Introduction

Bird calls are functionally rich signals and are used in a wide range of behavioral contexts by passerine and non-passerine birds. In the past recent years, the birds have been rigorously identified with automatic recognition using the audio signals [1], and this approach has become more advanced with time due to the reason that autonomous recording units (ARUs) are affordable, power sufficient and the data can be obtained from distant areas which can be transmitted directly to the laboratory for analysis through GSM or satellite connection [2].

Spectrogram analyses have demonstrated that birds produce two kinds of vocalizations, a) typically short and simple calls, and b) usually lengthy and well-structured organized songs mainly associated with reproduction and aggression [3]. Singing lengthy songs is not the usual feature of non-passerine birds, instead they typically use short and simple calls in a variety of behavioral contexts like interactions between territories [4, 5], coordination of reproductive activities [6], foraging [7], avoidance of predator [8] and contact maintenance [9]. These vocalizations are less complex and characterized by mono syllabic and changing frequency patterns [10]. Ornithological studies are largely dependent on recordings of vocalizations about the subject bird [11]. Compared to passerine birds whose communication behavior is well represented in literature [12, 13], call–based communication system of non-passerine birds has been given less attention in the field of bioacoustics [14, 15].

Birds of the cuckoo family (Cuculidae) are although famous for their loud vocalizations, but unlike Psittaciformes, Passeriformes, and Trochilidae, cuckoos are non-vocal-learners [16, 17]. Learners acquire their vocalizations by imitation [18] and non-learners through inheritance [19], the main functions of vocalizations seem to be similar in both groups, and are involved in mate attraction and deterrence of rival males [20, 21]. Despite the extensive research on cuckoos, the description of the Asian koel’s vocalizations is still lacking. Koels are typically shy and secretive birds but produce loud and distinguishable calls that are easily detectable [22]. Cuckoos produce hundreds of sufficiently loud calls on a daily basis but depending on the habitat, and if the weather is calm, the common cuckoo (Cuculus canorus) calls for example, can be heard at a distance of approximately 2–3 km [17].

The Asian koel is a migratory bird that spends its summer in the plains of Pakistan and migrates back towards India as winter begins [23]. It is primarily a brood parasite of mynas and crows all over the Indian subcontinent [24]. In the Pothwar region of Pakistan, its breeding season begins from late May and continues till early September [25]. Its habitat differs from the other family members because, unlike other cuckoo species which are mostly insectivorous, the Asian koel is totally carpophagous (feeds on fruit) and occasionally adds insects to its diet [23, 26]. Asian koels can also be differentiated from other members of the family Cuculidae in that they often tolerate the young of their hosts, rather than forcing them out of the nest [26, 27]. This species lacks sexual dimorphism in body size, but shows prominent sexual dimorphism in plumage color [28].

Maller and Jones [29] previously described three main calls for the adult male koel in their breeding season in Australia and labeled them as “cooee call”, “wurroo call” and “whik call”. Data regarding vocalizations of the Asian koel are scant, and thus the present study was designed to provide a detailed description of vocalizations of the adult male Asian koel during its breeding season. We first categorized the vocal signals on the basis of audio and visual examination of the spectrogram. Subsequently, various spectral and temporal parameters were measured that were used for the statistical validation of the vocal categories. We also noticed some of usages for all the vocal categories in the field. Finally, besides standardized categories as mentioned above, onomatopoeic names were introduced to describe the vocal categories.

Methods

Acoustic analysis

In the acoustic analysis, we measured 10 call parameters and three categorical call variables for the selected vocalizations which have been used previously in the vocal repertoire studies [31–33]. Measurable variables defined various spectral, temporal, frequency and amplitude parameters (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Abbreviations list and description of various acoustic parameters and categorical call variables measured for each vocal category.

https://doi.org/10.1371/journal.pone.0186604.t001

All the call parameters and categorical variables mentioned in Table 1 were measured for all calls by using Praat software v. 6.0.20 [34]. Initially all of the calls were selected individually and F0 contour was extracted using the cross correlation method for the spectrograms of calls through Fast Fourier Transform method (FFT), Gaussian window shape, window length of 0.012 s, frequency view range of 700–24000 Hz, pitch floor of 75–700 Hz, dynamic range of 70 dB, intensity range of 30–100 dB. The F0 s (Hz), F0 e (Hz), F0 min (Hz) and F0 max (Hz) values were then measured from the extracted F0 contour. All measurements were made using the semiautomatic method using an onscreen cursor selection by the same person. The F0 contour was extracted according to the method previously described for the vocal repertoire of the African penguins (Spheniscus demersus) [31]. To measure F0 contour, each syllable (containing primary and secondary elements) was selected as a whole. To measure the jitter (%) and shimmer (%), [Sound: To Pitch (cc)] command was used. The jitter is the cycle to cycle fluctuation in the fundamental frequency while, the shimmer is the cycle to cycle fluctuation in amplitude [35, 36]. Both are used for speaker verification [37], to study the pathological characteristics of voice [38] and laryngeal pathologies [39, 40]. We also assigned score 1 to a full-length and 0.5 score to a half-length elements in all type of calls. However, in the case of type 1 coegh calls and type 2 cooee calls it was difficult to score the diffused secondary elements for statistical analysis, hence the scores were assigned on the basis of the number of secondary elements of the nearby type 1 cooee calls in the same bout.

Results

Statistical classification

Descriptive statistics of 13 different vocal parameters for a total of 6 vocal categories are presented in Table 2. Eight out of ten parameters, chosen by stepwise regression model were Dur (s), F0 s (Hz), Shimmer (%), F0 max (Hz), F0 e (Hz), F0 min (Hz), Pitch max (Hz) and Pitch m (Hz). The first two functions in DFA showed maximum variance (Function 1: 63.9%, eigenvalue = 59.42; Function 2: 26.6%, eigenvalue = 24.75) and showed highly significant differences between the different types of calls (Wilks’ λ DF1/5 = 0.001, df = 40, P < 0.001 and Wilks’ λ DF2/5 = 0.001, df = 28, P < 0.001). The function one in standardized canonical discriminant function coefficients showed positive coefficients for duration (0.64), F0 e (0.58), and F0 max (0.41) and high negative coefficient for Pitch max (-0.90). Function two showed high positive coefficients for duration (0.76), and Pitch max (0.60; S1 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. The descriptive statistics of acoustic parameters measured for all six vocal categories.

https://doi.org/10.1371/journal.pone.0186604.t002

Using the cross validated method, DFA classified the vocalizations correctly to 100% into the predicted vocal categories that we initially classified on the basis of spectrographic examination. The DFA correctly classified 100% (108/108) of type 1 cooee call, 100% (36/36) of type 2 cooee call, 100% (40/40) of type 1 coegh call, 100% (52/52) of type 2 coegh call, 100% (55/55) of wurroo call and 100% (43/43) of coe call (S2 Table). All six vocal types were distinctively separated in space in the plot defined by discriminants function 1 and 2 scores (Fig 1). Moreover, the remaining three parameters (No. of full-length secondary elements, no. of half-length secondary elements and no. of diffused secondary elements) that were not used in DFA analysis, were also found significantly different among all the six vocal categories. The statistical description is presented in Table 3.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Plot of discriminant scores generated by discriminant function 1 and 2 which shows distinct separation of the six vocal types of the adult male Asian koel.

Black dots show the centroids of the different vocal categories.

https://doi.org/10.1371/journal.pone.0186604.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Summary of statistical results of three nonnormalized parameters.

https://doi.org/10.1371/journal.pone.0186604.t003

Classification and description of calls

A total of six distinctive vocal categories were identified on the basis of initial visual and audible inspection of the spectrograms and subsequent validation via Discriminant Function Analysis (Fig 2). For each of the vocal category, the description of the spectrographic pattern and various characteristics are described below:

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Spectrograms of different vocal categories of the adult male Asian koel.

Two consecutive type 1 cooee calls (A), two consecutive type 2 cooee calls (B), two consecutive type 1 coegh calls (C), two connective type 2 coegh calls (D), a bout of wurroo calls (E), two consecutive coe calls (F). Spectrograms were constructed in Praat using the Fast Fourier Transform method (FFT), Gaussian window shape, window length of 0.012 s, frequency view range of 700–24000 Hz, pitch floor of 75–700 Hz, dynamic range of 70 dB, intensity range of 30–100 dB, 0.5 s zooming level.

https://doi.org/10.1371/journal.pone.0186604.g002

Type 1 cooee call (S1 Audio).

The type 1 cooee call is a harmonically rich call compared with all other types of calls. It contained a primary element of high spectral energy which produced maximum number of full-length secondary elements that added maximum melodies to this call (Fig 2A). This call was found repeated in a continuous manner at dawn and dusk until a female bird joined the male bird’s territory. This call type was also the most common type (e.g. 205/750 or 33% calls emitted by koels were of this call type) of the adult male Asian koel vocalizations and pleasant to human hearing. The shimmer (%) value for this call was found to be lower than all other type of calls (Table 2). The type 1 cooee call also overlapped slightly with type 2 cooee call and type 2 coegh call in DFA analysis (Fig 1), statistical analysis separated them (Table 3) whereby, the former was most common and the latter two were less common, thus further distinguishing these three call types.

Type 2 cooee call (S2 Audio).

This call type had fewer full-length secondary elements compared to the type 1 cooee call, while some of the secondary elements showed a diffused pattern (Fig 2B). The birds emitted these calls continuously at midday or at the end of dawn without including any other type of call in a bout or sometimes at the start of a bout which contained the type 1 cooee calls, type 1 coegh calls and type 2 coegh calls. The jitter and shimmer percent of this call had higher values compared to the type 1 cooee call (Table 2). This call type was frequently produced by the bird when it had produced type 1 cooee calls for a long time or in a situation where other birds like crows and male birds of same species entered its territory.

Type 1 coegh call (S3 Audio).

All the secondary elements which were produced by the primary element in this call type were diffused which clearly visually distinguished them from other type of calls. This call type had no distinguishing half-length or full-length secondary elements (Fig 2C). Jitter and shimmer percent values were higher in comparison to all other calls (Table 2). This type of vocalization was sharp and unpleasant to human hearing. The bird uttered this call in the middle of a bout of type 1cooee calls when it usually produced the latter calls continuously for up to 8.0 ± 2.0 minutes.

Type 2 coegh call (S4 Audio).

Spectrographic examination of type 2 coegh calls confirmed that the first and last parts of the primary element were not involved in the production of clear-cut secondary elements. In contrast the middle half portion of the primary element in the syllable produced clear secondary elements of half-length (Fig 2D). The jitter value was lower as compared to all other call types (Table 2). This vocalization was emitted by the birds in a bout usually before the type 1 coegh call or in between a long bout of the type 1 cooee calls. Similar to type 1 coegh call, the birds never started their bouts from type 2 coegh calls. Spectrographically both type 1 and type 2 coegh calls could be visibly distinguished from each other.

Wurroo call (S5 Audio).

Koels frequently produced wurroo calls in a continuous manner. Except the coe call, these calls were of shorter duration and were produced less frequently as compared to all other types of calls. The starting two calls in the bout of wurroo calls were usually short and produced fewer full-length secondary elements, while the latter calls had long length durations and produced maximum secondary elements of full-length. The full-length secondary elements in syllables of the starting wurroo calls had low spectral energy as compared to the full-length secondary elements of the middle wurroo call in a wurroo calls bout (Fig 2E). This vocalization was produced usually at early dawn and sunset and most often the birds used to take start of their vocalizations with this call type.

Coe call (S6 Audio).

This type of vocalization was of shorter duration and was found to be produced at times when other male or female birds of same species perched on the same tree. The call type was more frequently heard when the birds were involved in quick communications and foraging. When calling, the birds were found usually moving to and fro on the branches of a tree. In comparison to other type of calls, the F0s (Hz) and F0e (Hz) values of coe call were very low (Fig 2F, Table 2).

Discussion

The present study is the first to provide a comprehensive acoustic analysis of adult male Asian koels vocalizations. We recorded and analyzed a total of 334 high quality calls recorded in its wild habitat in two consecutive breeding seasons and determined six distinct call types in the vocalizations.

Previously Maller and Jones [29] identified three different vocal categories for the male adult common koel; “cooee”, “wurroo” and “whik”. In our findings the spectrograms of type 1 cooee call; the most common call type, appeared substantially different from the cooee call, the most common call type in their findings. For example, greater numbers of secondary elements are produced by the primary high spectral element in this type of call. We were so far unable to record any whik call from the adult Asian koel in two consecutive breeding seasons. Instead, we identified four other call types that we categorized as; “type 2 cooee”, “type 1 coegh”, “type 2 coegh” and “coe”, and also measured various structural and temporal parameters for all call types. Differences in the vocalization of koels found in Asia and Australia support the possibility of either species or sub-species differences among these geographically distinct populations. Although, the taxonomy of common koel is complex, both the findings of Maller and Jones [29] and ours provide useful evidences for systematic clarification of the common koel in future studies. Though, it has also been reported that common koels found in Australasia have similar calls [26]. The present study thus further reveals that the Asian koel produces six different vocalizations, while the koels found in Australia produce only three kinds of vocalizations [29]. In addition, like the most common call type “cu-coo” of the common cuckoo, previously studied for vocal distinctiveness in detail [42–44], the Asian koel also frequently produced the “type 1 cooee call” in its vocalizations. Future studies to examine this call type for intra and inter individual variations would be valuable.

The DFA results from 10 call parameters accurately classified 100% of the adult male Asian koel calls according to the predicted vocal categories which were identified on the basis of visual examination of spectrograms. These results were relatively higher than the previously reported vocal classification for birds: e.g. 90.5% in case of African penguins [31], 83.3% in case of great curassows (Crax rubra) [45], and 74.2% in the case of the smooth-billed anis (Crotophaga ani) [15]. Furthermore, in the statistical analysis of the three acoustic parameters (No. of full-length SEs, no. of half-length SEs and no. of diffused SEs) both the Mann-Whitney U-test (two—tailed) and Wilcoxon W test results were taken into consideration because the latter one is also considered a different version of Mann-Whitney U-test in SPSS software analysis [46].

The current study also confirms that, all the types of calls (type 1 cooee, type 2 cooee, type 1 coegh, type 2 coegh, wurroo and coe call) of the adult male Asian koel are harmonically rich calls, their syllables consist of a basic primary element with nearly similar fundamental frequency range but the secondary elements differ in composition and number. It has been described previously, that sounds which have frequencies occuring together add melody to the sound and are basically the multiples of the first element frequency and can be produced by a single source [11, 47]. Besides, some syllables contain structures which are not harmonically related to each other and may originate from separate sources of the sound [48]. Our study provides evidence that in the vocalization of the adult male Asian koel, the primary elements are harmonically connected with the secondary elements in the syllables of all types of calls. This is so because during spectrogram analysis we found a reduction in the number of secondary elements that coincided with a frequent decrease in the spectral energy of the primary element.

Most of the host bird species employ various tactics, such as high level of rejection toward cuckoo eggs or high level of aggressiveness toward the female parasite to escape brood parasitism [49]. The aggression of the host toward the brood parasites has already been reported in common cuckoo [50] and several African cuckoos [51]. Parasitic cuckoos however use the strategy to evolve quickly to overcome resistance by their host species [49]. Furthermore, birds also show aggression by producing usually low pitched [52, 53] and harsh vocalizations [54, 55]. It is clear from our findings that type 1 coegh call and type 2 coegh call lack harmonics, have low pitch values and, contain more noise compared to the type 1 cooee call. These findings indicate that the adult male Asian koel may use these call types to display some kind of aggression to compete efficiently with its host species to well utilize its short breeding season in order to achieving an ultimate high breeding success. It also came to our observation that during the courtship display, male koels got agitated, became more vocal on seeing a female koel and began spectacular chases toward the females. In addition, the type 1 cooee call was frequently produced by the birds during the whole day and occasionally also at night time. Again, this type of activity indicates that this call type is most likely the advertising call. However, assigning functions to calls was beyond the scope of the present study and therefore need extensive additional work.

In conclusion our study provides first comprehensive description of various acoustic parameters and the statistical validation of the adult male Asian koels vocalization recorded in their breeding seasons. Although current findings are limited only to the breeding season and did not fully explore the entire vocalizations of the male Asian koel, still we feel confident that we recorded all type of vocalizations of the adult male Asian koel in their breeding season. This is so because we recorded the birds i) from initial arrival in Pakistan till the start of back journey toward India, ii) three times in a day, that is at early dawn, noon and at dusk. However, we consistently found that noon time recordings had greater noise. To our understanding this was due to the reason that several other bird species were active at these times. We did initial analyses on these recordings to identify any other potential vocal category, but we found that most of the vocal categories were similar to those identified at dawn and dusk time recordings. Thus these recordings were omitted from further analyses.

Our findings therefore enrich the biological knowledge about the adult male the Asian koel vocalizations and provide a foundation for future comparative studies among different species of other cuckoos. Future studies should focus on exploring the entire vocal repertoire of male Asian koels and to understand the functions of the vocalizations in sexual selection, social interaction and individual recognition.

Supporting information

Acknowledgments

The authors are indebted to Quaid-i-Azam University, Islamabad for support and funding the present research.