1. Anthropogenic noise can affect intra-pair communication and therefore interfere with reproductive success. However, many animals have various signal strategies to cope with noise, although it is unclear whether they rely on direct auditory feedback from their own perception of noise or signal-to-noise level or on indirect social feedback from receivers.
2. We studied the role of social feedback on male great tit (Parus major) song adjustment by exclusively exposing females to artificial traffic noise inside their nest box.
3. We found a delay in female response latencies to male song in the noisy condition compared to the control condition on the first day of noise exposure. Males from the noise treatment group, not directly exposed to noise themselves, sang closer to the nest box within 3 days after the start of exposure.
4. The male's closer proximity to the nest box most likely led to the observed higher song amplitudes at the noisy nest boxes compared to quiet control nest boxes, and explains why the inside signal-to-noise ratios were restored to equal levels between treatment and control nest boxes after several days of exposure.
5. The initial difference between treatment groups in female response latencies at the start of exposure also disappeared accordingly.
6. Our results strongly suggest an active role for female birds in steering male song behaviour under noisy conditions. Males did not receive direct exposure during intra-pair communication, but adjusted their behaviour in the predicted direction. These data are important to understand the mechanisms related to communication in noise and reveal the critical role of ecology in shaping animal interactions.
Communication between members of a breeding pair plays an important role in many different contexts (Bradbury & Vehrencamp 1998; Helm, Piersma & Van Der Jeugd 2006). Pair members that cooperate during territory defence or parental care can benefit from exchanging signals as this enables them to synchronize their behaviours and investments (Hall 2009; Johnstone 2011). Optimal communication requires signal efficacy, which strongly depends on environmental conditions and signals are likely to match the properties of habitats to maximize transmission between individuals (Naguib et al. 2011; Endler 1992; Wiley & Richards 1978). However, habitats can change rapidly, especially in areas occupied by humans, forcing strong selection on signalling behaviour (Van Der Sluijs et al. 2011; Seehausen, Vanalphen & Witte 1997; Smith et al. 2008).
Many animals use sounds to communicate with their mates, but also using this medium becomes increasingly difficult in an urbanizing world (Barber, Crooks & Fristrup 2009; Warren et al. 2006; Brumm & Slabbekoorn 2005). Human-generated noise coming from heavy machinery, such as factories and traffic, is known to interfere with signal detection and may affect intra-pair communication and consequently reproductive success (Slabbekoorn & Ripmeester 2008; Barber, Crooks & Fristrup 2009). Anthropogenic noise has been reported to affect communication in frogs (Bee & Swanson 2007; Lengagne 2008), mammals (Brumm et al. 2004; Parks et al. 2011) and fish (Vasconcelos, Amorim & Ladich 2007; Slabbekoorn et al. 2010; Codarin et al. 2009). There are also several reports on masking of male–female communication by anthropogenic noise in birds and has been related to reduced mate attraction (Gross, Pasinelli & Kunc 2010; Habib, Bayne & Boutin 2007) and breeding performance (Halfwerk et al. 2011b; Francis, Ortega & Cruz 2011, 2009).
Noise-related selection pressures on communication have likely led to the evolution of a variety of strategies to cope with fluctuating noise levels, both on the side of the sender and on the side of the receiver (reviewed in Brumm & Slabbekoorn 2005). Senders can raise amplitude or call rate (Potash 1972; Brumm et al. 2004), or avoid overlap between their signals and the noise (Zelick & Narins 1985; Potvin, Parris & Mulder 2011; Slabbekoorn & Peet 2003). Receivers have evolved various perceptual mechanisms that allow signal extraction from noisy environments, referred to with often partly overlapping terminology such as spatial release from masking, auditory stream segregation and the ‘cocktail party effect’ (Bee & Micheyl 2008; Brumm & Slabbekoorn 2005). Furthermore, both senders and receivers can affect signal transmission by choosing a particular location during intra-pair communication. Birds can improve detection and discrimination thresholds by moving closer (Slabbekoorn 2004), choosing higher song posts (Mathevon, Aubin & Dabelsteen 1996) or staying in- or outside their nest cavities (Blumenrath, Dabelsteen & Pedersen 2004), but we currently lack insight into whether such spatial strategies are exploited under fluctuating noise conditions.
Birds can use either an internal or an external feedback mechanism to sing louder, higher, faster or closer to their intended receivers when confronted with low-frequency urban noise (Halfwerk & Slabbekoorn 2009). Male changes in singing behaviour can be based on direct auditory feedback from noise level or the signal-to-noise ratio of their own vocal output (internal feedback). Males may also rely on indirect social feedback from conspecific receivers, such as females or territorial neighbours (external feedback) to adjust their songs in response to changing noise conditions (Halfwerk & Slabbekoorn 2009). Distinguishing between these two types of mechanisms requires noise exposure to either the sender or the receiver during communication, which is challenging for field as well as laboratory conditions, given the spherical spread of sound and the ability by sound waves to bend around obstacles.
The great tit (Parus major) provides an excellent study system to expose only one side of the communication channel to increased noise levels. Females, at the start of the breeding season, interact with their mates from within their nest cavities during the dawn chorus ritual (Gorissen & Eens 2004; Halfwerk et al. 2011a). Females have been found to call and emerge less in response to playback of their mate's song under noisy conditions (Halfwerk et al. 2011a), but they can also be exposed to noise inside an artificial nest box during natural dawn singing of their own male. Their response to playback was found to be masking-specific as females responded less to low song types compared to high song types when exposed to artificial traffic noise (Halfwerk et al. 2011a). Therefore, males could theoretically use female response as an external social feedback mechanism, for instance during masking-dependent song-type switching (Halfwerk & Slabbekoorn 2009), but which remains to be tested empirically.
In the present study, we exposed females at the peak of intra-pair interactions to artificial traffic noise inside their nest box, while leaving the singing male outside unaffected. We monitored signal-to-noise ratios of male song inside the nest box and measured female calling response to assess whether females in the noise treatment reduced or delayed calling response, which could serve as external noise-dependent feedback to the singing males outside the nest box. We recorded male song behaviour throughout the experimental period and additionally noted song post use after 3 days of exposure and expected males to sing higher or louder songs, or to sing in closer proximity, depending on the noise exposure to their mates inside the nest box.
Materials and methods
Study site and species
The study was conducted in a nest box population of great tits at ‘Nationaal Park Dwingelderveld’, the Netherlands, between March and May, in 2009 and 2010. The nest boxes were divided over four different sites that either consisted of deciduous forest or mixed woodland. The great tit (Parus major) is a hole-nesting passerine that uses song in both male–male as well as male–female communication (Krebs, Ashcroft & Webber 1978; Mace 1987). Females start to roost inside their nest cavity (or wooden nest box in our population) at this stage and are visited by their mates who will sing towards them from a nearby song perch. Males typically start to sing 30–45 min before sunrise and end their dawn chorus song when the female emerges from the nest box, after which the pair often copulates (Mace 1987; Kluyver 1951).
Male–female interaction rapidly increases when the female begins with nest building (Mace 1987). Females are actively listening to their singing mates from inside their nest box and occasionally call back in response (Fig. 1; Gorissen & Eens 2004). Female calling starts a few days before beginning of egg-laying and rapidly decreases again when the first eggs have been laid (Halfwerk et al. 2011a). Males have a small repertoire of song types (2–6 in our study population) that they display with eventual variety (Lambrechts & Dhondt 1986; Rivera-Gutierrez, Pinxten & Eens 2010): the same song type is repeated for several minutes before a switch is made to a different song type (Fig. 1). The majority of song types consist of a low-frequency note and a high-frequency note, in the range of 2–9 kHz (Fig. 2; Halfwerk & Slabbekoorn 2009; Franco & Slabbekoorn 2009). Male–female communication also rapidly decreases after the first few eggs have been laid, with males spending less time in close proximity of the nest box and females responding less to the song of their mates (Halfwerk et al. 2011a).
The behavioural data presented here are part of a larger study on the impact of noise on great tit breeding behaviour. Territories were mapped in March and early April, and nest boxes were checked for nest building every other day. Nest box treatment was randomly assigned and 67 great tit pairs started nest building in a control box, whereas 68 pairs started building in noise box. A total of 29 pairs abandoned their nest box before the incubation phase, but the rates were similar among treatment groups (12 control, 17 noise).
Noise playback of artificially generated low-frequency traffic noise (filtered white noise in the range of 1–10 kHz with a decrease of 6·5 dB/kHz; Halfwerk & Slabbekoorn 2009) was carried out using full-range speakers (Peerless, 2·5 inch) connected to an mp3-player and battery-pack hidden under the leaf litter. We extended the normal nest box by removing the roof and adding a second box on top (made of the same material), inaccessible by the birds, but with a hole at the bottom, for both noise and control territories. In 2009, we added the second box and started the treatment during the final stage of nest building. In 2010, we added the second box to all nest boxes in a pair's territory and started the treatment at the beginning of nest building.
We inserted a speaker at a height of 15 cm within this second box to allow playback of noise mimicking conditions inside as if the nest box was situated at 50 m from a major highway (Halfwerk et al. 2011b). Noise exposure started when the nest was nearly finished and the sound level was increased over 2 days time in two steps to let the female gradually habituate. On the first day of the experiment, we increased the noise level to ~56 dB (SPL, A-weighted, measured at the position of the female with a Cesva SC-30 sound analyser). The second day, we increased the noise to ~65 dB, which was the level at which we exposed females throughout the rest of the experiment. Noise was played continuously day and night in 30 min loops with 10 s in between subsequent playback of the loop file with a ramp off to silence and a ramp on to full exposure again to avoid abrupt changes in noise level. Owing to high spring temperatures in both years, females started quickly with egg-laying (on average 1·8 ± 2·6 SD days after start of the experiment), which therefore coincided with the first day of full noise exposure.
Noise levels outside the nest box, recorded on a microphone positioned on the tree at the same height as, and within 50 cm of the nest box entrance, did not differ in the range of 1–10 kHz (anova; n = 29; F1,28 = 0·36; P = 0·85), which fully covers the great tit song range, and the noise was not audible to a human observer at 10 m from the nest box, which corresponds to the average singing distance of great tits in our area. The Leiden Committee for animal experimentation approved the study under number 08073.
We used SongMeters (16-bit, 24-kHz sample rate; Wildlife Acoustics Inc., Concord, MA, USA) to automatically record male and female behaviour. In 2009, we recorded behaviour at 20 nest boxes (11 control and nine noise) and we complemented the set in 2010 to a total of 29 (16 control, 13 noise). A microphone placed inside the nest box was used to record female calls and male song signal-to-noise ratio (with a fixed gain of +24 dB), while the other microphone outside recorded only the male's dawn song (fixed gain +42 dB). Recording microphones were also used to assess time of female emergence by the sounds of her claws on the nest box when taking off.
In 2009, two human observers made simultaneous focal observations at a different set of nest boxes without recorders (n = 22; 11 control, 11 noise) from both treatment groups to score the position of the male song post. Observations were carried out after 3 days of full noise exposure (which was 5 days after the start of the experiment). The observers were switched between treatments every other day to correct for inter-observer differences. The observers noted the song perch at 1-min intervals to determine the nearest song post. After the dawn chorus, the horizontal distance to the nest box was measured with a yard stick, and the vertical distance was estimated to the nearest metre to get a Euclidean distance measure to the nearest song post.
We scored female behaviour using the automatic dawn chorus recordings, including the time of nest box emergence, call rate and response latency (Fig. 1). Inter-individual female call variation is high and we therefore selected only the first and second call bout from a recording for the latency analysis. We measured the time (in ms) between start of female calling and start of male song or call bout (Fig. 1).
Male song behaviour was analysed by assessing the beginning of the dawn chorus and by identifying the different song types sung by the male until the female emerged. We estimated for each song type the proportion of time it was sung on a particular morning and selected for each song type the longest bout for further analysis. From each song-type bout, we selected two strophes from the start, mid and end of the bout. We determined the peak frequencies, as well as the onset and offset times of each note within a strophe in the program Luscinia (Fig. 2; Lachlan 2007) and averaged the measurements over the lowest notes (hereafter low-note) and loudest notes (hereafter peak-note) for each song-type strophe.
We band-pass-filtered each note (150 Hz above and below peak frequency of the note, which corresponds to the critical bandwidth of the great tit; Fig. 2; Langemann, Gauger & Klump 1998) and calculated the root-mean-square (RMS) value in matlab (the Mathworks, Nattick, MA, USA). We selected a noise sample of similar length after the song-type strophe and used the same band-pass filter settings to calculate the RMS value of the noise for each individual note. RMS values of notes and noise were transformed to a dB-scale and adjusted according to microphone gain. Noise amplitude was subtracted from note amplitude dB(note − dB(noise) to get signal-to-noise ratios for both low-note and peak-note (Pohl et al. 2012). In addition, we determined maximum song amplitude (loudest song type based on dB-values, Fig. 1). Song frequency and signal-to noise ratio measurements were averaged over song type, adjusted for the percentage of time sung. For the signal-to-noise measurements, we used recordings made inside the nest box; and for the frequency and song amplitude measurements, we used recordings made outside the nest box.
We analysed male and female acoustic behaviour on the first day (Noise Day 1) of full noise exposure (which was the third day of the experiment for the control treatment group) and compared this with measurements taken 2 days later (Noise Day 3). When a female had not started calling by the third morning of the experiment, we selected the first morning of calling as Noise Day 1. Additionally, we analysed female emergence times and call rates, as well as male song frequency and signal-to-noise ratios on day 1, 4 and 7 from the start of laying as these variables have been shown to covary strongly with egg-laying phase (Halfwerk et al. 2011a). Male song perch was only analysed on Noise day 3.
An impact of continuous noise exposure on male and female behaviour was tested using full factorial generalized linear mixed models (GLMM, spss 17·0, Armonk, NY, USA), with loglink-function for response latency (log-transformed) and call rate. Treatment and noise day or treatment and egg-day was included as fixed effects and site, year and date as random effects. Song post distance (log-transformed) was compared in a linear mixed model (LMM) with treatment as fixed factor and site and date as random factors.
Noise levels at the position of the female inside the nest box differed substantially between treatment groups (noise = 67·7 ± 1·8 SD, control = 36·9 ± 3·2 SD dB SPL, A-weighted), but the majority of spectral energy of the experimental noise was largely outside the frequency range of great tit song (see Fig. 2 for an example of a song under both noise and control conditions). As a result, noise levels differed more subtly in the low-note frequency range (3·74 ± 0·30 SD kHz), by 5·5 dB (anova; F1,28 = 10·1; P = 0·004; see also Fig. 2) and noise levels in the peak-note frequency range (4·29 ± 0·28 SD kHz) differed non-significantly by 3·3 dB (anova; F1,28 = 3·09; P = 0·09).
Female response latencies to songs or calls of their social male increased on the first morning of full noise exposure (Fig. 3a), but differences with females in control boxes disappeared within 2 days (GLMM; interaction day/treatment: n = 29; d.f. = 1; χ2 = 10·2; P = 0·001; Fig. 3a). Noise exposure had no effect on the moment of female emergence or female call rate (all P > 0·3).
Male maximum song amplitude recorded at the position of the nest box showed an opposite pattern (GLMM: interaction day/treatment: n = 29; d.f. = 1; χ2 = 10·2; P = 0·001; Fig. 3b), with amplitudes only differing between treatments after 3 days of full noise exposure (Fig. 3b). The interaction effect was mainly due to a decrease in amplitude in the control group (15 of 16 males decreased in recorded amplitudes), whereas in the noise group the amplitude either increased (four of 13), decreased (five of 13) or remained the same (four of 13, change of less than 1 dB). The noise treatment had no effect on the start of male dawn singing or the low-note or peak-note frequency (all P > 0·6).
The difference in recorded amplitude levels at the position of the nest boxes were related to song posts occupancy found in a different subset of males. Song post use after 3 days of exposure (on average on egg-day 3·6 ± 1·1 SD) differed between treatments, with males in the noise group singing from perches that were on average half the distance closer to the nest box compared to males in the control group (LMM: n = 22; F1,20 = 10·12; P = 0·005; Fig. 4).
The overall decrease in amplitude between noise days suggests that males moved away from the nest box, which can also explain why signal-to-noise ratios generally decreased with egg-laying (Fig. 5). The signal-to-noise ratios differed between noise and control treatment groups for the low-notes (GLMM; n = 29; d.f. = 1; χ2 = 4·07; P = 0·044), but these differences disappeared at later stages in the laying phase (Fig. 5a). We did not find significant differences in signal-to-noise ratios between treatment groups for the peak-notes. (χ2 = 1·70; P = 0·19; Fig. 5b).
We examined the role of female feedback on noise-dependent male song behaviour during the great tit dawn chorus ritual. We exposed females to artificial traffic noise inside their nest box, while leaving the singing male outside unaffected. We found females to delay their calling response in the noise treatment, which was related to increased song masking levels. Males in the noise treatment, not directly exposed, were found vocalizing from closer song perches after several days of full noise exposure. We also recorded male song amplitude at the position of the nest box and found no difference in amplitude on the first day of noise exposure. However, after 3 days of exposure, birds in the noise treatment group were recorded with higher amplitudes at the position of the nest box. The difference in song post use and the song amplitude change between noise and control groups were related to restored signal-to-noise ratios and restored female response behaviour.
Effects of laying stage
We aimed at exposing great tits during the peak in intra-pair communication, which coincides with the start of egg-laying. However, by doing so, we introduced unwanted variation in both male song behaviour and female response behaviour. Female calling behaviour in response to the male singing outside the nest box rapidly decreases progressively with egg-laying and almost completely disappears 5 days after the start of laying (Halfwerk et al. 2011a). This explains why females in the control treatment group increased response latencies to male song between experimental day 1 and 3. Importantly, females from the noise treatment group decreased call latencies, and consequently responded similarly compared to females from the control group after 3 days of exposure. Males likewise change their song behaviour in relation to egg-laying (Halfwerk et al. 2011a), spending less time near the nest box of their social mate and singing from a song post further away, just a couple of days after laying has started. This explains why all males in the control group were recorded with lower song amplitudes between experimental day 1 and 3, and why males from the noise group were on average recorded with similar amplitudes between days. Importantly, song amplitude did not differ on the first day of exposure between treatment groups, whereas 2 days later amplitudes differed by ~6 dB, which corresponds to halving the distance between song post and nest box, as was observed at a different set of nest boxes.
Internal or external feedback mechanisms
We found males to sing at closer distance in the noise condition, thereby restoring signal-to-noise ratios, even though males did not receive exposure directly. This strongly suggests that males relied on external feedback from the females to adjust their signalling behaviour appropriately. The most likely cue that may have triggered males in the noise treatment to move or remain closer to the nest box than control males is the delay in response behaviour that we found for their females. We did not find females to change call rate or emergence behaviour. Alternatively, males may have relied on a visual cue, provided by subtle movements of females, for instance at the nest box entrance. A similar social feedback mechanism was found in brown cowbirds, for which it was shown that selective female response tendencies played a determinant role in shaping male songs (King, West & Goldstein 2005).
Short-term noise-dependent signalling strategies have been proposed to be the result of an internal or external feedback mechanism (Halfwerk & Slabbekoorn 2009). Males can use direct auditory feedback from their own perception of noise or signal-to-noise level (internal) and change singing behaviour accordingly or use indirect social feedback (external) from conspecific receivers, such as females or neighbours, to adjust their songs in response to changing noise conditions (Halfwerk & Slabbekoorn 2009). Our study suggests that noise-dependent spatial song behaviour is (at least partly) driven by an external mechanism in great tits, although it remains to be tested how directly exposed males will respond in the absence of noise-dependent female feedback.
Most knowledge on noise-dependent feedback mechanisms comes from studies on amplitude regulation, which is generally presumed to reflect an internal mechanism, known as the Lombard effect (Brumm & Slabbekoorn 2005). The Lombard effect specifically refers to an involuntarily control of amplitude in response to noise (Pick et al. 1989; Lombard 1911), but animals can also adjust signal amplitude outside the context of noise, as male birds have been shown to sing louder when their mates are further away (Brumm & Slater 2006). Males may have an internal mechanism that matches information on receiver-distance to song amplitude, but it seems more likely that males in the experiment of Brumm & Slater relied on an external feedback mechanism in the form of (a lack of) female response. Although noise-dependent amplitude regulation has been shown to occur in many animals in the absence of a receiver (Potash 1972; e.g. Manabe, Sadr & Dooling 1998), it does not prove that individuals are not affected by external cues while fine-tuning vocal amplitude as well. Interestingly, this latter possibility has never been adequately tested and our study shows how distinguishing between the two types of feedback mechanisms can be more complex than we would expect at first sight.
Exposing only senders to noise, or only receivers as in our experimental setup, seems a useful tool to study whether birds adopt an internal or external mechanism when singing higher, louder or faster in response to noise during intra-pair communication. Males responded in our experiment by moving closer to females, a simple yet effective way to increase signal-to-noise ratios at the receiver's side (Slabbekoorn 2004). Males could also have produced songs at higher amplitudes, or changed the radiation pattern of their songs by aiming their songs at the nest box (Brumm, Robertson & Nemeth 2011), but it is likely that the theoretical increase of ~6 dB, related to half the distance between song post and nest box, was sufficient to overcome the 5·5 dB masking impact on low-notes. We do know that male great tits did not change the frequency of their songs in the noise treatment, despite the fact that female great tits can provide frequency-dependent feedback to males in noise (Halfwerk et al. 2011a). The lack of frequency change in the present study suggests that noise-dependent frequency use in great tits is not driven by an external feedback, or at least not during male–female communication.
Costs of communication in noise
We found an impact of anthropogenic noise on intra-pair communication and although birds seemed to restore communication rapidly, such an impact may still have negative fitness consequences as suggested by a previous study, which revealed a correlation between high noise levels and reduced clutch sizes and lower numbers of fledglings (Halfwerk et al. 2011b). The masking of the acoustic interaction can affect synchronization of reproductive behaviour between pair members and can have a negative impact on the pair bond between males and females (Swaddle & Page 2007). The negative impact of masking may be crucial even for short periods of exposure, especially when they coincide with the peak of female fertility, as it did in our experiment. Such impact will have crucial fitness consequences as it may reduce the pair's reproductive investment, for instance during food provisioning to the chicks (Johnstone 2011), and may explain our previous findings of noise-related reduction in fledgling success (Halfwerk et al. 2011b). A negative impact on male–female communication may also lead to less investment by females in their clutch, which could explain the relationship between clutch size and noise levels that was previously reported (Halfwerk et al. 2011b).
We also found male great tits to change song perches during noise treatment. Males occupied songs posts that were closer to the nest box in both horizontal and vertical direction. As a result, males could have been singing from more exposed branches and suffer higher predation risks (Moller, Nielsen & Garamszegi 2006). Furthermore, a change of song post can affect a song's spatial ecology (Naguib et al. 2011; Slabbekoorn 2004). During the peak in female fertility, male great tit dawn song is typically delivered from a song post that is close to the roosting cavity of the social mate (Mace 1987). However, males also interact with neighbouring males around dawn during so-called song-type matching contests (Krebs, Ashcroft & Webber 1978). Under normal circumstances, a trade-off related to signal detection for different types of receiver determines optimal signal design, including song post choice (Naguib et al. 2008, 2011). A noise-dependent change in song post affects this trade-off and especially a reduction in song post height can have a dramatic effect on long-range transmission for male–male communication and territory defence (Mathevon et al., 1996).
Anthropogenic noise can additionally affect an animal's cognitive demands, either through distraction (Chan et al. 2011b) or through increased sensory processing (Bee et al., 2008). Consequently, anthropogenic noise has the potential to shift allocation of cognitive capacity with crucial fitness consequences. For instance, shifting attention away or towards predator risk assessment immediate affects survival probabilities as many species face a trade-off between vigilance and foraging behaviour (Bernays & Funk 1999) and anthropogenic noise has been found to reduce predator detection (Chan et al. 2011b) and to reduce feeding efficiency as a result of increased vigilance (Quinn et al. 2006). Interestingly, as sensory processing is often multimodal (Driver & Spence 1998), it is very likely that acoustic noise affects behaviours that depend on other sensory modalities as well (Chan & Blumstein 2011a).
We have experimentally shown that females can provide noise-dependent acoustic feedback on male song performance during intra-pair communication, which may have caused males to decrease singing distance and increase signal-to-noise ratios. Males did not adjust song frequency in response to the feedback from females, but the spatial adjustment of song perch may have been already sufficient to mitigate song masking by noise and restore critical communication conditions. Our findings suggests that great tits have a suite of strategies to compensate for detrimental noise impact, which may explain why this species survives well in the urban habitat, despite some loss in reproductive success (Halfwerk et al. 2011b) and limitations in terms of information transfer associated with signal adjustment (Halfwerk et al. 2011a). The experimental approach of testing noisy urban conditions on just the receiver side strongly suggest an active role for female birds in steering male communicative behaviour and reveals the critical role of spatial ecology in shaping animal interactions.
We thank C. ten Cate, K. Riebel, J. Podos & H. Brumm for their feedback on previous aspects of the manuscript. We thank Staatsbosbeheer and Natuurmonumenten for allowing us to work in their forest reserves. We are very grateful to C. Both for letting us work at his research sites and for his valuable feedback during field work. This study was funded through a NWO grant (no. 817·01·003) to H.S. The authors report no conflict of interest.