Prey are sensitive to even subtle cues of predation risk, which provides the evolutionary potential for parasites to exploit host risk perception. Brood parasitic common cuckoos (Cuculus canorus) lay their eggs in the nests of host species and their secretive laying behaviour enables them to evade host defences. Therefore, it seems paradoxical that female cuckoos often give a conspicuous ‘chuckle’ call after parasitizing a host’s clutch. Here, we show that this hawk-like chuckle call increases the success of parasitism by diverting host parents’ attention away from the clutch and towards their own safety. In our field experiments, reed warbler (Acrocephalus scirpaceus) hosts paid no more attention to the ‘cuck-oo’ call of the male common cuckoo than the call of a harmless dove. However, the chuckle call of the female cuckoo had the same effect as the call of a predatory hawk in distracting the warblers’ attention and reducing rejection of a foreign egg. Our results show that the female cuckoo enhances her success by manipulating a fundamental trade-off in host defences between clutch and self-protection.
Parasites evolve not only to evade host defences but also to manipulate host behaviour1. Endo-parasites do this inside the bodies of their hosts by physiological manipulation of host risk taking to enhance parasite transmission2. Here, we test whether a brood parasitic cuckoo manipulates host perception of predation risk using an acoustic signal—a hawk-like call—that might misdirect host defences and thereby reduce the chance that hosts detect parasitism. It is well known that adult birds distinguish threats to themselves from those to their offspring3; for example, parents flee from hawks but readily attack nest predators of no direct threat to themselves4. In theory, cuckoos could exploit this fundamental trade-off in host defences by using deceptive signals.
Obligate brood parasites lay their eggs in the nests of other species—the hosts—which are then tricked into raising parasite young at the expense of some, or all, of their own offspring5. Previous studies have shown that hosts of the common cuckoo Cuculus canorus defend against parasitism by mobbing adult cuckoos6 (a first line of defence)7 and by rejecting eggs that differ from their own8,9,10. Hosts also monitor cuckoo activity in the vicinity of their nest and vary these defences in relation to local parasitism risk11,12,13,14. In response, cuckoos have evolved host egg mimicry15 and remarkable secrecy and speed when they parasitize a host nest16. Therefore, it seems paradoxical that female cuckoos often call while they monitor host nests, and especially just after parasitizing a clutch16. Their chuckle (or ‘bubble’)16 call—a rapidly repeated ‘kwik-kwik-kwik…’—is similar in fundamental frequency and rate to the ‘kiii-kiii-kiii…’ call of Accipiter hawks and strikingly different from the familiar two-note call of the male cuckoo (Fig. 1a).
We propose that the female cuckoo chuckle call tricks the hosts into responding vigilantly as if they were exposed to danger from a hawk, instead of from a cuckoo. This would divert host attention from clutch protection to self-protection3, and so reduce the chance of the hosts detecting that they have been parasitized. As noted by Wallace17, many parasitic cuckoos also resemble hawks in appearance. Indeed, experiments have shown that this visual resemblance makes hosts more reluctant to mob18. Therefore, an alternative hypothesis is that the female cuckoo’s chuckle call might provide an additive benefit to enhance her hawk mimicry in order to bypass the hosts’ first line of defence. In this study, we test both of these potential benefits of the female cuckoo chuckle call in overcoming the host’s mobbing and egg rejection defences.
First, we tested whether female cuckoo calls provoke vigilance in reed warblers, a favourite cuckoo host in marshland8. Our playback experiment had four treatments (Fig. 1a): the call of a female cuckoo (a threat to the clutch but not to adults), the call of a Eurasian sparrowhawk Accipiter nisus (a threat to the adults but not to the clutch), the call of a male cuckoo (no direct threat to the clutch nor to the adults, but a potential cue to parasitism risk) and the call of a collared dove Streptopelia decaocto (a harmless control). All four calls were frequently encountered on the study site. At 24 nests where reed warblers were incubating a recently completed clutch, we placed a speaker 5 m from the nest and recorded host responses on video to each of the four calls in sequence (see Methods and Fig. 1d). There were marked differences in responses across the four treatments (Fig. 1b and experiment 1 in Table 1). As predicted, reed warblers were more likely to become vigilant (scanning the surroundings for danger; see Methods) during hawk calls than dove calls (generalized linear mixed-effects model: χ2 = 12.02; P < 0.001). There was little response to male cuckoo calls and this did not differ from the response to dove calls (χ2 = 0.37; P = 0.54). In contrast, the hosts responded strongly to female cuckoo calls (Fig. 1b) and this did not differ from the response to hawk calls (χ2 = 0.62; P = 0.43). When vigilance responses occurred they were rapid, occurring within the first few syllables of playback (see Methods).
The increase in vigilance to both the female cuckoo and hawk calls may arise from their acoustic similarity or because both are independently recognized as a threat to reed warblers. We therefore repeated the playbacks to tits (Paridae)—frequent victims of sparrowhawks19, but typically unsuitable hosts for cuckoos in Europe. Hence, they should not respond to female cuckoo calls as a threat unless they mistake them for hawk calls. We presented the playbacks to 60 individually recognizable tits at experimental feeders (28 blue tits Cyanistes caeruleus and 32 great tits Parus major). Each individual experienced just one of the four treatments broadcast from a speaker 5 m from the feeder. The playback order was randomized and we recorded responses on video to each of the four calls in separate trials (see Methods and Fig. 1e). There were no differences in the responses between blue and great tits (generalized linear model (GLM): χ2 = 1.62; P = 0.20). Once again, the responses occurred rapidly and the same marked differences in vigilance were noted as for reed warblers (Fig. 1c and experiment 2 in Table 1). Tits were more likely to become vigilant during hawk calls than dove calls (χ2 = 9.36; P = 0.002), but the response to male cuckoo calls was no different from the response to dove calls (χ2 = 0.83; P = 0.36). In contrast, female cuckoo calls increased vigilance as much as hawk calls (χ2 = 2.00; P = 0.16). As cuckoos are no threat to tits, their similar response to the calls of female cuckoos and hawks is likely to have resulted from the perceived acoustic similarity.
Next, we tested whether exposure to the four calls influenced reed warbler nest defences (egg rejection and mobbing). We removed one egg at random from 72 reed warbler clutches on the day they laid their fourth egg (when they would still be vulnerable to parasitism), painted it brown and returned it to the nest to simulate parasitism (Fig. 2a; female cuckoos typically remove a host egg and then lay their own egg in its place; see Methods). We then placed a balsa wood model of an adult cuckoo on the nest with a speaker concealed next to it. Each reed warbler pair then received just one of the four playbacks. We measured host mobbing responses (mandible snaps and rasp calls) for one minute after the first member of the pair returned to within 1 m of the nest, then playback was triggered remotely and we recorded host mobbing responses for another minute (see Methods). This experiment allowed us to test whether the female chuckle influenced the first line of defence (mobbing) and/or egg rejection defences.
The playback treatment had a marked effect on egg rejection (Fig. 2b and experiment 3 in Table 1). When we checked the nests one day after the trial, two clutches had been depredated and of the remaining 70 clutches, the foreign egg had been rejected in 32 cases (one by nest desertion and all others by targeted ejection from the nest). As predicted from our hypothesis that increased vigilance diverts host attention away from the clutch, reed warblers that had been exposed to hawk or female cuckoo calls were more likely to accept the foreign egg (Fig. 2b). The effect of playback treatment was still apparent when we checked the clutches again three days after the trial, once there had been an opportunity for delayed rejection (n = 68 nests; two clutches had been depredated since day 1). Reed warblers were still more than twice as likely to retain a foreign egg in their clutch following female cuckoo calls compared with male cuckoo calls (χ2 = 5.99; P = 0.014).
In contrast, the call type did not affect mobbing responses (experiment 3 in Table 1 and Supplementary Fig. 1). Neither the propensity to mob after playback (GLM: χ2 = 4.84; n = 72 nests; P = 0.18) nor the mobbing intensity (F = 0.76; n = 44 nests; P = 0.52) differed significantly across the treatments. As in previous studies, individual mobbing responses also did not predict egg rejection12 (current study: χ2 = 0.69; P = 0.40). These results are perhaps not surprising given that mobbing is a generalized defence against all intruders at the nest, and individuals show consistent differences in mobbing intensity that are not specific to cuckoos20.
Male cuckoo calls had no more effect on host responses than the harmless dove control. Male cuckoos call conspicuously from exposed perches to repel rival males21 and attract females22, but their calls are likely to be a poor predictor of local parasitism risk because males roam widely and call frequently even when females are scarce22. Conversely, the presence of a female cuckoo is a strong predictor of parasitism risk6, which explains why they are more secretive than males and call less frequently16. This would reduce the potential for hosts to learn to discriminate female cuckoo chuckles from hawk calls. Our results also explain why female cuckoos typically call just after laying16,22, which is precisely when it would pay them to distract host attention from the clutch23. A female cuckoo can choose an opportune time to glide down to the nest when the hosts are away, but there is an increasing probability that the hosts will return or at least see her leaving, and this is when it might be most beneficial to distract them with a call. Similar vocal trickery has been demonstrated in kleptoparasitic drongos (Dicrurus adsimilis), whose false alarm calls enable them to steal food by distracting the attention of foragers24.
Hawk-like calls are typical for female cuckoos of the Cuculus genus and are quite unlike the male calls, which are simple coos and whistles. A comparison across the cuckoo subfamily Cuculinae suggests that sexually dimorphic calls have evolved with parasitism: 19 of the 58 parasitic species exhibit sex differences, whereas none of the 32 non-parasitic species do so25. In many species, sex-specific calls have socially selected functions; for example, to attract mates and repel rivals26. Female cuckoos rarely call22, which suggests the calls are not important for territory defence, although they may function in attracting males. However, their timing (after laying), acoustic similarity with hawk calls and our experimental results all suggest that their calls have been shaped by host defences. Our results suggest that female chuckles play an important role in a suite of specialized female traits associated with a brood-parasitic lifestyle, including secretive behaviour to avoid alerting hosts8,16, polymorphic plumage to confuse host recognition27,28 and brain specialization to facilitate spatial memory of the locations of host nests29.
To the human ear, there are clear differences between female chuckle calls and hawk calls. Nevertheless, manipulation by imperfect mimicry is frequent in the natural world, and resemblance to hawk calls in some key features might be sufficient to trick hosts30,31. If hosts respond to a female cuckoo call as though it were a hawk, they are less likely to reject a cuckoo egg, but if they fail to respond to a hawk call they may lose their life. Predators are secretive so it is not surprising that even brief encounters, including auditory cues, can have long-lasting effects on prey behaviour19,32,33. The benefits of a more rapid response to hawk-like signals inevitably lead to increased discrimination errors34, leaving hosts vulnerable to exploitation by cuckoo chuckles. As a result, the female cuckoo might have ‘the last laugh’ in this particular battle between host defence and parasite trickery.
Study species and field sites
Our experiments were conducted from March to July in 2016 at three field sites in Cambridgeshire, United Kingdom. The playback experiments with great tits and blue tits were conducted in the Cambridge University Botanic Garden (52° 19′ 35′′ N, 0° 12′ 58′′ E) and Madingley Wood (52° 21′ 71′′ N, 0° 04′ 89′′ E). The experiments with reed warblers were conducted on Wicken Fen (52° 18′ 29′′ N, 0° 16′ 50′′ E), where we have studied reed warblers and cuckoos since 198512. Each year, circa 300 pairs of reed warblers nest along the reed fringes of waterways and defend 11–35 m linear territories. On average, around 5% of these nests are parasitized by cuckoos who monitor host nests from perches in trees and large shrubs near the reeds. Our experiments closely follow procedures detailed elsewhere12 and are described briefly below.
Each exemplar was extracted from original uncompressed WAV files obtained from XenoCanto recordists (http://www.xeno-canto.org/; Supplementary information). For all three playback experiments, we used the same exemplars of each call type: four different exemplars for each call type (16 in total). Each playback track of a female cuckoo or sparrowhawk call comprised one natural phrase of repeated syllables extracted from the recordings, whereas for the male cuckoo each exemplar comprised three natural ‘cuck-oo’ phrases and for the dove call, each exemplar comprised two natural ‘coo-cooo-coo’ phrases. The average duration across the tracks was 3.06 ± 0.17 s (mean ± standard error (s.e.m.); dove: 3.11 ± 0.18 s; male cuckoo: 3.73 ± 0.21 s; female cuckoo: 2.18 ± 0.13 s; sparrowhawk: 3.21 ± 0.31 s). Visual inspection of the data revealed no consistent difference between playback exemplars of a given call type on responses in any of the experiments. Vigilance responses by both reed warblers and tits occurred rapidly within the first few syllables of the playback (see below), so small differences in the playback duration did not affect the results. Each playback track was composed using Cool Edit Pro (version 2.0). First, we filtered low-frequency background noise (below 100 Hz) from each track, and then we added 10 s of silence before and after each call clip (to allow the observers time to prepare to record the responses). All responses were measured from the onset of the playback call and not the onset of the track. The tracks were then standardised to a peak amplitude of −15 dB and saved in stereo format as uncompressed WAV files.
Given that we used the calls of three different species, and there is no information available on natural production amplitudes for female cuckoo calls, it was not possible to exactly match the playback amplitudes to natural levels in our experiment. Instead, we standardized all playback amplitudes to the same level and based our amplitude levels on those used in previous studies using hawk call playbacks35. The playback amplitude was standardized both within and across treatments by calibration of the peak amplitude (HandyMAN TEK1345 Sound Level Meter; Metrel UK) at the distance of the fixed location for all subjects (either the nest or feeder; see experiments 1–3 below for further details). The example spectrograms in Fig. 1a were generated using Raven Pro version 1.436 with the Hamming window function at 1,024 points, 56 Hz, 96% overlap and 0.99 ms.
Video recording and analysis
The behavioural responses were recorded as video files using a HC-V270EB-K HD Camcorder (Panasonic) at 50 frames per second and a resolution of 1,920 × 1,080. The videos were scored blind to treatment by first marking the time point of the playback call onset and then subsequently watching muted videos slowed to × 0.25 of the recorded speed on VLC (VideoLan). The ‘Jump to time’ (version 2.1) extension was used to identify the exact frame in which the response began (see Supplementary Information for video examples).
Experiment 1: vigilance in reed warblers (cuckoo hosts)
At Wicken Fen and adjacent waterways, we conducted a repeated-measures playback experiment at 24 reed warbler nests (May to June 2016), with each nest exposed to all four treatments in an order defined by a Latin square to remove the potential effects of order exposure. In addition, a 10 min rest period was included between each trial and no effects of the playback order were found suggesting that this period was adequate (Table 1; see below). Calls were broadcast using a Pignose 7100 field speaker placed 1 m above the ground and calibrated to an 80 dBA peak amplitude 1 m from the speaker. The nests were sufficiently separated in space or time to avoid the effects of the playbacks on neighbouring nests12.
While incubating, reed warblers typically sit deep in the nest cup with their head below the nest rim. Occasionally, they stretch their necks to peer out over the nest rim and such vigilance scans are associated with approaching threats; for example, when a human or dog approaches the nest. The bird will subsequently leave the nest if the threat persists. On average, these vigilance scans last 2.5 s and range between 489 ms and 11.5 s (n = 25 scans measured from baseline incubation behaviour using the methods described above). We categorized subjects as exhibiting vigilance behaviour if they were observed to scan by peering over the nest rim with the neck extended for more than 1 s continuously during the trial. Vigilance responses (58 of 96 trials) began rapidly, with 52 (90%) beginning within 500 ms of the playback call onset (which equates to during the first 1–5 syllables; see Fig. 1a) and all but two beginning within the first second of the call playback.
Experiment 2: vigilance in great tits and blue tits (not cuckoo hosts)
We conducted a second experiment at Cambridge University Botanic Garden and Madingley Wood using 60 individually identifiable (colour-banded or pit-tagged) free-living parids: 32 great tits (P. major) and 28 blue tits (C. caeruleus). These species nest in tree holes that are inaccessible to female cuckoos, and in Europe they are not parasitized (although there are records of cuckoo parasitism in great tits in Asia37). We used experimental peanut feeders as a standardized location from which to conduct the playback trials during March and April 2016 (before cuckoos had arrived in the region). The trials began when a bird had been on the feeder for at least 10 s and when no other tits were present on the feeder or in close proximity. The speaker was located 5 m away from the feeder and playbacks were broadcast in a randomized order across individuals and at a standardized amplitude (as in experiment 1). Given unpredictable visits by individuals to feeders, it was not possible to ensure that each individual received all treatments, so each individual received just one playback treatment.
Behavioural responses were recorded as video files (as above). When feeding on peanut feeders, tits regularly survey the surroundings with short, regular ‘look-ups’ that last for an average of 539 ms (range: 172–3,303 ms; n = 50 look-ups measured from baseline feeding activity). Vigilance behaviour was defined as the subject scanning the surroundings for more than 1 s continuously during the trial or scanning the surroundings before immediately leaving the feeder during the trial. Vigilance responses (33 of 60 trials) began rapidly with 30 (91%) beginning within 500 ms of the playback call onset (which equates to during syllables 1–5; see Fig. 1a), and all began within the first second of the call playback.
Experiment 3: nest defences in reed warbler hosts
At 72 reed warbler nests at Wicken Fen and adjacent waterways, we conducted an experiment on the day the fourth egg was laid (most pairs lay a clutch of four eggs)8. We simulated parasitism with a foreign egg using previously validated methods8,12. We selected one egg from the nest at random, painted it uniform brown with Rowney acrylic ‘burnt sienna’ paint and then replaced the egg in the nest. This simulates the behaviour of female cuckoos, which first remove a host egg before they lay their own egg in the nest8. We used ‘non-mimetic’ brown eggs for two reasons: (1) these are similar to the eggs laid by some female cuckoos on our study site8 and (2) reed warblers have reduced their propensity to reject eggs over the past 30 years in concert with the decline in cuckoos and, hence, a decline in parasitism risk11. Highly mimetic eggs are now rarely rejected and so responses to non-mimetic eggs provide a better measure of host rejection11.
Having ‘parasitized’ the nest with a foreign egg, we placed a model adult cuckoo on top of the nest. As in previous experiments12, we alternated between two virtually identical balsa wood cuckoo models, which did not differ in their effects on responses. The models were painted with grey upper parts and pale under parts with barring. Grey females are the most common morph on our study site, and are similar to males in appearance. The responses to these models correlated strongly with those to taxidermic cuckoo mounts and were similar to responses to a live cuckoo38,39. We concealed a small speaker (Altec Lansing12) next to the nest and broadcast the playbacks calibrated to a peak amplitude of 75 dBA at 1 m—the distance from the nest of subjects when the playbacks began. Each nest received just one playback treatment chosen at random. Female cuckoos typically produce one chuckle phrase after laying so our playbacks mimicked natural call production16,22. Once again, the nests were sufficiently separated in space and time to avoid the effects of the playbacks on neighbouring nests12. They were also different pairs from those tested in experiment one.
We retreated from the nest to observe each pair’s behavioural response to the model cuckoo, following previous protocols6, after which we remotely triggered one of the four playback treatments chosen at random and recorded the behavioural responses to the playback for one minute afterwards. At the end of the minute, we removed the model and playback speaker. We then checked the nest contents one day and three days after the trial to assess whether the painted egg had been ‘accepted’ (that is, it was still present in the nest and clutch being incubated) or ‘rejected’ (that is, it was no longer present in the clutch being incubated or the clutch had been deserted and the pair had begun a replacement nest nearby). Our previous work has shown that most rejections of real cuckoo eggs8 and all rejections of experimental painted eggs12 occur within three days.
All statistical analyses were conducted using the statistical software R version 3.3.240 and were two sided. The models were checked for normality of residuals, homogeneity of variance and over-dispersion by manual visual inspection and using R package DHARMa41. Statistical modelling used a full model approach for the mixed-effects model (package lme442) and the GLM analyses: all terms of interest were fitted and then significance testing was performed via likelihood ratio tests to determine which factors resulted in a significant reduction in explanatory power when removed (Table 1). The significance of the factor levels for ‘call type’ was determined using likelihood ratio tests to assess whether collapsing the two levels of interest (for example, ‘hawk’ and ‘female cuckoo’) resulted in a significant reduction in the explanatory power of the model compared with a model with all four levels43.
For experiment 1 (vigilance in reed warbler hosts), a generalized linear mixed-effects model42 with a binomial error structure and logit-link function was used to test the prediction that female cuckoo calls provoke vigilance behaviour in reed warblers. Responses were coded as ‘vigilant’ (yes or no) according to the definition above. ‘Nest identity’ was fitted as a random term to control for repeated measures at each nest for each of the four playback call types. ‘Call type’ (dove, male cuckoo, female cuckoo or hawk) and ‘order of playback’ (first, second, third or fourth) were each fitted as four-level fixed effects.
For experiment 2 (vigilance in tits), a GLM with a binomial error structure and logit-link function was used to test the prediction that female cuckoo calls provoke vigilance behaviour in tits. Again, responses were coded as ‘vigilant’ (yes or no) according to the definition above, ‘call type’ (dove, male cuckoo, female cuckoo or hawk) was fitted as a four-level fixed effect and ‘species’ (blue tit or great tit) was fitted as a two-level fixed effect.
For experiment 3 (nest defences in reed warbler hosts), two GLMs with a binomial error structure and logit-link function were used to test the prediction that female cuckoo calls reduce egg rejection by reed warbler hosts by one day and three days after the trial. Egg rejection responses were coded as ‘reject’ = 0 or ‘accept’ = 1. ‘Call type’ (dove, male cuckoo, female cuckoo or hawk) was fitted as a single four-level fixed effect. A GLM with a binomial error structure and logit-link function was used to test the prediction that female cuckoo calls reduce mobbing behaviour by reed warbler hosts. Mobbing propensity responses were coded as ‘mob after playback’ (yes or no) based on whether or not the parents mobbed the model after the playback. ‘Call type’ (dove, male cuckoo, female cuckoo or hawk) was fitted as a four-level fixed effect and ‘mob before playback’ (yes or no) was fitted as a two-level fixed effect. For the individuals that mobbed after playback, an additional GLM with a normal distribution was used to investigate the effects of playback treatment on mobbing intensity (call rate). The response term for ‘mobbing intensity’ was the mobbing rate (number of mobbing calls per minute) after the playback. Again, ‘call type’ (dove, male cuckoo, female cuckoo or hawk) was fitted as a four-level fixed effect and ‘mobbing call rate before the playback’ was fitted as a covariate to control for the marked variation between individuals in mobbing responses often observed in this species12.
All protocols were reviewed and licenced by Natural England.
The data that support the findings of this study are available in the Supplementary Information.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We thank the National Trust for permission to work on Wicken Fen, Natural England for the licenses, H. Rowland, J. Mackenzie and T. Dixit for field assistance, C. Spottiswoode and A. Jungwirth for comments, and especially D. Cram for comments and assistance throughout. This work was funded by Natural Environment Research Council grant NE/M00807X/1.