Animal Biology, Vol. 54, No. 2, pp. 207-220 (2004) Koninklijke Brill NV, Leiden, 2004.
Also available online - www.brill.nl
Amplitude spectra of Corncrake calls: what do they signalise? TOMASZ S. OSIEJUK 1,∗ , BOGUMIŁA OLECH 2 1 Department
of Animal Morphology, Institute of Environmental Biology, Adam Mickiewicz University, 28 Czerwca 1956/198, 61-485 Pozna´n, Poland 2 Kampinoski National Park, Tetmajera 38, 05-080 Izabelin, Poland
Abstract—The territorial call of corncrake Crex crex males is a disyllabic, loud, pulse repetition signal repeated in long series at night. It seems to be relatively simple as it lacks the repertoire variation typical of passerines, but still seems to be the equivalent of a territorial and/or advertisement song. In this paper we tested: (1) whether the pattern of energy distribution within the call (i.e. amplitude spectrum) is individually invariant; (2) whether it could be a signal used in an individual recognition system; and (3) whether it is a sexually selected honest trait related to male body size. We found that amplitude spectra varied more between than within males, but it is rather unlikely that energy distribution is a single feature encoding identity of a male. We found some weak correlations between body size parameters and variables describing amplitude spectra, either supporting or contradicting the expected relationship (i.e. larger male — more energy in lower frequencies). We also found that the frequency of maximal amplitude, even within a single call, may rapidly change by a few kHz, which is most probably an effect of head and body movements, not changes in sound production itself. The conclusion is that all features of corncrake calls, including energy distribution, indicate that this signal has evolved under a strong pressure towards precise localisation of a sender, which is understandable as corncrakes inhabit dense vegetation and call almost exclusively at night. Keywords: amplitude spectrum; body size; corncrake; Crex crex; energy distribution; individual recognition.
INTRODUCTION
Acoustic signals are highly complex because changes in frequency and amplitude in time give a virtually unlimited number of combinations, which may efficiently code for any kind of information (Catchpole and Slater, 1995). Moreover, acoustic signals are often hierarchically organised: at different levels, described as repertoire ∗ Corresponding
author; e-mail:
[email protected]
208
T.S. Osiejuk, B. Olech
size, temporal organisation, or acoustic structure, they may encode different kinds of information for different receivers (e.g., Searcy and Nowicki, 1999). Repertoire size has proved to be an honest signal in sexual selection (Hasselquist et al., 1996). Temporal organisation (including the pattern of repertoire presentation in time) plays an important role in effective territorial defence, which enables tuning and directing the response towards a specific neighbour or floater (e.g., Naguib, 1999; Beecher et al., 2000). The fine acoustic structure of the signal may be used in individual recognition (Stoddard and Beecher, 1983; Rebbeck et al., 2001). This huge variety has not been studied equally among bird taxa, as the majority of researches conducted so far concerned songbirds (suborder Oscines), i.e. a group whose songs are modified through a learning process (Hansen, 2001). Songs of other, mostly non-learning, taxa are usually supposed to be simpler, with a lower between-individual variation. There is little doubt that repertoire variation among non-passerines is lower but, on the other hand, several studies have shown that between-individual variation in acoustic structure of songs or calls is sufficient for individual recognition or may correlate with some biologically important properties of a sender (Galeotti and Pavan, 1991, 1993; Galeotti et al., 1996; Appleby and Redpath, 1997; Galeotti, 1998). Corncrake Crex crex belongs to the family Rallidae and has a very characteristic, loud craking call, which is uttered by males during the breeding season (Cramp and Simmons, 1980). This call seems to be very simple in design, especially if we compare it to the elaborate songs of passerines. It consists of only two units (syllables SYL1 and SYL2, fig. 1) organised in a very stereotyped manner, and repeated a thousand times during night activity (Green et al., 1997). On the other hand, the corncrake call is structurally a toneless, pulse repetition signal. Such signals consist of a series of energy bursts, often with a broadband spectrum, and generally are more typical of anurans, marine mammals or insects than for birds (Beeman, 1998). The inner call structure, expressed as intervals between maximal amplitude peaks (called pulse-to-pulse duration or PPD), has proved to be individually characteristic and stable over longer periods (Peake et al., 1998; Peake and McGregor, 1999). Individuality in call structure has also been found in other rails (Huxley and Wilkinson, 1979; Clapperton, 1987). The duration of call syllables, and especially intervals, perceived as a specific rhythm of calling, varies both between and within individuals, with clear relations to parameters such as body size and aggressive motivation (Owsi´nski, 2002; Osiejuk et al., 2003, 2004). Those studies show that, despite relative simplicity, the corncrake call is functionally hierarchically organised, and that at least at the two levels recognised — i.e. (i) temporal pattern of amplitude pulses, and (ii) rhythm — the call may encode different information about identity, size and/or motivation. However, it does not mean that there is no further potential variation within the corncrake call. During a 3-year study we analysed over 60 000 calls uttered by ca. 200 males. Preliminary comparisons of call spectrograms, showing energy distribution across frequency domain, revealed that
Corncrake call amplitude spectra
209
Figure 1. Sonogram of a Corncrake call. An illustration of basic temporal call measurements: SYL1, SYL2, INT1 and INT2.
this feature is relatively individually stable and, simultaneously, significantly varies between males. Therefore, we may also expect this acoustic parameter to carry a specific message. Usually, energy distribution across sound frequency is negatively correlated with the body size of the animal. Fundamentally, this is a physical relationship, as it results from the extent that sound properties depend on the sound production apparatus (Bradbury and Vehrencamp, 1998). Body size co-varies with some anatomical and physiological factors, such as tracheal length and vocal tract resonance, which directly affect the signal produced (Wallschläger, 1980; Ryan and Brenowitz 1985; Lambrechts, 1996). Evolutionarily, it was a prerequisite that enabled the frequency component of a signal to function as an honest trait, signalling the size of the sender (index signal type sensu Vehrencamp, 2000). On the other hand, it is much easier to show such a relation for multi-species comparison than within a single species (e.g., Bowman 1979; Shy, 1983). The corncrake seems to be a good subject to study the relationship between voice and body size. Corncrakes inhabit dense habitats in which visual communication is severely limited, and a loud call serves as an elementary signal in long-range communication (Green et al., 1997). Thus, it may be expected to carry crucial information about size, condition, quality, or motivation for potential mates or rivals. As a pulse repetition signal, the corncrake call belongs to less extensively examined types of sounds since, at least in birds, most work has been carried out on tonal signals (Catchpole and Slater, 1995).
210
T.S. Osiejuk, B. Olech
Another possibility is that the pattern of energy distribution enables or facilitates individual recognition. This, however, does not exclude the first hypothesis, as both functions may be fulfilled simultaneously. Furthermore, the fact that the male identity is also encoded by temporal structure of amplitude pulses does not weaken this hypothesis, as redundancy in recognition systems is rather common, especially when many individuals may signalise simultaneously, i.e. in a noisy environment (Robisson, 1992; Aubin and Jouventin, 2002; Jouventin and Aubin, 2002). In this paper we examine to what extent the pattern of energy distribution within the call varies within and among individuals. Therefore, we focused on looking for a potential physical basis for individual recognition (Catchpole and Slater, 1995). Prediction of the individual recognition hypothesis is that betweenindividual variation in variables describing energy distribution within the call should be significantly higher than within-individual variation. Second, we checked how within-call energy distribution pattern depends on the body size of males. Prediction of the body size hypothesis is that there should be a negative relationship between body size and energy distribution pattern, e.g., larger males should have calls with a lower frequency of maximal amplitude.
MATERIAL AND METHODS
Study area and population The study was carried out in the western part of the Kampinoski National Park (Central Poland, 20◦ 23 E 52◦ 19 N), from May to July 2002. The study plot (ca. 24 km2 ), known as ‘Farmułkowskie Łaki’ ˛ is an open area of a narrow swamp belt, 2-3 km long, between two inland dune systems. It constitutes a mosaic of peat land as well as wet meadows and a small proportion of formerly arable land in an early stage of succession after being left uncultivated. The area is naturally closed from the north, east and west by forests, which ensure quite stable breeding conditions depending only on water levels and weather (Michalska-Hejduk, 2001). Corncrakes in the study area are distributed irregularly and the basic parameters of abundance and habitat selection are well-known from earlier studies (Juszczak and Olech, 1997; Owsi´nski, 2002). Recording of birds and individual recognition The study area was visited at night (23:00-04:00) during the calling activity period (7 May-13 July 2002). Birds were recorded (for 2-3 min) from a distance of about 10 m by using a Tonsil Mc-382 directional microphone and Marantz PMD222 cassette recorder. Locations of all calling males were determined by using a GARMIN 12 GPS receiver. After recording, birds were captured in the following way. Field workers trampled 1-2 m2 of vegetation near the place where males were calling. This area was then
Corncrake call amplitude spectra
211
lit using an electric torch and the craking call was played back from a loudspeaker placed nearby. During the playback, people remained in the shadows for up to 30 min. During that time, males often approached the lit area and were caught with a special net (80 cm diam.) attached to a 160 cm long stick. The behaviour of males during capture was noted (time of reaction). Captured birds were ringed with a numbered metal ring, weighed, and biometric measurements (tarsus, wing length and head with bill) taken, which took, in total, ca. 15 min. After this period, the birds were released. Call analysis and bioacoustics terminology All recordings were transferred from the tape recorder to a PC workstation with SoundBlaster Live! 5.1 (full version) using 22.05 kHz / 16 bit sampling. Recordings were analysed by Avisoft SASLab Pro 4.2 software within the following set of parameters: 1024 FFT-length, Frame [%] = 25, Window = Hamming and Temporal Overlap = 98.43%. This gave a 112 Hz bandwidth with 21 Hz frequency and 0.72 ms time resolution (Specht, 2002). We used One-dimensional function called Amplitude spectrum (linear) with Hamming evaluation window with bandwidth 7.806 Hz and resolution 5.363 Hz. Using this function, we measured the following spectral characteristics: FMA [Hz] = frequency of maximal amplitude; L25 [Hz] = frequency below which 25% of the total signal energy is distributed; M50 [Hz] = frequency below which 50% of the total signal energy is distributed; U75 [Hz] = frequency below which 75% of the total signal energy is distributed; MINF [Hz] = minimum frequency and MAXF [Hz] = maximum frequency, for which the syllables’ amplitude falls below −20 dB (relative to the maximum amplitude); BAND [Hz] = bandwidth, which equals MAXF − MINF. For bandwidth-related measurements, the Spectral Characteristics option ‘total’ was on and the minimum frequency range was limited to 0.5 kHz, to remove background noise. The value was established on the basis of an earlier inspection of amplitude distribution within calls. At least for recorded males, MINF of Corncrake calls was always over this 0.5 kHz cut-off frequency. For more details on measurement characteristics, see Specht (2002). For each syllable we also measured the maximum amplitude of a signal (MAXAMP [mV]), which was used as a covariate in some analyses. MAXAMP is a relative measure, but it reflects both the real call intensity and the distance between the male and microphone. Therefore, it could be used as an index of recording quality. We may expect that MAXAMP affects mainly variables like MAXF and BAND, while the others should be less affected by this factor. The measurements are illustrated in figure 2. Data selection and statistical analysis Two different sets of recordings were analysed. The first set consisted of repeated recordings made during a single night of the same ten males. Several factors affect acoustic parameters of the recorded vocalisations. Some of them are intrinsic (or
212
T.S. Osiejuk, B. Olech
Figure 2. Focused sonogram with amplitude spectrum of SYL1 from fig. 1. Arrows indicate all acoustic variables measured. Note that this is only an illustration; measurements were taken automatically with Avisoft SASLab Pro.
sender dependent) and some of them are generated by the environment through which the signal is transmitted (Catchpole and Slater, 1995). Therefore, if we want to find any biological relationships between a male’s quality or size and a signal he utters, we have to remove or take into account environmental effects. To do this, we compared calls of the same ten males recorded during a single night in three sessions separated by at least 1 h. We might expect that each session took place in more or less different recording conditions that arise from changes in temperature, humidity, distance, and position of the bird in relation to the person recording. In this case, males were recognised on the basis of pulse-to-pulse duration (PPD), according to the method described by Peak et al. (1998). We used only the first 11 PPD from each syllable, as this was the lowest number of amplitude pulses found in our population (Olech et al., 2002). We also calibrated the method for the studied population by comparing PPD variation within and between recordings of the same and different males identified by rings. Details of the method are described elsewhere. The second set consisted of single recordings of 84 males which were caught after recording and for which we had body size measurements. Sample size varied between analyses as we do not have complete measurements for all males. For each recording, we measured the above-described parameters for ten calls, i.e. for ten SYL1 and ten SYL2. Those ten calls were always chosen randomly from the whole set of recordings of the male, which usually consisted of ca. 100 calls. We measured only those calls that were free from any background noise, including sounds produced by other corncrakes and animals.
Corncrake call amplitude spectra
213
Statistics were calculated by SPSS 10 software (Norusis, 1993). Data are presented as means ± SE, if not indicated otherwise. P values are two-tailed, unless stated otherwise. We used multivariate general linear models (GLM) to estimate the effect of male, syllable and recording session on measured acoustic parameters. Discrimination analysis was applied to classify males based on the acoustic parameters of their call spectrum. In all cases Pearson’s correlation coefficients were calculated. To reduce the number of variables for body size, we used one parameter: the first principal component based on the three size parameters. We refer to this new variable as ‘body size’. It explained 52.0% of all body size measurements and correlated positively with tarsus (r = 0.64), head with bill (r = 0.81), and wing (r = 0.70). Body weight can vary due to differences in nutritional stores, but also to differences in structural body size. We therefore calculated residual mass, which in this study is the residual of a regression of body mass on body size (R 2 = 0.15, N = 78, P < 0.001; mass = 169.5 + 3.94 * body size), and may be used as a measurement of bird condition.
RESULTS
Within-individual variation in acoustic parameters of corncrake calls Amplitude spectra describing energy distribution across the frequency domain highly varied between males (ANOVA, all P < 0.001), and for some variables also between syllables within a call (table 1). We used repeated recordings of the same males to examine factors affecting within- and between-individual variation in acoustic parameters. We applied a multivariate GLM with seven acoustic parameters as explained variables and MALE (1-10), SYL (1-2) and SESSION (1-3) as independent variables, and MAXAMP as a covariate (table 2). The model showed that there were no significant differences in acoustic properties between SYL1 and SYL2 of a male. Only in the case of the L25 variable was the difference significant. Both MALE and SESSION significantly affected all acoustic parameters of calls. In all cases except MAXF and BAND, the effect of MALE was much stronger than the effect of recording SESSION (table 2). The effect of MAXAMP was significant in all variables except FMA. Discriminant analysis revealed that these seven basic acoustic measurements enable a correct classification of 68.8% of cases. Rapid changes in FMA Analysing calls from both data sets, we found that some males exhibited a relatively high variation in FMA variable, which seemed to be independent of recording quality, as FMA was the only variable unaffected by MAXAMP (table 2). A detailed comparison of successive calls of such males revealed that FMA may drastically change between successive performances. The amplitude spectrum of corncrake calls often exhibits two or more distinct energy peaks located on well-separated
214
T.S. Osiejuk, B. Olech
Table 1. Basic acoustic parameters of corncrake calls (N = 84). Variable
SYL1
SYL2
min-max
mean ± SE
min-max
mean ± SE
FMA
1.35-6.27
4.88 ± 0.093
1.55-6.02
4.83 ± 0.089
L25
2.24-4.84
3.66 ± 0.064
2.25-4.80
3.64 ± 0.064
M50
2.89-5.81
4.79 ± 0.062
2.90-5.83
4.77 ± 0.061
U75
4.13-6.63
5.71 ± 0.054
4.19-6.66
5.71 ± 0.053
MINF
0.68-2.21
1.03 ± 0.027
0.63-2.26
1.05 ± 0.032
MAXF
5.37-9.88
7.31 ± 0.077
5.50-9.70
7.33 ± 0.074
BAND
3.62-8.81
6.28 ± 0.088
3.90-8.77
6.28 ± 0.086
Significance of differences between SYL1 and SYL2 t = 2.44, df = 83, P = 0.017 t = 3.60, df = 83, P = 0.001 t = 4.01, df = 83, P = 0.003 t = −0.90, df = 83, P = 0.373 t = −1.66, df = 83, P = 0.101 t = −1.152, df = 83, P = 0.253 t = 0.09, df = 83, P = 0.925
Table 2. General Linear Model explaining variation of seven acoustic variables describing the corncrake call. Acoustic variable FMA L25 M50 U75 MINF MAXF BAND
MALE
SESSION
SYL
MAXAMP
F9,599
P
F2,599
P
F1,599
P
F1,599
P
97.18 2081.24 887.82 501.93 54.41 43.82 81.15