acta ethol (2002) 4:65–72 DOI 10.1007/s10211-001-0056-8

O R I G I N A L A RT I C L E

Torben Dabelsteen · Nicolas Mathevon

Why do songbirds sing intensively at dawn? A test of the acoustic transmission hypothesis Received: 26 September 2000 / Received in revised form: 24 September 2001 / Accepted: 2 November 2001 Published online: 11 January 2002 © Springer-Verlag and ISPA 2002

Abstract Among the ideas proposed to explain the existence of the dawn chorus in songbirds, the acoustic transmission hypothesis claims that birds sing most intensively at dawn because this is the time of the day when songs suffer least from environmentally induced degradation and hence propagate over the longest distances. In this article, we report on the first sound transmission experiment that directly tests this assumption using natural song from a typically forest-living dawn chorusing bird, the blackcap Sylvia atricapilla. Representative sound elements from the introductory twitter part and from the terminating motif part of the blackcap song were transmitted and re-recorded at three different times of the day: dawn, midmorning, and early afternoon. These recordings were then compared with respect to the following measures of sound degradation: signal-tonoise ratio (SNR), excess attenuation, blurring over song elements, and elongation of song elements by tails of echoes. As could be expected, both the background noise and the SNR varied considerably over the day. More surprisingly the excess attenuation decreased during the day, being lowest in the afternoon. There was no diurnal variation in blurring and elongation by echoes. The results may be explained by the diurnal variation in physical parameters such as temperature, relative humidity, and wind speed. The implications of this for different Communicated by P.K. McGregor T. Dabelsteen (✉) Department of Animal Behaviour & Centre for Sound Communication, Zoological Institute, University of Copenhagen, Tagensvej 16, 2200 Copenhagen N, Denmark e-mail: [email protected] N. Mathevon Laboratoire de Biologie Animale et Appliquée, Université Jean Monnet, 23 rue du Dr. Michelon, 42023 Saint-Etienne cedex 02, France N. Mathevon NAM Mécanismes de Communication, CNRS UMR 8620, Universite Paris XI-Orsay, France

communication activities are discussed. Overall, the results show that dawn conditions in a temperate deciduous forest do not always constitute the best circumstances for long-range communication and therefore that the dawn chorus cannot be explained by the sound transmission hypothesis. Keywords Dawn chorus · Sound transmission · Sound degradation · Bird song · Communication networks

Introduction In many bird species, the song activity of the males follow a circadian rhythm where the long-ranging advertising full song is sung most intensively at dawn, creating the so-called “dawn chorus”. A number of different hypotheses explaining the existence of the dawn chorus have been put forward (e.g. Kacelnik and Krebs 1982; Mace 1987; Cuthill and Macdonald 1990; Staicer et al. 1996). These can be classified into three main categories: those dealing with mechanisms where the dawn chorus is explained by, for instance, increased levels of song-controlling hormones; those dealing with the social functions of song in the sense that these are best served at dawn; and those dealing with how diurnal variations in environmental constraints favour dawn singing. The acoustic transmission hypothesis predicts that birds sing their full songs most intensively at dawn because this is the time of the day when these songs propagate most effectively and hence should be most effective for long-range communication (e.g. Henwood and Fabrick 1979). Henwood and Fabrick (1979) used a mathematical model developed for calculating sound attenuation and broadcast range under different microclimatic conditions to explore sound transmission at different times of the day. This model suggested that sounds will be transmitted over a wider area at dawn than at midday. Also in support of the acoustic transmission hypothesis is the common subjective experience that sounds may sometimes be heard over relatively long dis-

66

tances early in the morning before the start of the dawn chorus when there is no background noise from vocalising birds. However, the hypothesis has never, as far as we know, been tested in a song transmission experiment specifically quantifying the diurnal variation in those aspects of environment-induced song degradation that are important for song communication. There are at least three such aspects: (1) a reduction in signal-to-noise ratio (SNR) of the songs caused by addition of background noise, attenuation from spherical spreading (6 dB/dd, doubling of distance), and excess attenuation (EA, attenuation in excess of the above mentioned 6 dB/dd) due to absorption and multiple scattering; (2) an elongation of the song elements by “tails” of reverberation caused echoes that may fill in the inter-element pauses and overlap succeeding elements (forward masking); and (3) a blurring of amplitude and frequency patterns over time caused by selective frequency filtering, atmospheric turbulence, and reverberation (e.g. Wiley and Richards 1978, 1982; Dabelsteen et al. 1993; Mathevon et al. 1996; Holland et al. 1998, in press). Some of the factors responsible for these aspects of song degradation are likely to vary over the day, for example, atmospheric turbulence, absorption, and level of abiotic background noise (e.g. Henwood and Fabrick 1979; Wiley and Richards 1982; Larom et al. 1997). On a normal day during the breeding season, one would therefore predict a diurnal variation in sound degradation. Starting at dawn, the wind speed increases until some time later in the day, causing an increase in atmospheric turbulence and hence an increase in blurring and background noise from the vegetation. The latter will reduce the SNR. The temperature generally increases and may become highest near ground level, creating a negative temperature gradient, which refracts the sounds upwards and creates a sound shadow. In effect the sound shadow will increase the EA and therefore reduce the SNR. The same effects are obtained from the increase in atmospheric absorption induced by both the rise in temperature and a decrease in relative humidity. Finally, the background noise from the many different types of calls coupled with the general activity of the birds will normally decrease and therefore contribute to an increase of the SNR. The predictions for blurring and EA are clear whereas those for SNR are conflicting, and it is difficult to make predictions for the “tails” of echoes. We still have very few quantitative data for this new tail measure (Holland et al. in press), but it is probably less affected by time of day than the other measures because it primarily depends on large reverberating surfaces. Also, song communication in a network environment involves activities with conflicting requirements for sound transmission (e.g. McGregor and Dabelsteen 1996). It is therefore not easy to predict the time of day that offers the best conditions for song communication. An increase in EA and blurring, for instance, will constrain extraction of information from the songs even though information often is encoded into song features that are relatively resistant to

Fig. 1 Sonograms and waveforms of the blackcap test sounds: five T-sounds and five M-sounds

degradation (e.g. Brémond and Aubin 1990; Dabelsteen and Pedersen 1992; Dabelsteen et al. 1993; Robisson et al. 1993) and it will also constrain song-based addressing of particular receivers using song matching or special timing of singing (e.g. McGregor et al. 1992; Dabelsteen et. al. 1997; Naguib 1999). However, the same increase in EA and blurring might help privatise communication to counteract undesired eavesdropping on singing interactions (e.g. McGregor and Dabelsteen 1996; Dabelsteen et al. 1998). An increase in “tail” energy would probably have the same effect because “tails” may mask successive song elements (Holland et al. in press). An increase in SNR would have the opposite effect. Finally, since song degradation seems to provide cues for song-based ranging (e.g. McGregor 1994; Naguib 1995, 1997; Nelson and Stoddard 1998; Nelson 2000), a diurnal variation in the different aspects of degradation, in whatever direction, would probably complicate ranging. Here we present a direct test of the acoustic transmission hypothesis. We use a sound transmission experiment to investigate the idea that dawn provides the best conditions for song propagation and consider a number of different communication activities. The full song of the blackcap (Sylvia atricapilla), a species that typically sings a dawn chorus, constituted the experimental subject. The song starts with a highly variable and sometimes very long first segment of relatively weak and high-pitched twittering sounds (T-sounds) and ends with a louder and relatively short species-specific motif of

67 Table 1 Weather conditions during the experiment

Temperature (°C) Wind speed (m/s) Relative humidity (%)

Measurement height

Dawn 0435–0450 hours

Midmorning 0915–1000 hours

Afternoon 1305–1350 hours

3m 6m 3m 6m 3m 6m

3.7 1.7 0.2 0.3 81 98

6.9 7.1 0.7 0.9 70 72

11.3 9.5 1.3 1.3 57 62

pure fluting tones (M-sounds) (e.g. Cramp 1992; Fig. 1). Both sound types were broadcast and re-recorded at different times of the day, and the diurnal variation in all of the above-mentioned aspects of song degradation were quantified.

Methods Location and weather conditions The test sounds were transmitted through deciduous primeval forest at Strødam Biological Field Station, Denmark. The experimental area constitutes a typical blackcap habitat. It is dominated by 20-m-high trees, mainly beeches (Fagus sylvatica) and birches (Betula pendula), which have trunks reaching a diameter of 0.3–0.4 m and canopies starting about 6 m above the ground. The area below the canopy has occasional dense undergrowth of bushes and young trees of various heights and species. The ground is covered by grass and herbs, leaf litter, and fallen trees. The experiment was performed on 19 April 1998, a few days before the return of the first blackcap migrants. The weather was typical for the season (Table 1), with respect to both relative humidity, which decreased over the day, and temperature, which increased. A temperature gradient with the highest temperature near ground level developed in the afternoon. The wind speed also increased, but only slightly. The wind conditions can vary quite a lot at this time of year, but days with only little wind are common (e.g. see the Danish Meteorological Institute’s Web archive at http://www.dmi.dk/vejr/klima/webarkiv_dk/webarkiv.html). The present experiment was made on one of five such consecutive days. Test sounds and transmission equipment The test sounds were extracted from high-quality recordings made at a distance of about 3 m from each of two singing birds. The recordings were made by C. Chappuis using a Nagra type IIIB tape recorder and a Sennheiser MD211 N microphone. We selected ten representative sound elements, five T-sounds from the introductory part of the song (T1–T5), and five M-sounds from the terminating part of the song (M1–M5; Fig. 1). T1 originated from one of the males, the remaining sounds from a song of the other male. The two songs were digitised (16-bit OROS acquisition card, sample frequency 20 kHz) and the ten sounds successively isolated using the DSP program SYNTANA (Aubin 1994) and transferred to a DAT test tape. The test tape had five repetitions of a series of the five M-sounds followed by five repetitions of a series of the five T-sounds. We inserted 1-s pauses between the sounds and 1.5-s pauses began and ended the test sequence. The sounds were played back from a Sony TCD-D10 PRO DAT recorder through a Denon DCA-600 power amplifier, highpass filtered (1.2 kHz, f–3 dB), and broadcasted from a VIFA 1'' Neodymium tweeter (Larsen and Dabelsteen 1997). The system was calibrated to yield a sound pressure level (SPL) that preserved

the natural difference between T- and M-sounds. At a distance of 1.5 m from the loudspeaker T- and M-sounds therefore reached maximum SPLs of about 78 and 85 dB(A), respectively. The transmitted sounds were re-recorded through a Brüel & Kjær 1/2'' condenser microphone (type 4188) covered with a windscreen (type UA 0459) and fixed in horizontal position pointing towards the loudspeaker. The microphone was connected to a Brüel & Kjær precision integrating sound level meter (type 2236) with a 10-m extension cable. The output from the sound level meter was recorded on a HHb Portadat PDR 1000 DAT recorder. Both the loudspeaker and the microphone were mounted on telescopic Clark masts (type QT 15 N/HP). Experimental design The test sounds were transmitted according to a 10×2×3 (test sound×microphone height×time of day) factorial design. Since the test tape contained five repetitions of each sound, we had five replications for each playback trial. The positions of the loudspeaker and the microphone were representative of natural positions of respectively male singers and receivers of both sexes: loudspeaker 9 m above the ground (height of blackcap song posts 4.5–13 m above the ground; Cramp 1992); microphone 3 and 9 m above the ground; horizontal distance between loudspeaker and microphone 50 m, corresponding to a typical distance between two territory owners. The three times of the day were chosen as follows: at dawn 15 min before the start of the dawn chorus; in the middle of the morning a couple of hours after the dawn chorus had stopped; and early in the afternoon when the temperature reaches its maximal value (Table 1). The test tape was played at least twice at each microphone position to ensure that we obtained recordings that were not corrupted by loud vocalisations from other birds. In the afternoon and especially during the morning we had to play the tape several times to obtain such recordings. Since the broadcasting system modifies the original characteristics of the test sounds slightly, we also performed a recording close to the loudspeaker (both loudspeaker and microphone 9 m above the ground, distance between them 1.5 m). These sounds constituted the model sounds. The sounds transmitted over 50 m constituted the observation sounds. By comparing the observation sounds with the model sounds, which had suffered only negligible habitat-caused degradation, we determined the effect of the 50-m transmission and at the same time controlled for the effect of the equipment. Data analysis For each sound type and treatment we selected for analysis the first two observation sounds that were not overlapped by transient noise or had such noise in their successive pauses. Transient noise typically came from aeroplanes flying over or from birds emitting loud vocalisations close to our microphone. Weak vocalisations that could not be identified by clearly distinctive peaks on waveform displays were considered part of the stationary, ambient background noise. Our quantification of the different degradation measures involved a compensation for the contribution of the am-

68 bient noise to the observation sounds. We determined this ambient noise on the basis of a 1-s segment sliced from one of the 1.5-s pauses of each recording, assuming that this pause-noise was representative of the noise on top of the observation sounds of the same recording. Among the model sounds, the SNRs of which were very high, we selected those that were surrounded by the least background noise on the model-sound recording. The selections were made on the basis of spectrographic and waveform analyses. The sounds were band-pass filtered at 1–9 kHz (Stanford Research System, type SR650, 115 dB/octave) and then fed into a Kay Elemetrics Corp. DSP 5500 Sona-graph (bandwidth 300 Hz, frequency range 0–8,000 Hz, Hamming window, dynamic range 40 dB). The selected sounds were digitised using a sample frequency of 22,050 Hz. The A/D conversion and the subsequent filtering and signal analysis were made on a PC equipped with Signal Data DSP SPB2 signal processor board, using the program SIGPRO, version 1.3 (Pedersen 1998). The analysis allowed us to quantify all of the aspects of song degradation mentioned in the introduction. Each sound and noise segment belonging to it was first band-pass filtered using the sound-specific filter bandwidths that matched the frequency ranges of the sounds (sound: frequency range in kilohertz; T1: 3.7–8.7; T2: 2.1–6.6; T3: 2.3–4.7; T4: 2.2–8.0; T5: 3.9–7.7; M1: 1.7–3.3; M2: 2.1–3.7; M3: 2.5–3.5; M4: 2.1–3.3; M5: 2.4.6.2; Fig. 1). For each observation sound the SNR and the measures of the “tails” of echoes were then determined from the signal waveforms, whereas the EA and the blurring were determined from amplitude (envelope) functions of the sounds. The mean energy of the ambient background noise per unit time was first determined from the 1-s noise segment. The waveforms of the model and the observation sound were then aligned in time by maximising the cross-correlation between them. In the aligned position we determined the energy of the model part (SE) and the tail part (TE) of the observation sound and calculated the SNR: SNR=10 log [(SE–NEs)/NEs] where NEs was the noise energy over the duration of the model part of the observation sound. A tail-to-signal ratio (TSR) was also calculated: TSR=10 log [(TE–NEt)/(SE–NEs)] where NEt was the noise energy over the duration of the tail. The rate of tail decline (RTD) was determined from measurements of the time taken for TE to be reduced to three-quarters, one-half, and one-quarter. The amplitude functions (AFs) of the model and the observation sound (extracted through Hilbert transformation; Dabelsteen and Pedersen 1985) were also aligned in time by maximising the cross-correlation between them. In the aligned position we calculated the energies of the model part of the observation AF (SEAF) and the model AF, and the ratio kAF between them was determined. The EA was then determined: EA=–20 log (kAF)–AG, where AG is the attenuation of the sound due to geometric spreading. In the aligned position we finally determined the difference signal between the model part of the observation AF and the kAF-attenuated model AF. The blurring expressed as a blur ratio (BR) was then calculated as the ratio between the energy of this difference signal (XEAF) and SEAF: BR=XEAF/SEAF. For details of the analysis, see Dabelsteen et al. (1993) and Holland et al. (1998, in press).

Statistics The data for each measure were subjected to a multi-factorial analysis of variance (ANOVA) [3×2×2 (time of day×sound type×microphone height)] with ten replications. The ten replications result from each sound type being represented by five different sounds and the two repetitions of all possible factor combinations. The data sets for the different measures were slightly lower than 120 for different reasons. For instance, we lacked data for the AF measures (EA and BR) for T2 because this sound is rich in overtones and hence has an AF that is difficult to interpret and makes a calculation of EA and BR dubious. Background noise corruption of entire observation sounds or part of them constituted a general problem, especially for the broad-frequency-range Tsounds. Altogether, this resulted in data sets of 118 for background noise, 117 for SNR, 111 for the tail measures, and 107 for EA and

BR. The statistical calculations were made with Statgraphics Plus, version 1 (Manugistics Inc.) and all graphs with Sigma Plot 4.0 (SPSS Inc.).

Results Together, the main factors, time of day, sound type, and microphone height, and their two-factor interactions explained a substantial part of the variation in most of our measures, largest in those linked to the background noise and the attenuation (background noise: 87%; SNR: 80%; EA: 69%), and smallest in those resulting mainly from reverberation (BR: 34%; TSR: 38%; time taken for tail energy to decay to three-quarters, one-half, and onequarter: 15%, 28%, and 30%, respectively; see the Appendix, which shows the ANOVA tables). The measures linked to the background noise and the attenuation also constituted those that varied most reliably with the time of day. The level of the background noise varied significantly over the time of day. With the present design it was highest in the midmorning and lowest at dawn immediately before the start of the dawn chorus (Fig. 2a). The SNR varied accordingly, being highest at dawn and lowest in the midmorning (Fig. 2b). The time of day also had a significant effect on EA, which decreased gradually over the day (Fig. 2c). However, neither BR nor any of the tail measures showed a diurnal variation (Fig. 2d–f). As would be expected, the sound type (M- or Tsounds) had a significant effect on the background noise and nearly all the degradation measures. There was the most background noise in the frequency range of the Tsounds, (mean±SE, T: 19.5±1.2 dB, M: 6.1±0.6 dB), which also had the largest EA (T: 8.9±0.4 dB, M: 7.9±0.3 dB). Accordingly, the T-sounds had the lowest SNR (T: 19.3±0.6 dB, M: 26.0±0.8 dB). However, the Tsounds also had the lowest BR (T: 0.15±0.01, M: 0.20±0.01). The TSR did not vary between the two sound types (T: –4.4±0.4 dB, M: –4.6±0.3 dB), but the T-sounds had the fastest RTD. The average effects of the factors time of day and sound type hide some interesting significant two-factor interaction effects. For instance, the M-sounds were considerably less attenuated than the T-sounds in the afternoon (Fig. 3a). The significant two-factor-interaction effects of sound type by time of day on the tail decay measures (time from start of tail to where the energy of the tail was reduced to three-quarters, one-half, and onequarter) show that the RTD of the T-sounds was different from that of the M-sounds only at dawn and in the afternoon, but not in the midmorning (Fig. 3b). The microphone height had a significant effect on all the degradation measures except on the time taken for the tail energy to decay to three-quarters, one-half, and one-quarter, that is, on the RTD. As expected from the results of previous sound transmission experiments the low microphone height (3 m) resulted in more degradation than the high position (9 m), that is, the lowest SNR (3 m: 21.3±0.8 dB, 9 m: 24.2±0.8 dB), and the highest

69 Fig. 2a–f Main effects of the time of day on background noise and degradation cues. D Dawn, M midmorning, A afternoon. Error bars indicate standard errors, asterisks significant differences obtained with the 95% least significant difference multiple range test

EA (3 m: 9.9±0.3 dB, 9 m: 6.8±0.2 dB), BR (3 m: 0.22±0.01, 9 m: 0.14±0.01), and TSR (3 m: –3.3±0.3 dB, 9 m: –5.8±0.3 dB). The background noise did not vary with microphone height (3 m: 12.8±1.3 dB, 9 m: 12.6±1.3 dB).

Discussion

Fig. 3 Sound type by time of day interaction effects on excess attenuation (a) and tail decay rate (b). D Dawn, M midmorning, A afternoon; MD, MM, and MA: M-sounds at dawn, midmorning, and afternoon, respectively; TD, TM, and TA: T-sounds at dawn, midmorning, and afternoon, respectively. Error bars indicate standard errors

Only the background noise, the SNR, and the EA showed a diurnal variation. EA decreased from dawn to midmorning and further on until the afternoon, whereas the background noise was lowest at dawn and highest in the midmorning. The SNR was highest at dawn and lowest in the midmorning. However, in the present design “dawn” refers to a few minutes before the actual start of the dawn chorus. We did not include data directly from the time during the dawn chorus because the many vocalisations at this time would have resulted in a very high background noise and an extensive masking of the transmitted observation sounds. This would have made a calculation of most of the present aspects of the sound degradation dubious or at least very difficult. Even in the midmorning recordings where bird vocal activity is still very high although less intense than at dawn, it was diffi-

70

cult to obtain tail measures uncorrupted by vocalisation noise. For instance, the relatively slow decay rate of the T-sounds in the midmorning indicates that there is a diurnal variation in the decay rate of the T-sounds (Fig. 3b). However, the frequent masking of the tails in the midmorning makes it likely that the slow decay rate at this time of day could be an artefact of our inability to compensate with the correct background noise in the calculations of the RTD measures. Undoubtedly, the background noise would have been highest during the dawn chorus, and the SNR accordingly lowest. Since there will only be a slight change in the physical conditions from immediately before to during the dawn chorus, our other dawn measures would have been almost the same had they been taken during the actual dawn chorus. The significant effect of the time of day on EA, therefore, means that the EA had been highest during the dawn chorus and then decreased gradually over the day (Fig. 2c). Interestingly, the EA of the Tsounds did not decrease further between midmorning and afternoon, whereas that of the M-sounds showed a considerable fall in that period (Fig. 3a). The fall in EA during the day is difficult to explain. The increase in temperature and the decrease in relative humidity (Table 1) should both have increased the atmospheric absorption and hence increased the EA, at least for frequencies larger than about 3 kHz (e.g. Ingård 1953; Wiley and Richards 1982). However, a decrease in multiple scattering from leaves might have contributed to the fall in EA, because the surface impedance and the density of the leaves are likely to have decreased as a result of evaporation. Several factors may have contributed to the afternoon difference in EA between the low frequency Msounds and the high frequency T-sounds. For instance, the above-mentioned increase in attenuation with decreasing relative humidity is larger for high- than for low-frequency sounds (e.g. Ingård 1953; Delany 1977). Since the undergrowth was gradually coming into leaf in the days around the experiment it is also possible that there were more and slightly larger leaves in the afternoon. Such progress in the leafing process would cause more scattering of high-frequency sounds than of low-frequency ones (e.g. Price et al. 1988). Finally, the development of a negative temperature gradient in the afternoon may have caused an upwards refraction of the sounds, more so for the high-frequency T-sounds than the low-frequency M-sounds and therefore more attenuating for the T-sounds (e.g. Ingård 1953; Pridmore-Brown and Ingard 1955). The lack of diurnal variation in BR, TSR, and RTD is perhaps not surprising giving that these measures are all strongly affected by reverberation from large reflecting surfaces, and these, for instance, tree trunks and branches, do not change over the day. The possible effect of the progress in leafing seems not to have had a strong influence on these measures. Our results imply that blackcaps will not necessarily improve song-based information transfer and addressing of particular receivers over long ranges by singing at dawn rather than later in the day. On the contrary, the gradual de-

crease in EA during the day implies that these aspects of communication would be most effectively performed later in the morning and in the early afternoon. Using high perches seems to facilitate these aspects of information transfer in blackcaps as in other songbird species (e.g. Dabelsteen et al. 1993; Mathevon et al. 1996; Holland et al. 1998). Especially, the species-specific M-sounds are very effective for information transfer and addressing of receivers in the afternoon (Fig. 3a). For T-sounds the diurnal variation in transmission capability is less pronounced. The variation in SNR supports the finding that the transmission conditions generally are best in the afternoon. However, although the diurnal variation in SNR is influenced by the variation in EA, it is mainly a consequence of the high variation in background noise caused by vocalising birds and therefore cannot be used to explain why dawn chorusing has evolved. The BR and the tail measures, which are not directly related to the background noise, are believed to represent potential ranging cues (e.g. McGregor 1994; Naguib 1995), and these measures did not vary with the time of day. Therefore, blackcaps using these types of cues for ranging would not face complications caused by a diurnal variation of these cues. Of course, masking from background noise must impair all these communication aspects. On the other hand, the high EA and the low SNR do provide optimal conditions for privatising communication, and since the EA is higher for Tsounds than for M-sounds, especially in the afternoon, privatised singing would be facilitated by using T-sounds. To our knowledge, this is the first experiment that directly tests the acoustic transmission hypothesis that dawn choruses exist because songs propagate best and hence make song communication most effective at dawn. Our results from a temperate deciduous forest, especially those relating to information transfer and addressing of receivers, do not support the hypothesis. On days with weather conditions like on the present experimental day, blackcap song seems to propagate best in the afternoon. Of course, the weather conditions can be very different, for instance, more windy with an increase in the windcaused background noise and a resulting decrease in SNR during the day. However, the present conditions do occur regularly at the start of the blackcap breeding season, and the song activity is usually most intensive on such calm days, at dawn as well as later in the day. Altogether, our results suggest that there may be many days during the breeding seasons of many songbird species where their song propagation is not necessarily most effective at dawn, and that we need to consider other factors, for instance, social ones, to explain the existence of dawn choruses in temperate deciduous forests. Acknowledgements We thank Jo Holland for her helpful comments on an earlier version of this manuscript and Claude Chappuis for use of his high-quality recordings of blackcap song. Centre for Sound Communication (CSC), which is financed by the Danish National Research Foundation, funded the equipment used in this study and also the expenses in connection with Nicolas Mathevon’s stay in Denmark. The experiment complies with the current laws of Denmark.

71

Appendix ANOVA tables ANOVA tables are presented for the eight dependent variables: background noise, signal-to-noise ratio, excess % variance explained Background noise Main effects (A) Time of day (B) Sound type (C) Microphone height Interaction effects A×B A×C B×C Signal-to-noise ratio Main effects (A) Time of day (B) Sound type (C) Microphone height Interaction effects A×B A×C B×C Excess attenuation Main effects (A) Time of day (B) Sound type (C) Microphone height Interaction effects A×B A×C B×C Blur ratio Main effects (A) Time of day (B) Sound type (C) Microphone height Interaction effects A×B A×C B×C

df

F-ratio

attenuation, blur ratio, tail-to-signal ratio, and time from start to when, respectively, 75%, 50%, and 25% of tail energy remains (see the text).

P-value

35.0 46.8 0.0

2 1 1

132.36 353.66 0.23

<0.01 <0.01 0.63

5.2 0.0 0.0

2 2 1

19.68 0.10 0.23

<0.01 0.90 0.63

42.8 28.4 5.3

2 1 1

111.77 148.25 27.76

<0.01 <0.01 <0.01

3.0 0.0 0.1

2 2 1

7.88 0.11 0.24

<0.01 0.90 0.62

9.4 4.8 48.1

2 1 1

14.28 14.50 146.05

<0.01 <0.01 <0.01

3.3 0.4 3.0

2 2 1

5.04 0.61 9.04

0.01 0.55 <0.01

0.1 8.8 25.3

2 1 1

0.07 13.14 37.66

0.93 <0.01 <0.01

0.1 0.0 0.1

2 2 1

0.04 0.02 0.19

0.96 0.98 0.67

References Aubin T (1994) La communication acoustique chez l’oiseau. Etude des méchanismes de codage-décodage des signaux. Adaptation á des conditions défavorables de propagation. Habilitation á diriger des recherches, Université de Nancy I, France Brémond JC, Aubin T (1990) Responses to distress calls by blackheaded gulls Larus ridibundus: the role of non-degraded features. Anim Behav 39:503–511 Cramp S (ed) (1992) The birds of the western Palearctic, vol VI. Oxford University Press, Oxford Cuthill IC, Macdonald WA (1990) Experimental manipulation of the dawn and dusk chorus in the blackbird Turdus merula. Behav Ecol Sociobiol 26:209–216

Tail-to-signal ratio Main effects (A) Time of day (B) Sound type (C) Microphone height Interaction effects A×B A×C B×C

% variance explained

df

F-ratio

P-value

1.6 0.2 27.5

2 1 1

1.23 0.36 42.46

0.30 0.55 <0.01

2.7 0.8 4.8

2 2 1

2.09 0.59 7.45

0.13 0.56 0.01

Time from start of tail to when 75% of tail energy remains Main effects (A) Time of day 0.8 2 0.48 0.62 (B) Sound type 7.4 1 8.95 <0.01 (C) Microphone height 0.6 1 0.72 0.40 Interaction effects A×B 4.9 2 2.96 0.06 A×C 1.2 2 0.72 0.49 B×C 0.0 1 0.05 0.82 Time from start of tail to when 50% of tail energy remains Main effects (A) Time of day 3.1 2 2.23 0.11 (B) Sound type 14.5 1 21.03 <0.01 (C) Microphone height 1.1 1 1.61 0.21 Interaction effects A×B 5.5 2 4.00 0.02 A×C 2.5 2 1.81 0.17 B×C 1.3 1 1.88 0.17 Time from start of tail to when 25% of tail energy remains Main effects (A) Time of day 1.96 2 1.48 0.23 (B) Sound type 11.9 1 17.88 <0.01 (C) Microphone height 0.4 1 0.65 0.42 Interaction effects A×B 9.2 2 6.92 <0.01 A×C 4.8 2 3.62 0.03 B×C 1.5 1 2.25 0.14

Dabelsteen T, Pedersen SB (1985) A method for computerized modification of certain natural animal sounds for communication study purposes. Biol Cybern 52:399–404 Dabelsteen T, Pedersen SB (1992) Song features essential for species discrimination and behaviour assessment by male blackbirds (Turdus merula). Behaviour 121:259–287 Dabelsteen T, Pedersen SB, Larsen ON (1993) Habitat-induced degradation of sound signals: quantifying the effects of communication sounds and bird location on blur ratio, excess attenuation and signal-to-noise ratio. J Acoust Soc Am 93:2206–2220 Dabelsteen T, McGregor PK, Holland J, Tobias JA, Pedersen SB (1997) The signal function of overlapping singing in male robins. Anim Behav 53:249–256

72 Dabelsteen T, McGregor PK, Lampe HM, Langmore NE, Holland J (1998) Quiet singing in song birds: an overlooked phenomenon. Bioacoustics 9:89–105 Delany ME (1977) Sound propagation in the atmosphere: a historical review. Acustica 38:201–223 Henwood K, Fabrick A (1979) A quantitative analysis of the dawn chorus: temporal selection for communicatory optimization. Am Nat 114:260–274 Holland J, Dabelsteen T, Pedersen SB, Larsen ON (1998) Degradation of song in the wren Troglodytes troglodytes: implications for information transfer and ranging. J Acoust Soc Am 103:2154–2166 Holland J, Dabelsteen T, Pedersen SB, Paris AL (in press) Potential ranging cues contained within the energetic pauses of transmitted wren song. Bioacoustics Ingård U (1953) A review of the influence of meteorological conditions on sound propagation. J Acoust Soc Am 25:405–411 Kacelnik A, Krebs JR (1982) The dawn chorus in the great tit (Parus major): proximate and ultimate causes. Behaviour 83:287–309 Larom D, Garstang M, Payne K, Raspet R, Lindeque M (1997) The influence of surface atmospheric conditions on the range and area reached by animal vocalizations. J Exp Biol 200:421–431 Larsen ON, Dabelsteen T (1997) The 1” Vifa Neodymium Tweeter: a versatile speaker for playback experiments. Bioacoustics 8:323–326 Mace R (1987) Why do birds sing at dawn? Ardea 75:123–132 MatheN von, Aubin T, Dabelsteen T (1996) Song degradation during propagation: importance of song post for the wren Troglodytes troglodytes. Ethology 102:397–412 McGregor PK (1994) Sound cues to distance: the perception of range. In: Green PR, Davis M (eds) Perception and motor control in birds. Springer, Berlin Heidelberg New York, pp 74–94 McGregor PK, Dabelsteen T (1996) Communication networks. In: Kroodsma DE, Miller EH (eds) Ecology and evolution of acoustic communication in birds. Cornell University Press, Ithaca, N.Y. pp 409–425 McGregor PK, Dabelsteen T, Shepherd M, Pedersen SB (1992) The signal value of matched singing in great tits: evidence from interactive playback experiments. Anim Behav 43:987–998

Naguib M (1995) Auditory distance estimation of singing conspecifics in Carolina wrens: the role of reverberation and frequency-dependent attenuation. Anim Behav 50:1297–1307 Naguib M (1997) Use of song amplitude for ranging in Carolina wrens, Thryothorus ludovicianus. Ethology 103:723–731 Naguib M (1999) Effects of song overlapping and alternating on nocturnally singing nightingales. Anim Behav 58:1061–1067 Nelson BS (2000) Avian dependence on sound pressure as an auditory distance cue. Anim Behav 59:57–67 Nelson BS, Stoddard PK (1998) Accuracy of auditory distance and azimuth perception by a passerine bird in natural habitat. Anim Behav 56:467–477 Pedersen SB (1998) Preliminary operation manual for signal processor SIGPRO, ver. 1.3. Centre for Sound Communication, Odense University, Denmark Price MA, Attenborough K, Heap NW (1988) Sound attenuation through trees: measurements and models. J Acoust Soc Am 84:1836–1844 Pridmore-Brown DC, Ingard U (1955) Sound propagation into the shadow zone in a temperature-stratified atmosphere above a plane boundary. J Acoust Soc Am 27:36–42 Robisson P, Aubin T, Brémond J-C (1993) Individuality in the voice of the Emperor penguin, Aptenodytes forsteri: adaptation to a noisy environment. Ethology 94:279–290 Staicer CA, Spector DA, Horn AG (1996) The dawn chorus and other diel patterns in acoustic signaling. In: Kroodsma DE, Miller EH (eds) Ecology and evolution of acoustic communication in birds. Cornell University Press, Ithaca, N.Y. pp 426–453 Wiley RH, Richards DG (1978) Physical constraints on acoustic communication and the atmosphere: implications for the evolution of animal communication. Behav Ecol Sociobiol 3:69–94 Wiley RH, Richards DG (1982) Adaptations for acoustic communication in birds: sound transmission and signal detection. In: Kroodsma DE, Miller EH (eds) Acoustic communication in birds, vol 1. Academic Press, New York, pp 131–181