the initial planning of our project. Thank also to the staff of Mari-it Conservation Park and BCC-FFI. We are grateful to DENR officers for permission to conduct this research. Richard Policht and Milada Petrů were supported by Grant Agency of the Czech Republic (No.206/05/H012). We thank also Xiao-Yin Zhang, Tristan Farrow and Gareth Bloomfield for language corrections in the early version of this manuscript. The research equipment was funded by the Grant Agency of the Charles University in Prague, project No. B-BIO-185/2004.
