Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Downloaded from http://rspb.royalsocietypublishing.org/ on June 18, 2017 rspb.royalsocietypublishing.org Crying wolf to a predator: deceptive vocal mimicry by a bird protecting young Branislav Igic1,†, Jessica McLachlan1,2, Inkeri Lehtinen1,3 and Robert D. Magrath1 Research Cite this article: Igic B, McLachlan J, Lehtinen I, Magrath RD. 2015 Crying wolf to a predator: deceptive vocal mimicry by a bird protecting young. Proc. R. Soc. B 282: 20150798. http://dx.doi.org/10.1098/rspb.2015.0798 Received: 8 April 2015 Accepted: 11 May 2015 Subject Areas: behaviour, ecology Keywords: alarm call, anti-predator, deception, interspecific eavesdropping, mimicry, nest defence Author for correspondence: Branislav Igic e-mail: [email protected] † Present address: Department of Biology, University of Akron, Akron, OH 44325 – 3908, USA. Electronic supplementary material is available at http://dx.doi.org/10.1098/rspb.2015.0798 or via http://rspb.royalsocietypublishing.org. 1 Research School of Biology, The Australian National University, Canberra, Australian Capital Territory, Australia Department of Zoology, University of Cambridge, Cambridge, UK 3 Department of Environmental Sciences, University of Helsinki, Helsinki, Finland 2 Animals often mimic dangerous or toxic species to deter predators; however, mimicry of such species may not always be possible and mimicry of benign species seems unlikely to confer anti-predator benefits. We reveal a system in which a bird mimics the alarm calls of harmless species to fool a predator 40 times its size and protect its offspring against attack. Our experiments revealed that brown thornbills (Acanthiza pusilla) mimic a chorus of other species’ aerial alarm calls, a cue of an Accipiter hawk in flight, when predators attack their nest. The absence of any flying predators in this context implies that these alarms convey deceptive information about the type of danger present. Experiments on the primary nest predators of thornbills, pied currawongs (Strepera graculina), revealed that the predators treat these alarms as if they themselves are threatened by flying hawks, either by scanning the sky for danger or fleeing, confirming a deceptive function. In turn, these distractions delay attack and provide thornbill nestlings with an opportunity to escape. This sophisticated defence strategy exploits the complex web of interactions among multiple species across several trophic levels, and in particular exploits a predator’s ability to eavesdrop on and respond appropriately to heterospecific alarm calls. Our findings demonstrate that prey can fool predators by deceptively mimicking alarm calls of harmless species, suggesting that defensive mimicry could be more widespread because of indirect effects on predators within a web of eavesdropping. 1. Introduction Prey animals often use dishonest signals to fool predators and escape predation. For example, prey species can make themselves appear larger or displeasing, and thus deter predators from attacking [1,2]. Mimicry is another classic example of a dishonest signal and is often used by otherwise undefended organisms to deceive predators into misidentifying them as toxic or dangerous species (defensive mimicry [3]). This is generally accomplished by mimicking attributes of dangerous models, including predators themselves, or the aposematic coloration or warning sounds of noxious prey (Batesian mimicry [1,4–6]). However, mimicking dangerous species is not always possible, and it is unknown whether mimicking harmless species can provide protection against predators. Although the protective benefits of traditional examples of mimicry are well known, such as for Batesian and Müllerian mimicry [7,8], it is still unclear if avian vocal mimicry has an anti-predator function [9,10]. Mimicry of other species’ vocalizations or sounds is widespread among songbirds, but the function of these mimetic vocalizations is known for only a handful of species (reviews: [9,10]). In response to predators, birds often mimic sounds that are associated with danger, such as calls of predators or alarm and aggression calls of heterospecifics, suggesting that vocal mimicry confers anti-predator protection by similar mechanisms as Batesian mimicry [11–14]. Indeed, mimicry of rattlesnake (Crotalus viridis) rattling sounds by burrowing owls (Athene cunicularia) has been shown to deter heterospecific competitors from entering owl burrows and thereby potentially evicting owls [4]. Yet surprisingly, how predators respond to mimetic & 2015 The Author(s) Published by the Royal Society. All rights reserved. Downloaded from http://rspb.royalsocietypublishing.org/ on June 18, 2017 (a) brown goshawk (top predator) (b) 16 2 rspb.royalsocietypublishing.org thornbill alarm 8 honeyeater alarm 8 Proc. R. Soc. B 282: 20150798 frequency (kHz) pied currawong (nest predator) 0 16 0 16 mimicry of honeyeater alarm brown thornbill (mimic) New Holland honeyeater (mimicked species) 8 0 0 0.3 time (s) 0.6 Figure 1. Predator –prey relationships between species, and aerial alarm vocalisations used in experiments. (a) From top to bottom: brown goshawk (top predator), pied currawong (nest predator), brown thornbill (mimic), New Holland honeyeater (harmless species mimicked); arrows indicate direction of predator– prey relationship. (b) Spectrograms of brown thornbill non-mimetic aerial alarms, New Holland honeyeater aerial alarms and corresponding mimicry by brown thornbills. Photo credits: brown goshawk, Geoffrey Dabb; rest, Steve Igic. (Online version in colour.) vocalizations of prey has never been tested. As a result, avian vocal mimicry is often placed outside the evolutionary framework of traditional mimicry systems (for discussions, see [10,15]). Alarm calls can be used to deceive competitors, and such deception can entail non-mimetic or mimetic calls [16–18]. Several bird species produce alarm calls in the absence of danger to distract competitors and steal their food (e.g. Dicrurus adsimilis [18]), or use deceptive alarms as a mate-guarding strategy (e.g. Hirundo rustica [19]). This deception is possible between species because individuals often eavesdrop on and respond to alarm calls of other species [20]. There are two main types of avian alarm calls: (i) aerial alarm calls are produced to predators in flight and provoke listeners to scan the sky or fly into cover, and (ii) mobbing alarm calls are produced to terrestrial predators or stationary avian predators, and provoke listeners to inspect and harass the predator [21]. Aerial alarm calls are particularly effective signals for startling others because predators in flight are extremely dangerous and require listeners to respond immediately [21]. However, just like the story of the boy who cried wolf, the effectiveness of deception can decline when deceptive alarms are used too frequently [17,22,23]. Callers can potentially avoid resistance to deception by mimicking the alarm calls of other species, thereby reducing the frequency of any one type of deceptive call [17]. Although the use of non-mimetic and mimetic alarms to deceive competitors is well documented, it is unclear whether alarms of any type are used to deceive predators. For most bird species studied, human disturbance or predator attacks on a nest provoke parents to produce mobbing alarm calls, which is appropriate when predators are not in flight [21]. However, brown thornbills (Acanthiza pusilla; figure 1a) mimic a chorus of other species’ aerial alarm calls when humans disturb their nests [14]. As no flying predator is present, the production of aerial alarm calls by thornbills during nest defence conveys inaccurate information about danger and may serve to deceive predators into responding as if they themselves are in immediate danger from above [14,24]. Mimicking a chorus of aerial alarm calls may also simulate multiple callers, a reliable indication of immediate danger [20,25–27], which may increase the effectiveness of deception. Here, we experimentally tested the anti-predator function of mimetic aerial alarm calls used by brown thornbills during nest defence. We examined the interactions between thornbills and their primary nest predators, pied currawongs (Strepera graculina; figure 1a), and tested the hypotheses that (i) thornbills use aerial alarm calls to deceive currawongs that attack their nest, and that (ii) mimicking a chorus of alarm calls is better at deceiving currawongs than using a single, nonmimetic type of alarm call. Currawongs are susceptible to being deceived by aerial alarm calls because they themselves are prey to Accipiter hawks, such as A. fasciatus (figure 1a; [28]), which are dangerous top predators that provoke aerial alarm calls from thornbills and other sympatric species. 2. Material and methods (a) Study site, species and experimental overview We studied free-living colour-banded thornbills and unbanded pied currawongs in and around the Australian National Botanic Gardens, Canberra, Australia (358160 S, 149860 E). Both species are highly territorial, pair-breeding passerines common throughout southeastern Australia [29,30]. Brown thornbills are 7 g and build dome-shaped nests typically 0.5–1 m above the ground and in dense vegetation [29]. Currawongs are 280 g specialized nest predators that can capture up to 2 kg of young birds to raise a single brood to maturity [31]. An estimated 52% of brown thornbill nests with nestlings suffer mortality at our study site, with 61–83% of these losses attributed to nest Downloaded from http://rspb.royalsocietypublishing.org/ on June 18, 2017 To test when thornbills use mimetic aerial alarms, we simulated three levels of threat to 8 – 10-day-old nestlings of breeding thornbill pairs in 2011, approximately 6 days before normal fledging age. We presented 10 thornbill pairs with taxidermic mounts 1 m in front of their nests in combination with broadcasting sounds to simulate three levels of threat: (i) no threat control—a crimson rosella (Platycercus elegans, a harmless sympatric parrot) near the nest in combination with rosella contact calls, (ii) potential threat from a searching nest predator—a pied currawong near the nest in combination with rosella contact calls, and (iii) immediate threat from a predator attacking the nest—a pied currawong near the nest in combination with nestling distress calls, which are produced by nestlings only when disturbed (electronic supplementary material, figure S1). All three treatments were presented to a pair in random order during a single day. Additional details of experimental protocol are provided in the electronic supplementary material. To test the specificity of responses in relation to nestling distress calls, each sound was broadcast only once for 4 s at the start of a 5 min period when the model was presented. Sounds were broadcast at the average amplitude of nesting distress calls (mean ¼ 49.2 dBA from 4 m; s.e.m. ¼ 0.63; n ¼ 10). We recorded vocalizations produced by parents following protocols described in Igic & Magrath [14]. For ethical reasons, we conducted experiments before nestlings were old enough to prematurely fledge as that may reduce their survival. We examined production of non-mimetic and mimetic alarm calls by parents: (i) over the full 5 min threat treatment, and (ii) specifically during the 4 s period that nestling distress calls were broadcast, simulating a period of nest attack. We inspected spectrograms using RAVEN PRO v. 1.4 [34] and identified the calls produced by parents using protocols described in Igic & Magrath [14]. We categorized parental calls into five types based on their use by thornbills and heterospecifics outside the context of nest defence: non-alarm, non-mimetic mobbing alarm, non-mimetic aerial alarm, mimetic mobbing alarm and mimetic aerial alarm [14]. Summaries of the thornbill’s mimetic repertoire are presented in Igic & Magrath [14,24]. Thornbills rarely mimic predator calls [14], and never did so in this study. Both parents were present during our experiments and we considered their combined call production in analyses. Mimicry of aerial alarms when a currawong is nearby but young are undisturbed would suggest a function in hindering predators from finding the nest, whereas mimicry during the simulated attack would suggest a function in helping young escape. (c) Currawong response to thornbill alarms at simulated nests Next, we tested how currawongs themselves respond to thornbill non-mimetic and mimetic aerial alarm calls. We broadcast three sets of 4 s thornbill vocalizations to 18 currawongs during their breeding season in 2011: (i) non-mimetic aerial alarms alone, (ii) a combination of non-mimetic and mimetic aerial alarms, and (d) Currawong response to thornbill and real honeyeater alarms We conducted a follow-up experiment in 2013 to examine how currawongs respond to New Holland honeyeater aerial alarms and test whether these alarms are more salient to them than thornbill alarms. Currawongs might respond more strongly to a mixture of nonmimetic and mimetic alarms than non-mimetic alarms alone because (i) the mimetic alarm itself is more salient or (ii) the mixture of different species’ alarms is more salient than alarms of a single species. We therefore broadcast four types of calls alone over 0.8 s to currawongs feeding in the open: (i) thornbill song as a control, (ii) thornbill non-mimetic aerial alarms, (iii) thornbill mimetic honeyeater aerial alarms, and (iv) the honeyeater’s own aerial alarms (figure 1b). We used the same currawong population as in the previous experiment, but only 10 pairs produced nests during 2013 and could be used for this experiment. We used a modified protocol compared with the previous experiment. We placed food on the ground, rather than in a feeder, to examine the salience of alarms as opposed to their effectiveness in nest defence. Recordings were broadcast using the same equipment and methods, but the equipment was mounted around the waist of an observer standing 10 m away from the focal bird. We again video recorded currawong responses and 3 Proc. R. Soc. B 282: 20150798 (b) Thornbill response to nest threat (iii) thornbill song as a control. We used a mixture of non-mimetic and mimetic alarms, rather than mimetic alarms alone, because thornbills generally use mimetic alarms in combination with nonmimetic alarms [14]. For mimetic alarms, we used mimicry of New Holland honeyeater (Phylidonyris novaehollandiae) aerial alarm calls, the call most frequently mimicked during nest defence [14]. We predicted that the aerial alarms would distract currawongs and that mimicry would prolong this distraction. Under natural circumstances, a delay would allow older thornbill nestlings to scramble out of the nest and hide in surrounding vegetation [29]. We trained currawongs to feed from custom-made feeders that simulated a thornbill nest environment (electronic supplementary material, figure S2), where we broadcast the recordings of thornbill vocalizations. Full details of training protocol and creation of playbacks are provided in the electronic supplementary material. Once currawongs were habituated to feeders and our experimental equipment, we broadcast each recording on a separate date, in random order, and in the absence of species aggressive towards currawongs or any aerial alarm calls nearby in the previous 10 min. Upon setting up the feeder and equipment, we allowed currawongs to make two visits to the feeder before broadcasting a recording the moment a currawong was about to pick up food on its next visit. All recordings were broadcast from a Response Dome Tweeter speaker mounted on a tripod 0.7 m from the ground and 5 m from the feeder. We played sounds from an Edirol R-09HR solid-state digital player and custom-built amplifier at an amplitude of 60 dBA at 5 m (average amplitude of thornbill aerial alarms recorded during nestling banding [14]), and we video recorded currawong behaviour. We examined how currawongs responded to playbacks by analysing video recordings using MICROSOFT WINDOWS MOVIE MAKER v. 2.6 (Microsoft Corporation). We classified the immediate behaviour of currawongs following sounds into three categories that reflected the intensity of their alarm response: flight from the feeder’s location, aerial scanning (defined as looking up for more than 1 s), or other response. Currawong behaviour was classified into these categories by a scorer blind to the playback treatment: the video recordings were in random order, the sound was muted and the timing of playbacks was marked visually. The intensity of an individual’s alarm response can also be measured as the delay until their behaviour returns to normal (e.g. [35,36]), so we also measured how long currawongs delayed collecting food after playbacks, again using blind scoring (to the nearest 0.1 s). rspb.royalsocietypublishing.org predation by avian predators, primarily pied currawongs [32]. There are no data on the rate of predation by hawks for either species; however, the annual rate of survival for brown thornbills and pied currawongs is, respectively, 63% and 82% [32,33]. To test our hypotheses, we first tested experimentally whether thornbills mimic aerial alarm calls in response to currawongs on the ground during nest defence, and if so whether these are likely to function in preventing attack or helping young escape when attacked. We then tested whether currawongs are deceived by thornbill alarm calls and mimicry at simulated thornbill nests. Finally, we examined the response by currawongs to thornbill, heterospecific, and mimetic alarms to test their salience when broadcast alone. Downloaded from http://rspb.royalsocietypublishing.org/ on June 18, 2017 (a) (b) 20 (c) 25 20 10 5 0 15 10 150 100 50 5 0 0 0 30 60 90 120 150 180 210 240 270 300 time (s) no threat potential threat immediate threat Figure 2. Alarm calling by thornbills during nest threats. (a) Alarm calls (fitted means + s.e.m.) produced by thornbills (n ¼ 10) specifically during the 4 s in which nestling distress calls were broadcast: simulated attack on the nest. (b) Alarm calls (mean + s.e.m.) produced by thornbills (n ¼ 10) at every 15 s interval following a simulated attack on the nest by a predator. Arrow indicates the period when nestling distress calls were broadcast (simulated attack on the nest). (c) Calls (mean + s.e.m.) produced by thornbills (n ¼ 10) over 5 min in response to different levels of threat: no threat; potential threat—a predator near the nest; and immediate threat—following a simulated attack on the nest by a predator. Analyses are shown in the electronic supplementary material, tables S1–S3. (Online version in colour.) used blind scoring to classify the initial response of currawongs to playbacks and the amount of time they spent aerial scanning following alarm calls (to the 0.1 s). (e) Statistical analyses We used statistical models to compare the number and types of alarm calls produced by parents: (i) in response to the three levels of threat over the full 5 min, and (ii) specifically during 4 s of nestling distress calls. Analyses compared: (i) the long-term responses of parents to degrees of nest threat, and (ii) the specific response of parents during nest attack. For the long-term response (i), we constructed a general linear model with the number of alarm calls produced by parents over 5 min as a response; pair ID, treatment order, threat type, alarm call type, and the interaction between threat type and alarm call type as covariates. To improve fit we transformed the response (log10 þ 1). The interaction between alarm call type and threat type was significant (electronic supplementary material, table S1) indicating that the four types of alarm calls were used differently across threat treatments. Therefore, we constructed separate linear models to compare production of alarm calls during potential and immediate threats. We used the log number of alarm calls as a response; pair ID and alarm call type as covariates (electronic supplementary material, table S2). For the short-term response to nest attack (ii), we used a zero-inflated negative binomial (ZINB) model to test if mimetic aerial alarm calls were used more frequently than other alarms during the 4 s of nestling distress calls (electronic supplementary material, table S3). Parents did not produce any calls during 4 s of rosella calls, and therefore this treatment could not be included in the analysis. We constructed this model with the number of alarm calls produced during the 4 s as a response; alarm call type and pair ID as fixed effects. A ZINB model fitted our data better than a negative binomial model (Vuong closeness test: V ¼ 23.82, p , 0.001; [37]). We compared currawong responses to playbacks using McNemar’s tests and linear mixed models. We used McNemar’s tests to compare categorical responses of currawongs to playbacks (fleeing, aerial scanning or other non-aerial-specific response). We constructed linear mixed models to compare the latency for currawongs to resume feeding after playbacks (electronic supplementary material, table S4) or time spent scanning (electronic supplementary material, table S5). Currawongs never scanned the sky following song in the second experiment (electronic supplementary material, figure S3), so this control treatment was not included in subsequent analysis. We used the time currawongs delayed feeding or spent scanning following a playback as responses, recording type and playback order as fixed effects, and currawong identities as random effects. To alleviate problems with unequal variances, we square-root transformed the latency for currawongs to resume feeding. All statistical analyses were conducted in R [38]. We constructed linear mixed models using identity link functions, REML and the lme() function of the nlme package [39]. Our ZINB model was constructed using the zeroinfl() function and we conducted Vuong closeness tests using the vuong() function, both from the pscl package [40]. For each model, we visually checked violations of normality and equality of variance using q–q normal plots and residuals plots. We used ANOVA and likelihood ratio tests to identify significant effects, and conducted pairwise post hoc tests and Tukey contrasts using the glht() function of the multcomp package [41]. McNemar’s tests were conducted using the mcnemar.test() function [38]. 3. Results In support of a function in helping young escape when attacked, thornbills mimicked heterospecific aerial alarm calls specifically during simulated nest attack, and not when a predator model was merely nearby. Mimetic aerial alarms were used despite the absence of a flying predator and were the most common type of call used by parents during the playback of nestling distress calls (ZINB effect: x26 ¼ 20:85, p , 0.01; all post hoc comparisons, p , 0.05; figure 2a; electronic supplementary material, table S3). Mimicry of aerial alarms coincided specifically with the 4 s that nestling distress calls were broadcast (figure 2b), beginning on average 1.9 s (+1.1 s.e.m.) after the onset of nestling distress calls and continuing for an average 5.9 s (+3.2 s.e.m.). During this time, parents mimicked between one and four different species’ aerial alarms (average 1.9 + 0.4 s.e.m.). After the simulated attack had finished or when a currawong was merely present nearby, parents produced non-mimetic and mimetic mobbing alarms, such that over the full 5 min mimetic aerial alarm calls were used less often than non-mimetic (potential threat: t27 ¼ 10.28, p , 0.001; immediate threat: t27 ¼ 4.44, p , 0.001) and mimetic ( potential threat: t27 ¼ 7.43, p , 0.001; immediate threat: t27 ¼ 2.82, p ¼ 0.04) mobbing alarms (figure 2b,c; electronic supplementary material, tables S1 and S2). Proc. R. Soc. B 282: 20150798 non-mimetic mimetic non-mimetic mimetic mobbing mobbing aerial aerial 200 no. vocalizations no. alarm calls no. alarm calls 15 4 non-alarm non-mimetic mobbing mimetic mobbing non-mimetic aerial mimetic aerial rspb.royalsocietypublishing.org non-mimetic mobbing mimetic mobbing non-mimetic aerial mimetic aerial Downloaded from http://rspb.royalsocietypublishing.org/ on June 18, 2017 (a) (b) 20 15 10 5 time spent scanning (s) latency to resume feeding (s) aerial scan flight other response 8 aerial scan flight rspb.royalsocietypublishing.org 25 5 10 30 6 4 2 0 0 non-mimetic non-mimetic and alarm mimetic alarms non-mimetic alarm mimetic alarm honeyeater alarm Figure 3. Currawong response to thornbill and honeyeater vocalizations. (a) Mean (+s.e.m.) number of seconds that currawongs (n ¼ 18) delayed feeding at simulated nests after playback of thornbill vocalizations. Analyses are shown in the electronic supplementary material, table S4. (b) Mean (+s.e.m.) number of seconds that currawongs (n ¼ 10) feeding in the open spent scanning after playback of thornbill or honeyeater vocalizations; currawongs did not scan the sky following thornbill song, so this treatment was not included here (electronic supplementary material, figure S3). Pie-charts above bars illustrate the proportion of currawongs that flew away from the location (black), scanned the sky for more than 1 s (grey) or did not respond with a typical aerial alarm response (white). Analyses are shown in the electronic supplementary material, table S5. In support of a deceptive anti-predator function, currawongs treated all aerial alarms as if they were themselves threatened by a predator in flight, and the specific benefit of incorporating mimicry was prolonging this distraction (figure 3a; electronic supplementary material, video S1). Currawongs were more likely to scan the sky or flee following either alarm playback than following song, suggesting that they were deceived by thornbill alarms (McNemar’s test versus song: non-mimetic alone: x21 ¼ 15:1, p , 0.001; mixture: x21 ¼ 12:5, p , 0.001; figure 3a). As a result of distraction, currawongs delayed feeding 8.3 s (+0.4 s.e.m.) longer following non-mimetic aerial alarms alone than following song (t32 ¼ 5.11, p , 0.0001), and prolonged this delay a further 8 s (+0.9 s.e.m.) if non-mimetic and mimetic aerial alarm calls were broadcast together (t32 ¼ 2.47, p ¼ 0.02; figure 3a; electronic supplementary material, table S4). We found no evidence that thornbill aerial alarms were less salient to currawongs than aerial alarms of the most frequently mimicked heterospecific. When foraging in the open, currawongs scanned the sky for danger following honeyeater alarms, thornbill non-mimetic alarms and mimetic honeyeater alarms (figure 3b). There was no difference in the duration of scanning across the alarm types (linear mixed model effect: F2,13 ¼ 1.91, all p ¼ 0.2; figure 3b; electronic supplementary material, table S5). Two out of 10 currawongs fled to mimetic alarms, but this was not statistically different compared with real honeyeater alarms (McNemar’s test: x21 ¼ 0:5, p ¼ 0.5; figure 3b). 4. Discussion Taken together, our findings demonstrate that brown thornbills deceive a major nest predator by mimicking alarm calls of other harmless species. This sophisticated behavioural strategy exploits the interactions among multiple species across several trophic levels (figure 1a): thornbills exploit currawong eavesdropping on the salient alarm calls of heterospecifics, signalling about top predators that are dangerous to all three species, in order to deceive currawongs into delaying attack on their nestlings. Given the 40-fold size difference between thornbills and currawongs, physical attacks by thornbills are likely to be both ineffective and risky, and deception is perhaps the only effective defence once a nest is attacked [42]. Although used in the presence of an avian predator, the aerial alarm calls used by thornbills during nest defence were deceptive. The alarms were not simply communicating about danger to mates (e.g. [43]) because they were produced in the absence of a flying threat—the currawongs were always stationary on the ground. Indeed, in other circumstances, thornbills produce aerial alarm calls to avian predators in flight and mobbing alarm calls to stationary avian predators, including to currawongs [14]. They were also not merely communicating the presence of danger to offspring [44] because the alarms were only used once nestlings were already under attack. Lastly, these aerial alarms are not used in aggressive contexts by thornbills or the mimicked heterospecifics, and thereby do not signal aggression [13,45]. Deception is a likely explanation whenever an alarm signal conveys inaccurate information about the type, or presence, of danger and in doing so provokes a response by the predator that benefits the caller but is costly to the predator [2]. Currawongs indeed treated the alarms as if they were themselves in danger from above (e.g. [18]). Although predators can pay attention to prey alarms [46,47] and use them to locate feeding opportunities [48], the use of alarm calls by prey to fool predators has not been shown previously. This is therefore a novel diversionary display employed by birds during nest defence, similar to injury feigning and false brooding [49]. A surprising feature of thornbill mimicry is that they mimicked harmless species rather than dangerous or toxic ones [4,13]. Unlike many taxa that mimic related or similarly sized noxious prey, such as the abundant examples of Batesian and Müllerian mimicry among invertebrates, reptiles and fishes [1,50], the opportunity for such mimicry is rare for birds. We know of only three cases. The New Guinean Pitohui and Ifrita genera are toxic, resulting in one case of Müllerian, and potentially also Batesian, mimicry within Pitohui [51,52]. The nestlings of Laniocera hypopyrra are also considered to be Batesian mimics of a toxic caterpillar [53]. Finally, burrowing owls mimic rattlesnake sounds to deter competitors [4]. The absence of noxious models, or difficulties in accurately mimicking vocalizations of much larger predators, might constrain thornbills into mimicking the alarm calls of harmless species, which are cues of dangerous flying predators. Aerial alarms might also signal a more urgent threat than predator calls because predators are usually Proc. R. Soc. B 282: 20150798 song Downloaded from http://rspb.royalsocietypublishing.org/ on June 18, 2017 Ethics. This work was carried out with permission from the Australian Bird and Bat Banding Scheme, Environment ACT, Australian National Botanic Gardens and Australian National University Ethics Committee (B.EEG.06.10). Data accessibility. Data available from the Dryad Digital Repository: http://dx.doi.org/10.5061/dryad.fh40b. Authors’ contributions. B.I. and R.M. conceived the study; B.I., J.M. and I.L. conducted the study; and all authors contributed to writing. Competing interests. We have no competing interests. Funding. This work was supported by an Australian Postgraduate Award to B.I., the Australian National University, and an Australian Research Council Discovery Grant (DP0665481) to R.D.M. Acknowledgements. We thank the Australian National Botanic Gardens for allowing us access to do our research; Andrew Cockburn, David Green, Tonya Haff, Helen Osmond, Nicolas and Elodie Magraf, and Martijn van de Pol for assistance in the field; the Magrath and Shawkey labs, the EEG reading group, Loeske Kruuk, Andrew Cockburn, Nick Davies and two anonymous reviewers for comments on the manuscript; Eben Goodale, David Green, Naomi Langmore, Amanda Ridley, Graeme Ruxton and many other colleagues for helpful comments and discussions. References 1. 2. 3. Ruxton GD, Sherratt TN, Speed MP. 2004 Avoiding attack: the evolutionary ecology of crypsis, warning signals, and mimicry. New York, NY: Oxford University Press. Caro T. 2014 Antipredator deception in terrestrial vertebrates. Curr. Zool. 60, 16– 25. Bates HW. 1862 Contributions to an insect fauna of the Amazon Valley (Lepidoptera, Heliconidae). Trans. Linn. Soc. Lond. 23, 495 –566. (doi:10.1111/j. 1096-3642.1860.tb00146.x) 4. 5. 6. Rowe MP, Coss RG, Owings DH. 1986 Rattlesnake rattles and burrowing owl hisses: a case of acoustic Batesian mimicry. Ethology 72, 53 –71. (doi:10. 1111/j.1439-0310.1986.tb00605.x) Rota J, Wagner DL. 2006 Predator mimicry: metalmark moths mimic their jumping spider predators. PLoS ONE 1, e45. (doi:10.1371/journal. pone.0000045) Barber JR, Conner WE. 2007 Acoustic mimicry in a predator–prey interaction. Proc. Natl Acad. Sci. 7. 8. USA 104, 9331 –9334. (doi:10.1073/pnas. 0703627104) Pinheiro CEG. 2003 Does Müllerian mimicry work in nature? Experiments with butterflies and birds (Tyrannidae). Biotropica 35, 356–364. (doi:10. 1111/j.1744-7429.2003.tb00589.x) Huheey JE. 1976 Studies in warning coloration and mimicry. VII. Evolutionary consequences of a Batesian– Müllerian spectrum: a model for Müllerian mimicry. Evolution 30, 86–93. (doi:10.2307/2407675) 6 Proc. R. Soc. B 282: 20150798 honeyeater and thornbill alarms broadcast alone, although a larger sample might detect small effects. A third explanation is that including mimicry simulates a chorus of alarm calling species, thereby increasing the perceived urgency or reliability of alarm information. Not only may alarm calls from multiple callers be a more reliable indication of danger than alarm calls from a lone caller [25,26], but more dangerous predators, such as hawks, generally provoke aerial alarms from multiple species [27]. Regardless of the exact mechanism, we show that mimicry confers an anti-predator advantage. Further work is required to disentangle the mechanisms by which the inclusion of mimicry increased the effectiveness of deception. In summary, we revealed the use of deceptive alarms to manipulate predator behaviour and an unexpected protective mechanism of mimicry. Similar anti-predator mechanisms for alarm signals and mimicry may be widespread in nature. For example, wild potato plants (Solanum berthaultii) deter herbivory by aphids (Myzus persicae) by mimicking aphid alarm pheromones [59], and myrmecophilous beetles mimic alarm pheromones of their ant hosts to evade aggressive interactions [60]. Eavesdropping on heterospecific alarm calls is widespread among vertebrates [20], so the use of either nonmimetic or mimetic alarm calls to deceive predators may also be common but overlooked. This is because researchers may assume incorrectly that alarm calls function only to warn others, or signal honestly to predators, rather than being used as deceptive signals directed at the predator. rspb.royalsocietypublishing.org silent when hunting prey. Demonstrating that dangerous model species are not required for protective function shows that mimicry can confer protection through unexpected mechanisms and expand upon the circumstances in which defensive mimicry could evolve. Indeed, it is the information conveyed by the mimetic signal that is important for its function, rather than any danger posed by the model itself. As mimicry of heterospecific alarm calls is common among songbirds [9,10], similar protective mechanisms for mimicry may also be common and requires further investigation. Aerial alarms caused temporary distraction such that deception would specifically be advantageous when nestlings are capable of fleeing the nest. Older nestling birds, including those of thornbills, commonly scramble or ‘explode’ from the nest prematurely when disturbed [29,54], and on an evolutionary timescale there is selection for young to fledge earlier and be more mobile when predation risk is higher [55]. Thornbill nests are surrounded by thick vegetation into which young can flee, so even short distraction of predators should increase nestlings’ chances of survival. Protecting older broods is more profitable because the fitness value of broods increases, and renesting opportunity decreases, with brood age [42]. Older nestlings can also be more vulnerable to predation because of louder begging that can attract predators [56,57]. A potential anomaly is that thornbills used aerial alarms in circumstances when nestlings are too young to flee, such as during our experiments. However, producing aerial alarm calls probably carries little risk to parents when a predator is attacking their nest, and in the absence of selection to restrict usage to circumstances when nestlings are capable of fleeing, mimicking aerial alarm calls could have evolved as a general response to nestling distress calls. Responses of thornbill parents to nestling distress calls appear similar throughout the nestling period, and older thornbill nestlings do explode from their nests when disturbed by humans while parents simultaneously mimic aerial alarm calls (B. Igic 2010–2011, personal observation). The effectiveness of this defensive strategy during natural nest predation situations requires further investigation. There are three potential mechanisms by which including mimicry prolonged distraction. First, mimicry may help evade identification, and thereby reduce the risk that targets identify alarms as deceptive [17]; however, non-mimetic aerial alarms still caused some distraction, suggesting this is not the complete explanation. Second, mimicry may enable thornbills to use alarms that are more salient to currawongs, such as alarms of larger heterospecifics that are more likely to share predators with currawongs [58]. Contrary to this hypothesis, currawongs foraging in the open responded similarly to Downloaded from http://rspb.royalsocietypublishing.org/ on June 18, 2017 9. 11. 12. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. drongo (Dicrurus paradiseus) in mixed-species bird flocks. Behav. Ecol. 21, 396 –403. (doi:10.1093/ beheco/arp203) Jurisevic MA, Sanderson KJ. 1998 Acoustic discrimination of passerine anti-predator signals by Australian raptors. Aust. J. Zool. 46, 369–379. (doi:10.1071/ZO97052) Wood SR, Sanderson KJ, Evans CS. 2000 Perception of terrestrial and aerial alarm calls by honeyeaters and falcons. Aust. J. Zool. 48, 127–134. (doi:10. 1071/ZO99020) Krama T, Krams I. 2005 Cost of mobbing call to breeding pied flycatcher, Ficedula hypoleuca. Behav. Ecol. 16, 37–40. (doi:10.1093/beheco/arh116) Armstrong EA. 1949 Diversionary display: part 1. Connotation and terminology. Ibis 91, 88 –97. (doi:10.1111/j.1474-919X.1949.tb02239.x) Pough FH. 1988 Mimicry of vertebrates: are the rules different? Am. Nat. 131, S67–S102. (doi:10. 2307/2461943) Dumbacher JP, Fleischer RC. 2001 Phylogenetic evidence for colour pattern convergence in toxic pitohuis: Müllerian mimicry in birds? Proc. R. Soc. Lond. B 268, 1971–1976. (doi:10.1098/rspb.2001.1717) Dumbacher JP, Spande TF, Daly JW. 2000 Batrachotoxin alkaloids from passerine birds: a second toxic bird genus (Ifrita kowaldi) from New Guinea. Proc. Natl Acad. Sci. USA 97, 12 970 –12 975. (doi:10.1073/pnas.200346897) Londoño GA, Garcı́a DA, Sánchez MM. 2015 Morphological and behavioral evidence of Batesian mimicry in nestlings of a lowland Amazonian bird. Am. Nat. 185, 135–141. (doi:10.1086/679106) Streby HM, Peterson SM, Lehman JA, Kramer GR, Iknayan KJ, Andersen DE. 2013 The effects of forcefledging and premature fledging on the survival of nestling songbirds. Ibis 155, 616 –620. (doi:10. 1111/ibi.12051) Cheng Y-R, Martin TE. 2012 Nest predation risk and growth strategies of passerine species: grow fast or develop traits to escape risk? Am. Nat. 180, 285–295. (doi:10.1086/667214) Leonard ML, Horn AG. 2006 Age-related changes in signalling of need by nestling tree swallows (Tachycineta bicolor). Ethology 112, 1020–1026. (doi:10.1111/j.1439-0310.2006.01259.x) Haff TM, Magrath RD. 2011 Calling at a cost: elevated nestling calling attracts predators to active nests. Biol. Lett. 7, 493–495. (doi:10.1098/rsbl.2010.1125) Carrasco MF, Blumstein DT. 2012 Mule deer (Odocoileus hemionus) respond to yellow-bellied marmot (Marmota flaviventris) alarm calls. Ethology 118, 243–250. (doi:10.1111/j.1439-0310.2011. 02002.x) Gibson RW, Pickett JA. 1983 Wild potato repels aphids by release of aphid alarm pheromone. Nature 302, 608– 609. (doi:10.1038/302608a0) Stoeffler M, Maier T, Tolasch T, Steidle JM. 2007 Foreign-language skills in rove-beetles? Evidence for chemical mimicry of ant alarm pheromones in myrmecophilous pella beetles (Coleoptera: Staphylinidae). J. Chem. Ecol. 33, 1382 –1392. (doi:10.1007/s10886-007-9315-0) 7 Proc. R. Soc. B 282: 20150798 13. 28. flocks. Auk 122, 108 –120. (doi:10.1642/00048038(2005)122[0108:acislm]2.0.co;2) Marchant S, Higgins PJ. 1993 Handbook of Australian, New Zealand and Antarctic birds, volume 2: raptors to lapwings. Melbourne, Australia: Oxford University Press. Higgins PJ, Peter JM. 2002 Handbook of Australian, New Zealand and Antarctic birds, volume 6: pardalotes to shrike-thrushes. Melbourne, Australia: Oxford University Press. Higgins PJ, Peter JM, Cowling SJ. 2006 Handbook of Australian, New Zealand and Antarctic birds, volume 7: boatbill to starlings. Melbourne, Australia: Oxford University Press. Wood KA. 1998 Seasonal changes in diet of pied currawongs Strepera graculina at Wollongong, New South Wales. Emu 98, 157 –170. (doi:10.1071/ MU98023) Green DJ, Cockburn A. 1999 Life history and demography of an uncooperative Australian passerine, the brown thornbill. Aust. J. Zool. 47, 633 –649. (doi:10.1071/ZO99052) Yom-Tov Y, McCleery R, Purchase D. 1992 The survival rate of Australian passerines. Ibis 134, 374–379. (doi:10.1111/j.1474-919X.1992.tb08017.x) Charif RA, Waack AM, Strickman LM. 2010 RAVEN PRO 1.4 user’s manual. Ithaca, NY: Cornell Laboratory Australia of Ornithology. Fallow PM, Gardner JL, Magrath RD. 2011 Sound familiar? Acoustic similarity provokes responses to unfamiliar heterospecific alarm calls. Behav. Ecol. 22, 401– 410. (doi:10.1093/beheco/arq221) Waite TA, Grubb TC. 1987 Dominance, foraging and predation risk in the tufted titmouse. Condor 89, 936 –940. (doi:10.2307/1368550) Vuong QH. 1989 Likelihood ratio tests for model selection and non-nested hypotheses. Econometrica 57, 307– 333. (doi:10.2307/1912557) R Development Core Team. 2011 R: a language and environment for statistical computing, 2.14.1v. Vienna, Austria: R Foundation for Statistical Computing. Pinheiro J, Bates D, DebRoy S, Sarkar D. 2012 nlme: linear and nonlinear mixed effects models, R package version 3.1 –106. See http://cran.r-project. org/web/packages/nlme/index.html. Zeileis A, Kleiber C, Jackman S. 2008 Regression models for count data in R. J. Stat. Soft. 27, 1 –25. Hothorn T, Bretz F, Westfall P. 2008 Simultaneous inference in general parametric models. Biom. J. 50, 346 –363. (doi:10.1002/bimj.200810425) Montgomerie RD, Weatherhead PJ. 1988 Risks and rewards of nest defence by parent birds. Q. Rev. Biol. 63, 167 –187. (doi:10.1086/4158) Witkin SR, Ficken MS. 1979 Chickadee alarm calls: does mate investment pay dividends? Anim. Behav. 27, 1275–1276. (doi:10.1016/0003-3472(79) 90078-2) Magrath RD, Haff TM, Horn AG, Leonard ML. 2010 Calling in the face of danger: predation risk and acoustic communication by parent birds and their offspring. Adv. Stud. Behav. 41, 187–253. (doi:10. 1016/S0065-3454(10)41006-2) Satischandra SHK, Kodituwakku P, Kotagama SW, Goodale E. 2010 Assessing ‘false’ alarm calls by a rspb.royalsocietypublishing.org 10. Kelley LA, Coe RL, Madden JR, Healy SD. 2008 Vocal mimicry in songbirds. Anim. Behav. 76, 521–528. (doi:10.1016/j.anbehav.2008.04.012) Dalziell AH, Welbergen JA, Igic B, Magrath RD. 2015 Avian vocal mimicry. Biol. Rev. 90, 643– 658. (doi:10.1111/brv.12129) Goodale E, Kotagama SW. 2006 Context-dependent vocal mimicry in a passerine bird. Proc. R. Soc. B 273, 875–880. (doi:10.1098/rspb.2005.3392) Ratnayake C, Goodale E, Kotagama S. 2010 Two sympatric species of passerine birds imitate the same raptor calls in alarm contexts. Naturwissenschaften 97, 103 –108. (doi:10.1007/ s00114-009-0617-7) Kelley LA, Healy SD. 2012 Vocal mimicry in spotted bowerbirds is associated with an alarming context. J. Avian Biol. 43, 525–530. (doi:10.1111/j.1600048X.2012.05863.x) Igic B, Magrath RD. 2014 A songbird mimics different heterospecific alarm calls in response to different types of threat. Behav. Ecol. 25, 538–548. (doi:10.1093/beheco/aru018) Wickler W. 2013 Understanding mimicry – with special reference to vocal mimicry. Ethology 119, 259–269. (doi:10.1111/eth.12061) Munn CA. 1986 Birds that ‘cry wolf’. Nature 319, 143–145. (doi:10.1038/319143a0) Flower TP, Gribble M, Ridley AR. 2014 Deception by flexible alarm mimicry in an African bird. Science 344, 513–516. (doi:10.1126/science.1249723) Flower T. 2011 Fork-tailed drongos use deceptive mimicked alarm calls to steal food. Proc. R. Soc. B 278, 1548 –1555. (doi:10.1098/rspb.2010.1932) Møller AP. 1990 Deceptive use of alarm calls by male swallows, Hirundo rustica: a new paternity guard. Behav. Ecol. 1, 1–6. (doi:10.1093/beheco/1.1.1) Magrath RD, Haff TM, Fallow PM, Radford AN. 2015 Eavesdropping on heterospecific alarm calls: from mechanisms to consequences. Biol. Rev. 90, 560–586. (doi:10.1111/brv.12122) Caro T. 2005 Antipredator defenses in birds and mammals. Chicago, IL: The University of Chicago Press. Koops MA. 2004 Reliability and the value of information. Anim. Behav. 67, 103 –111. (doi:10. 1016/j.anbehav.2003.02.008) Searcy WA, Nowicki S. 2005 The evolution of animal communication: reliability and deception in signaling systems. Princeton, NJ: Princeton University Press. Igic B, Magrath RD. 2013 Fidelity of vocal mimicry: identification and accuracy of mimicry of heterospecific alarm calls by the brown thornbill. Anim. Behav. 85, 593–603. (doi:10.1016/j.anbehav.2012.12.022) Sloan JL, Hare JF. 2008 The more the scarier: adult Richardson’s ground squirrels (Spermophilus richardsonii) assess response urgency via the number of alarm signallers. Ethology 114, 436–443. (doi:10.1111/j.1439-0310.2008.01479.x) Blumstein DT, Verneyre L, Daniel JC. 2004 Reliability and the adaptive utility of discrimination among alarm callers. Proc. R. Soc. Lond. B 271, 1851–1857. (doi:10.1098/rspb.2004.2808) Goodale E, Kotagama SW, Holberton RL. 2005 Alarm calling in Sri Lankan mixed-species bird