Voices in the dark: predation risk by owls influences dusk singing in a diurnal passerine
- First Online:
- Cite this article as:
- Schmidt, K.A. & Belinsky, K.L. Behav Ecol Sociobiol (2013) 67: 1837. doi:10.1007/s00265-013-1593-7
- 398 Views
Predation is an important cost of communication in animals and thus a potent selection pressure on the evolution of signaling behavior. Heterospecific eavesdropping by predators may increase the vulnerability of vocalizing prey, particularly during low light, such as at dusk when nocturnal predators are actively hunting. Despite the risk it entails, dawn and dusk chorusing is common in passerines. However, the dusk chorus has not been studied much, neglecting the opportunity for understanding how eavesdropping between predators and prey may shape communication in birds. Here, we report the first demonstration of simulated predation risk (playback of owl vocalizations) altering the dusk chorus of a diurnal passerine, the veery (Catharus fuscescens). Veeries have a pronounced dusk chorus, singing well after sunset and potentially exposing themselves to predation by owls. In response to brief playbacks of owl calls (~30 s of calls presented three times over 25 min), veeries sang fewer songs post-sunset and stopped singing earlier relative to control trials. These changes in singing remained evident 30 min after the last owl stimulus. Although the avian dusk chorus has received relatively little attention to date, our results suggest that the dusk chorus may pose a higher predation risk to singing males that may influence the evolution of singing behavior in diurnal birds.
KeywordsAnimal communicationDusk chorusHeterospecific eavesdroppingPredation riskVeery
Predation risk is an important cost of signaling in animals and is thus a potent selection pressure on the evolution of signaling behavior (Searcy and Nowicki 2005; Catchpole and Slater 2008). In birds, a cost of predation is often thought to accompany visual displays or structures, such as conspicuous coloration or long tails (Balmford et al. 1993; Andersson 1994). In addition, predators may be attracted to vocalizations used for mate attraction and pose a significant cost to singing (i.e., interceptive eavesdropping behavior; Peake 2005). These costs been observed in both avian (e.g., Krams 2001; Møller et al. 2008) and non-avian taxa (e.g., Ryan et al. 1981; Lang et al. 2005). A classic example of an acoustic mating signal that incurs a predation cost is the “chuck” call produced by túngara frogs (Physalaemus pustulosus). Ryan et al. (1981) demonstrated that the chuck call increases predation of frogs by fringe-lipped bats (Trachops cirrhosus), which eavesdrop on chorusing frogs.
In birds, song may also attract predators as shown by Krams (2001) and Mougeot and Bretagnolle (2000). Similarly, Møller et al. (2008) observed that singing birds have greater flight distances (i.e., nearest distance of approach by a predator) than non-singing birds, especially when singing from more exposed perches. These studies suggest that birds may be exposed to and perceived a higher risk of predation when singing. A cost of predation associated with singing activity in birds has been inferred in nocturnal species from two different orders: Procellariiformes and Camprimulgiformes. Mougeot and Bretagnolle (2000) demonstrated that the predatory brown skua (Catharacta antarctica) is attracted to nocturnal vocalizations of its prey species, the blue petrel (Halobaena caerulea), which in turn reduces singing rates in response to playback of skua calls. Nocturnal singing in poorwills (Phalaenoptilus nuttallii) in response to increased perceived risk is more complicated (Woods and Brigham 2008). Poorwills increase singing in response to conspecific songs, but the response is contingent on moon phase and the presence of great horned owls (Bubo virginianus).
Nocturnal vocalizations are not limited to nocturnal bird species. Indeed, 30 % of North American breeding birds, over 70 % of which are considered diurnal, vocalize at night (La 2011). However, less than 4 % of the avian vocalization literature is directed at nocturnal vocalization, and an even smaller percentage pertains to diurnal birds. The veery (Catharus fuscescens) has a pronounced dusk chorus (Belinsky et al. 2012), which has been demonstrated in other thrushes, e.g., Bicknell’s Thrush (Catharus bicknelli; Ball 2000), in closely related chats, including the common nightingale (Luscinia megarhynchos; Amrhein et al. 2004), and in more distantly related taxa such as wrens, Troglodytidae (e.g., Erne and Amrhein 2008). Veeries often sing intensely prior to sunset and continue for 40–50 min past sunset (see “Results” section). The presence of owls may make dusk singing especially costly to diurnal birds that lack the night vision of nocturnal petrels and caprimulgids (Nicol and Arnott 1974). Thus, predation risk may represent a largely uninvestigated cost of nocturnal singing in diurnal passerines (see Searcy and Nowicki 2008; Lima 2009), and dusk singing in birds may be a potentially important but largely neglected behavior (see Staicer et al. 1996; La 2011).
Here, we report the first demonstration of simulated predation risk (i.e., playback of owl vocalizations) altering the dusk chorus of a diurnal passerine, the veery. First, we hypothesized that veeries experience an increased cost of predation from nocturnal predators (although we do not exclude crepuscular activity of hawks) associated with singing in low light. If true, we predicted that veeries would decrease total song output in response to the perceived presence of owls, which we manipulated through an owl song playback experiment. Second, veeries exposed to owl playbacks would stop singing earlier in the evening (post-sunset) than veeries not exposed to owl playbacks.
Materials and methods
We studied a color-banded breeding population of veeries on the property of the Cary Institute of Ecosystem Studies, located in Dutchess County in southeastern New York, USA. The Cary Institute contains approximately 325 ha of eastern deciduous forest. Veeries are one of the dominant forest passerines at the site. Barred owls (Strix varia) are the primary nocturnal avian predators with three pairs (plus offspring) typically present each spring (KAS, unpublished data). Barred owls are known to kill wood thrush (Hylocichla mustelina) at our site (Schmidt et al. 2008a) and have been documented killing Catharus spp. by others (e.g., Streby et al. 2008). More generally, Livezey (2007) recorded avian prey in 13.5 % of barred owls pellets. Thus, barred owls are a true predation threat to veeries.
To test whether veeries would alter singing rates during the dusk chorus in response to simulated owl presence, we conducted paired 75-min playback/recording trials beginning at 25 min before sunset. These were conducted on the territories of known individuals and only during evenings with low wind speeds (<10 km/h) and no precipitation. We choose locations for setup based on observations of the locations of banded birds that were recorded singing no more than 1–2 days prior. In veeries, singing rates are significantly higher for males without active nests, and nest predation is high in this species (Schmidt and Ostfeld 2008). Hence, we observed an overall pattern of bouts of singing at a location, then near silence during nesting, and another bout of singing if and when a nest failed and territory borders potentially shifted. This pattern of singing between active nests also explains why a much smaller percentage of males sing at any one time during the breeding season. Based on hundreds of hours of focal recordings of veeries singing during the dawn, day, and dusk, we observed that each male sings in a specific location for roughly 5–7 days. In addition, recent analyses of radio-tracked individuals yielded an average territory size of 1.80 ha (KLB and KAS, unpublished data) for an average of 150-m spacing between the centers of abutting territories. Lastly, the recordings used for our analyses were based on passive recording with an omnidirectional microphone pointing upward (see below), and the resulting sonograms rapidly degrade with distance. For these reasons, we are confident that only a single individual was recorded across the two evenings at a location.
On a given evening, we conducted up to four concurrent trials spaced at least 300 m apart. Prior to the start of recording, we attached a Sennheiser ME 62 omnidirectional microphone (with K-6 power module) to a rebar post 1 m off the ground. The microphone was connected to a digital recorder (Marantz PD630 or Olympus LS 10; the same recorder was used for both trials for each bird) housed in a plastic box. We recorded at a sampling rate of 44.8 kHz and 24 bits with the gain set at a standardized level. We played all acoustic stimuli using Apple iPod shuffles connected to an amplified speaker (Saul Mineroff SME-AFS) and attached to a tree at ~2 m above ground. The speaker was positioned 25 m away from and facing the microphone. Playback amplitude was 92 dB SPL measured at 1 m in distance. This is lower than natural owl calls; however, sound distortion was evident at much higher volumes. Since we were targeting individual males on territories, we felt it conservative to reduce the playback volume, which would be sufficient to manipulate the perceived presence of owls near or on a given territory.
After equipment setup, we waited until 25 min before sunset to start the iPod and used a test tone to insure the system was operating. The first 10 min of the playback was silent to allow us time to quietly vacate the area before the first stimulus began. At 15 min before sunset, three barred owl “who cooks for you” calls were broadcast, followed by two calls at sunset, and one call and 10 s of caterwauling (also called raucous hoot; Mazur and James 2000) at 10 min post-sunset. All calls were separated by 20 s of silence, simulating natural calling rates. Not earlier than 60 min after sunset (last veery songs recorded were <54 min post-sunset), we returned to collect the equipment. Four different owl playbacks (i.e., exemplars) were used, each constructed of calls from four different commercially available recordings adjusted to the same amplitude using Raven Pro 1.4 software (Cornell Lab of Ornithology). Owls responded in some trials (0 of 13 silent, 1 of 13 frog, and 3 of 26 owl treatments). Two of these trials were dropped from analysis because the real owl calls continued for several minutes. The remaining trials with owls contained only short bursts similar to the playback stimuli and were retained. Sample sizes were too small to test for an owl exemplar effect. However, variation in song rate over time (see below) was greater within owl exemplars than among exemplars (range 0.5 < f ratios < 1.61; P > 0.50), dispelling concerns that results could be attributed to individual exemplars that we used. All exemplars were used in pilot experiments (in 2008 using more frequent playbacks) that induced counter-singing in barred owls. Thus, we are certain that all the owl stimuli accurately simulated the presence of owls.
Each bird had two trials: owl and control. The order of stimuli was randomly assigned, and we attempted to conduct the two treatments over two consecutive days unless weather interfered. We used two types of controls: silent (no iPod/speaker setups used) and gray tree frog stimulus. The sequence of calls in the frog stimulus was arranged identically to owl stimuli (i.e., three shorts bursts over a span of 25 min) and was made from previous recordings of tree frogs at the site. Trials were held from 21 May to 27 June 2009. The actual time of setup was standardized relative to sunset as described above. Paired trials were conducted on 24 individual birds; however, we eliminated six trials from individuals that sang <25 total songs (fewer songs than this, we did not consider chorusing behavior) during the recording period, leaving 18 birds (36 total trials) for analyses.
To quantify the veery’s singing behavior, three trained, unbiased observers scrolled through spectrographs of the dusk chorus recordings using Raven Pro 1.4. For each individual trial, observers quantified the number of songs per 3-min interval (i.e., song rate) through 54 min post-sunset (veeries did not sing later than this) and noted the time each song was sung. The same observer quantified songs for both of the paired treatments for each individual veery.
To quantify the decline in song rate as evening progressed, we divided the dusk period into eighteen 3-min intervals to track changes in song rates over time. In response to the owl stimulus, we predicted (1) song rates to be lower overall, relative to controls, (2) song rates to be lower in early dusk because of the recent owl stimulus, and (3) song rates to be lower in late dusk because visual acuity will be diminished and birds will be at greater risk as evening progressed. We performed two sets of analyses to test these predictions. In the first set, we included all songs per individual throughout the time period. In the second set of analyses, we quantified songs only during periods of extended singing bouts defined as a mean song rate of ≥18 songs during a 3-min interval (mean number of songs per 3-min interval was 24.91 ± 0.95 SE and did not differ between treatments; t16 = 0.56, P > 0.60). When birds are engaged in extended singing bouts, they are less able to perform vigilance behaviors and are sedentary at a singing post (KAS and KLB, personal observation). Both of these activities should be riskier when owls are perceived as present. While sedentary behavior can reduce predator encounters in some contexts, notably foraging (Werner and Anholt 1993), sedentary behavior while broadcasting one’s position and under reduced vigilance should not (also see Møller et al. 2008; Campos et al. 2009). For these reasons, we expected extended singing bouts to better reveal a cost of predation risk during the dusk chorus.
To test prediction 1, we compared the number of 3-min intervals in which song rates (all songs and extended singing bouts in separate analyses) were higher, on average, in control vs. owl treatments. Because an individual’s song rate in a 3-min interval may not be independent of the past interval, we used a bootstrapping procedure (Matlab® v. R2012b) to randomize song rates per 3-min intervals among all individuals independent of treatment while preserving the time post-sunset. This procedure breaks any potential correlation in an individual’s song rate across intervals as well as being blind to treatment. We repeated this procedure 1,000 times to produce a null distribution against which we tested our hypotheses. The P value was calculated as the number of randomizations in which the observed outcome (i.e., number of intervals or more in which control > owl) occurred in 1,000 replicates.
To take advantage of the paired design, we used paired t tests to compare the total number of songs (prediction 1), summed number of songs in the first and fourth 3-min intervals, and time of last song (prediction 3), across treatments. The first and fourth intervals immediately recorded singing following the timing of the last two stimulus calls and served a proxy for motivation to sing after owl vocalizations (prediction 2). The last song (time is negatively correlated with light levels and presumed lower likelihood of spotting a predator) was easily determined because birds abruptly ceased singing in the evening. We performed similar analyses using paired t tests for extended singing bouts including the total time in extended singing bouts, total number of songs in bouts, and the timing of the first (prediction 3) and last bouts. All tests were two tailed.
To examine whether there were any significant differences between the two types of controls, we performed a complementary set of comparisons between frog and silent controls similar to those described above, although it was necessary to replace paired t tests with non-paired t tests. There were no significant differences between the two types of controls in any of the analyses described below; therefore, control trials were subsequently pooled for the analyses. Specifically, frog (N = 10) and silent (N = 8) control treatments did not differ in which had the greater number of songs or singing bouts per 3-min intervals (frog = 11, silent = 7, P > 0.40; extended singing bouts: frog > silent in 10 of 16 trials, P > 0.40), nor in the mean number of songs throughout the trial (frog = 181.3, silent = 170.1, t16 = 0.80, P > 0.40). All other analysis were likewise non-significant (number of songs after stimulus: frog = 29.6, control = 27.2, t16 = 0.26, P = 0.80; time of last song: frog = 38.4, silent = 36.0; t16 = 0.64, P > 0.50; time of first bout: frog = 3.0, silent = 1.5, t13 = 0.72, P = 0.40 time of last bout: frog = 33.75, silent = 27; t13 = 1.27, P > 0.20). These analyses argue against an interpretation that any auditory stimulus would have the effect of decreasing song rates in veeries.
Extended singing bouts
Veeries had a greater number of songs in extended singing bouts in control trials (bootstrap analysis: P = 0.005; Fig. 1b, c). This was due to a greater number of bouts per interval in control trials (control 5.41 ± 0.84; owl 3.65 ± 0.63; bootstrap analysis: P = 0.005) since, within bouts, birds sang at similar rates regardless of treatment (see analyses). In pairwise comparisons of the 14 individuals with extended singing bouts, individuals sang a greater length of time in bouts in control trials (18.0 ± 2.28 vs. 10.9 ± 1.89 min; t13 = 2.97, P = 0.01; D of Fig. 2), leading to more total songs in extended singing bouts in control trials (150.71 ± 21.37 vs. 97.71 ± 19.31; t13 = 2.40, P = 0.028). Extended singing bouts occurred significantly earlier in control trials (interval 2.35 ± 0.54 vs. 4.57 ± 0.94; t13 = 2.45, P = 0.028; E of Fig. 2) but did not extend significantly later (interval 10.28 ± 0.89 vs. 8.57 ± 0.75; t13 = 1.15, P = 0.26; F of Fig. 2).
Heterospecific eavesdropping is common in predator–prey systems (Ryan et al. 1981; Mougeot and Bretagnolle 2000; Rainey et al. 2004; Deecke et al. 2005; Randler 2006; Laumann et al. 2007; Schmidt et al. 2008b; Emmering and Schmidt 2011), but in passerines, it has received less attention in the context of singing (i.e., intra- or intersexual context) than in a predatory context (i.e., alarm calling). By playing owl vocalizations to veeries near the time of sunset, we tested the hypotheses that veeries (1) eavesdrop on vocalizing owls and (2) perceive a cost of predation associated with singing in the low light of dusk when owls are perceived to be present, resulting in reduced song output and terminating dusk singing earlier.
In accordance with our predictions, veeries significantly decreased song rates at dusk (i.e., after sunset) in owl playback treatments compared to controls. This pattern was evident both at the start of the experimental trials and proximity (in time) to the owl stimuli as well as late in the experiment when light levels were lowest. Song reduction was also more pronounced when we restricted our analyses to extended singing bouts. Birds sang longer in extended bouts and bouts occurred earlier during control trials. In contrast, birds exposed to the owl stimulus reduced the amount of time in extended singing and avoided singing bouts early into the trial. More pronounced song reduction in the analysis of singing bouts is not surprising. During extended singing, birds are generally more exposed (Catchpole and Slater 2008; KAS and KLB, personal observations of veeries) and are less vigilant (song length itself is 2–3 s during which bird are not actively vigilant), and song production can potentially attract a predator to the fixed location of the singing bird (Mougeot and Bretagnolle 2000; Krams 2001).
At our field site, barred owls call in both short bursts and over extended periods much longer than our stimuli (KAS, personal observation). Since one call is sufficient to indicate an owl’s presence, we designed, a priori, our experiment to limit the length of the stimuli to three short bouts of 30 s each over a 25-min period. It is compelling that 30 min after the last owl stimulus, the effects of exposure to the stimulus were still evident in the early cessation of singing in veeries, relative to control trials. Thus, our study demonstrates that veeries eavesdrop on the communication of barred owls and alter their dusk singing behavior in the perceived presence of owls. This is the first example we know of to demonstrate that dusk singing in a diurnal passerine is affected by perceived predation risk.
The response of veeries to an increase in perceived predation risk is similar to that observed in nocturnal petrels, which also reduced song rates in response to experimental playbacks of predator calls (Mougeot and Bretagnolle 2000). Comparing veeries to common poorwills is less straightforward since the latter’s response is predicated on moon phase and conspecific singing (Woods and Brigham 2008). Unfortunately, interference due to weather prevented us from conducting trials during the second quarter when the moon is both relatively bright and high enough in the sky at sunset to affect forest brightness. Thus, we were unable to examine this variable in our analyses.
It is well known that many diurnal birds have a second peak of singing activity in the evening (Staicer et al. 1996; La 2011). Although the dusk chorus is not simply the dawn chorus in reverse, work on both the dawn and dusk choruses may complement one another. For example, Thomas et al. (2002) noted dawn singing initiated at lower light intensity (and hence earlier) in songbirds with larger exposed eyes and thus higher visual capacity in low light (also see Berg et al. 2006). Since rapid flight, necessary for escape, also requires high spatial resolution and aerial predators preferably hunt at dawn and dusk (Roth and Lima 2007), diurnal birds should begin singing later at dawn and cease singing earlier at dusk when they are under higher perceived predation risk. Interestingly, passerines may cease dusk chorusing at higher relative light levels than initiate singing at dawn (Leopold and Eynon 1961). Unfortunately, such comparisons are rare. In his review, Lima (2009) posited that the well-established dawn chorus in birds might be sensitive to the risk of predation because of, in part, low light levels. He then writes that he knew “no studies that have addressed predation on dawn-singing birds”. This gap continues; however, our study now demonstrates that predation risk can affect dusk singing in birds. Still, the study of the avian dusk chorus has been largely ignored relative to the dawn chorus (La 2011); the latter of which has a sizable literature (see Staicer et al. 1996; Catchpole and Slater 2008). We hope our study can motivate others to diminish this bias.
Lastly, it is interesting to conjecture whether, if predation risk imposes a fitness cost (i.e., handicap; Zahavi 1975) to singing in low light, dawn and dusk singing may be a reliable indicator of male quality, perhaps more so than singing at other times of day (see Otter et al. 1997; Poesel et al. 2006). If so, sexual selection may drive dawn and dusk singing in some passerines. However, we acknowledge that many of the hypotheses put forth to explain the dawn chorus (e.g., sound transmission, acoustic masking, advertise resource availability, mate-guarding; see Staicer et al. 1996; La 2011) may also operate at dusk, or alternatively, dawn and dusk choruses may have independent functions. The types of ecological, social, and sexual factors that ultimately explain the dusk chorus fall outside the purview of our study. Nonetheless, we believe that a systematic study of the dusk chorus in birds, both diurnal and nocturnal, and other taxa warrants more attention. Further studies of dusk chorus singing may reveal a profound influence of predation risk on the evolution of singing behavior.
The authors thank Stacy Tekstar, Angela Olsen, and Claire Randall for patiently and meticulously identifying all of the veery songs throughout these long recordings, and Janice Kelly for additional support in the field. Two anonymous reviewers provided constructive comments to an earlier draft of the manuscript. This research was supported in part by a grant to KAS from the National Science Foundation (DEB 0746985).
Our study was conducted in compliance with the ethical standards of animal care and use at Texas Tech University and the United States of America.