Abstract
The effect on anthropogenic noise on acoustic signals is widely studied although the extent to which this is due to short-term flexibility or long-term adaptation is unclear. Anthropogenic noise may disrupt signals used to mediate aggressive interactions, leading to more physical aggression between opponents. One solution to this problem is to switch signaling effort to a less noisy modality (e.g. the visual modality). In the present study we ask whether urban and rural European robins (Erithacus rubecula) use multi-modal signals flexibly in response to anthropogenic noise during territorial defense. We predicted that during simulated intrusions with experimental noise, European robins would increase their physical aggression, respond with more visual threat displays and sing fewer songs, i.e. show a multi-modal shift. We also predicted that urban birds, living in noisier habitats, would be generally more aggressive than rural birds. The results showed that urban birds were more aggressive than rural robins, but an increase in aggression with experimental noise was seen only in the rural birds. Urban birds also used visual signals more often than rural birds. However, birds did not use visual signals more in experimental noise. Instead, both urban and rural robins sang at higher rates under noise conditions, thus increasing signaling effort in the noisy modality. These results point to a complex role of immediate plasticity and longer-term processes in affecting communication during aggressive interactions under anthropogenic noise.
Significance Statement Human activity has an enormous effect on wildlife, including on their social behavior. Animals living in urban areas often tend to be more aggressive than those living in rural areas, which may be due to urban acoustic noise making communication between individuals more difficult. In a study with a common songbird, the European robin, we investigated the role of urban acoustic noise in aggression and territorial communication. Urban robins were more aggressive than rural robins, and additional noise in the territory increased aggression in rural but not urban robins. Robins increased their singing effort but did not increase visual signals in acoustic noise. These results suggest that noise can indeed make animals behave more aggressively although the effect may depend on how noisy it is already. These results further our understanding of how human-made noise changes animal communication and social behavior.
Introduction
Urban habitats are polluted with anthropogenic noise, often in multiple modalities, which creates challenges for urban-living wildlife (Brumm & Slabbekoorn, 2005). Many species rely heavily on signals for communication in contexts such as mate attraction and territorial defense, and noise from vehicles, buildings and other human activities often interferes with these signals (Francis et al., 2009; Halfwerk & Slabbekoorn, 2015; Lee & Thornton, 2021). A well-studied example of the effect of anthropogenic noise on communication is vocal signaling in urban birds: in response to anthropogenic noise commonly found in cities, many species of birds may increase repetition rates, amplitude or frequency characteristics of their acoustic signals (Brumm, 2004; Gil & Brumm, 2014; Slabbekoorn & den Boer-Visser, 2006; Wood & Yezerinac, 2006).
Urban living also leads to increased aggressiveness of individuals in urban habitats compared to the rural habitats (Scales et al., 2011). The reasons for increased aggression in urban habitat are not yet fully understood. It may result from several factors, including selection for bolder individuals (Evans et al., 2010), increased food resources (Foltz et al., 2015), increased exposure to harmful chemicals such as lead (McClelland et al., 2019) and less stable social environment due to high rates of territory turnover in urban habitats (Davis et al., 2013). Anthropogenic noise in urban habitats may also be responsible for increased aggression.
Animals often use signals in aggressive interactions (e.g. during territory defense) to resolve conflicts with opponents. Use of signals is often beneficial for both parties if they can avoid costly physical fights in this way (Maynard Smith & Price, 1973). Consequently, if signaling is prevented or the signals are rendered ineffective, individuals may need to resort to higher levels of physical aggression (Logue et al., 2010). Applied to urban habitats, this hypothesis suggests that the high levels of urban noise may render long-distance aggressive signals less effective, which in turn may lead to higher levels of aggression (Phillips & Derryberry, 2018). Consistent with this hypothesis, some studies found a positive correlation between ambient noise levels and aggressive behaviors (Akçay et al., 2020; Phillips & Derryberry, 2018; but see Kleist et al., 2016).
Signalers employ various strategies to overcome interference from anthropogenic noise. We focus here on the flexibility afforded by having signals in more than one modality (Bro-Jørgensen, 2010; Halfwerk & Slabbekoorn, 2015; Partan & Marler, 1999). In such cases, animals may shift their signals from the noisy modality to the less noisy modality to increase the likelihood that the message of the signals gets through to the receivers (Partan et al., 2010; Partan, 2017).
Few studies tested the multi-modal shift hypothesis in signals used in territorial interactions. In one study, Ríos-Chelén et al. (2015) found that male red-winged blackbirds (Agelaius phoeniceus) did not use more intense visual signaling in noisier territories although they modified their acoustic signals. Another study on song sparrows (Melospiza melodia) found that males in noisier urban habitats were both more aggressive and used proportionally more visual threat signals (wing waves) during territory defense compared to the males in rural habitats, consistent with a multi-modal shift (Akçay et al., 2019). In another experiment, however, individual song sparrows did not increase their visual signaling effort when experimentally presented with noise, suggesting that multi-modal shift was not due to immediate plasticity (Akçay & Beecher, 2019).
Here we investigate the responses of European robins (Erithacus rubecula) living in urban and rural habitats in Istanbul, Turkey, to simulated territorial intrusions with or without experimental noise playback. European robins have both visual and acoustic signals that are used in agonistic interactions (Lack, 1965). Previous studies have found that robins respond both to song and visual signals during territorial intrusions (Chantrey & Workman, 1984). Territory holders sing in response to the song of an intruder, while visual signals are used when the intruding male is within range of vision. Their most prominent visual signal is the neck display, which has been observed in response to the sight of a rival male’s red neck, or indeed even a ball of red feathers (Lack, 1965). Other visual territorial defense signals include wing flutters, pricking the tail up, and swaying, where the resident male moves his head from one side to the other (Lack, 1965).
There have been several studies on the singing and aggressive behaviors of robins in the presence of noise. A study by Mclaughlin and Kunc (2013) found that robins, after being lured by playback of a robin song from a speaker, tended to move away from the speaker when the speaker switched to playback of low-frequency noise mimicking typical traffic noise, particularly when the amplitude of noise was high (90 dB at 1m). In response increasing amplitude of noise, the robins sang shorter songs with fewer notes and increased the minimum frequency of their songs. In another study, experimental presentation of wind turbine noise during simulated territorial intrusions led to a decrease in low frequency elements in the songs, at the same time leading to an increase in flight rates (Zwart et al., 2016). Song rates did not significantly differ between the noise and no-noise conditions. Interestingly, fewer robins used visual threat postures under experimental presentation of noise compared to no noise, although the difference was not significant (Zwart et al. 2016).
The presence of both acoustic and visual signals in territorial defense makes the European robin a suitable candidate for testing the multimodal shift hypothesis but to our knowledge no previous study compared visual signaling between urban and rural robins. We predict that robins in urban habitats will exhibit higher levels of visual signaling. Additionally, if such a multi-modal shift is due to phenotypic plasticity, robins should increase their visual signaling under experimental noise. Finally, in accordance with earlier studies, we also expect to see a greater level of aggression from urban robins compared to rural robins. If increased aggression is due to individual plasticity in response to noise, we also expect higher levels of aggression from robins in response to experimental noise.
Methods
Study sites and species
We carried out playback experiments with male European robins that held territories in rural areas (forests around Sariyer, Istanbul, 41° 9’ 50.73971”N, 29° 0’ 32.25243”E) and urban parks and green areas in Sariyer, Istanbul, Turkey in April and May 2021 (urban: n=9; rural: n=12). Robin territories were detected by the presence of an already-singing male robin before the first playback or during recording sessions prior to playback. We determined a central location by observing the robin’s flights for about 5 minutes, although we did not attempt to map the entire territory. In all trials reported below, only a single bird responded to the playback (in one case, we aborted the trial when a second male came to within 10 m of the speaker). It was not possible to record data blind as our study involved observing focal individuals in the field and noise manipulation was audible to all observers.
Stimuli
Playback stimuli were generated on the software Syrinx (John Burt, Portland, OR) from male European robin songs recorded in March 2021 in four of the nine study sites. We generated stimuli tapes by extracting high quality songs from each recording and filtering out low frequency noise below 1000 Hz. We added a silent period after each song so that stimuli were presented at a rate of one song per seven seconds. The songs lasted on average (±SD) 2.36 (±0.49). We created one-minute stimuli (consisting of nine different songs) which were repeated three times to make up three-minute stimuli to be played during the trials. In total, we used 17 tapes created from the songs of 17 different robins. The stimulus played for each subject came from a robin whose territory was separated by at least one km from that of the subject’s territory. Each subject received the same stimulus in both trials.
We generated the experimental acoustic noise stimuli by filtering white noise (created with Audacity) with the average amplitude spectrum from a 1-minute recording (made with a Marantz PMD660 and ME66/K6 microphone) of traffic noise in Sariyer, Istanbul using the package seewave in R (Sueur et al., 2008).
As a visual stimulus, we used a 3-D printed bird model (dimensions, height: 8 cm, length: 12 cm, width: 4.5 cm) which was hand-painted to resemble an adult robin.
Experimental procedure and design
We started each trial by locating a European robin territory and observing the singing posts for about ∼5m. We then placed the robin model attached to a speaker (Anker Soundcore Bluetooth Speaker, Anker, Inc.) on a natural perch at the estimated center of the resident male’s territory, approximately 1.5 m above the ground. A second Bluetooth speaker (same model as above) was placed on the ground, face-up below the first, for noise playback. In the control condition, the second speaker was placed but not turned on, so the resident male received only song playback. In the noise condition, in addition to the song playback, traffic noise was played at 75 dB SPL at 1 m. The noise playback lasted for the entire duration of the song playback.
Each subject received two 3-minute trials, one with experimental noise and one without noise. The order of the conditions was counterbalanced, and the two trials were separated with a break period that lasted for approximately an hour. Two observers, about 10 m away from the experimental setup, recorded the songs and calls of the resident robin. The observers also narrated the trials, noting flights (any airborne movement by the bird), distance with each flight, and visual displays described in Table 1. We continued recording songs for 3 minutes after the end of each trial. Recordings were made on a Marantz PMD660 with a Sennheiser ME66/K6 microphone, or on a Zoom H5 handheld recorder with a Zoom SGH6 shotgun microphone.
Visual disp lays of European robins during territorial interactions
In 31 of the trials, the bird was already singing when we started the trial. For these birds, the 3-minute playback period started with the first song played. In 11 of the trials, where the subject was quiet when the playback started, the 3-minute playback period started with their first response (song or approach). The average duration of pretrial playback for these 11 trials was 64.9 seconds (SD = 40.8).
After each trial, we measured the ambient noise with a VLIKE VL6708 sound-level meter with the method described in (Brumm, 2004). We took eight measurements (two in each cardinal direction) within a minute period, which were then averaged. Noise levels were highly repeatable across trials (intra-class coefficient, r=0.96, p<0.00001; for three subjects, we only had noise measurements from a single trial).
Response variables and data analysis
We scanned recordings using the Syrinx software (John Burt, Portland, OR). All analysis was carried out in R. The number of songs and visual displays (neck displays, swaying, wing flutters) were extracted form sonograms created on Syrinx. We only analyzed song rates and durations, as overlapping stimulus and subject songs made it impossible to determine with certainty the note compositions for most songs. We also extracted from recordings the number of flights, closest approach to the model/speaker and proportion of time spent within 5m of the speaker. The latter three spatial variables were taken as aggressive behaviors. Because these variables were significantly correlated with each other, we carried out a principal component analysis (PCA) using the principal function in package psych (Revelle, 2021). The first component of PCA (PCA1) explained 61% of variance and was taken as our primary measure of aggression (see Table 2 for loading coefficients). Since only 8 subjects used any visual displays during the experimental period, we coded this variable as a binomial variable (visual signal present vs. absent).
Loading coefficients of the Principal Component Analysis.
We first checked whether urban territories had higher ambient noise with a linear mixed model (LMM) using habitat type (urban vs. rural) as the predictor variable and territory ID as the random variable. We also assessed whether noise levels were repeatable using the rptR package (Stoffel et al., 2017).
We then checked whether order of trials had a significant effect on aggression scores, song rates and visual signaling. The order of trials did not have a significant effect on song rate (LMM, coefficient = 0.51, SE = 0.78, p = 0.52) or aggression score (LMM, coefficient = -0.22, SE = 0.15, p = 0.16). However, there was a significant order effect on the incidence of visual displays (GLMM, estimate = -15.32, SE = 6.22, p = 0.01). Eight subjects used visual displays in the first trial, compared to two in the second trial (both of these subjects also used visual displays in the first trial). Because of this order effect, we only used the first trial for each subject when analyzing the presence or absence of visual displays.
We analyzed song rates and aggression scores with linear mixed models (LMM), using the lme fuction in the package nlme (Pinheiro et al., 2022). We took habitat type (urban vs. rural) and experimental condition (noise vs. control) as the predictor variables, and male ID as the random variable. We applied a generalized linear model (with log-link, using the “glm” function in base R) with visual displays as binomial response variable, and habitat and condition as predictor variables, using only the first trials for each subject. Since only two rural subjects used visual displays, we also carried out this analysis with the subset of only urban birds (see Supplementary Materials).
Results
Urban habitats had significantly higher levels of ambient noise than rural habitats (urban: M = 49.0, SD = 7.1; rural: M = 39.9, SD = 3.6) and noise measurements were highly repeatable between the two trials (intra-class correlation coefficient; r=0.96, standard error: 0.02; p < 0.0001).
Urban birds were significantly more aggressive than rural birds. There was no main effect of condition but there was a significant interaction effect of habitat and condition. To understand this interaction effect, we carried out separate LMMs with rural and urban birds with condition as predictor variable and territory ID as random variable. Rural birds were more aggressive under the experimental noise condition, whereas there was no effect of experimental noise on aggression in urban birds (Table 3).
Coefficients (SE) from the linear mixed models and the p-values from Wald t tests, examining the effect of habitat and experimental noise treatment. Statistically significant values are shown in bold type
Song rates did not differ significantly between urban and rural birds. Both urban and rural birds sang at higher rates under experimental noise. The interaction effect of habitat and experimental condition was not significant (Table 3). There was no effect of habitat and condition on song length (see Table 3).
Urban birds used visual threat displays in the first trials significantly more than their rural counterparts (GLM; X2= 9.75, p=0.0018; Figure 2). Experimental noise did not have a significant effect on the use of visual signals, although more birds (5 out of 5) used visual threat displays in no noise condition than in noise condition (1 out of 4 birds; X2= 3.22, p=0.07).
The relationship of aggression scores (A) and song rate (B) with habitat and condition. The boxes indicate interquartile ranges, the middle line indicates median, and whiskers indicate 95% confidence intervals. Dot connected by dotted lines represent data from individual subjects.
Proportions of first trials where the resident male used visual signals, grouped by habitat and treatment condition. The numbers at the bottom of each bar indicates the total number of subjects for each combination.
Discussion
The present study was designed to examine the role of noise in determining aggressiveness and aggressive signaling. We had predicted that urban robins, living in noisier territories will be more aggressive compared to rural robins in simulated territorial intrusions and experimental noise during simulated intrusions will change both their aggressive behaviors and signaling behaviors. Particularly, we expected that experimental acoustic noise should increase aggression during intrusions and lead to an increase in using signals in the visual modality. In line with our first prediction, we found that urban robins responded with significantly more aggressive behaviors (approach and flights) to simulated intrusions than rural robins. The effect of experimental noise on aggressive approach was dependent on the habitat: while experimental noise led to increased aggression in the (comparatively quiet) rural habitats, it had no effect in the noisy urban habitats. In both habitat types, experimental noise led to an increase in song rates, but no change was observed in song duration or the presence of visual signals which tended to be less common under experimental noise in the urban habitats.
Noise and aggressive behaviors in territory defense
Our results on the effect of habitat on aggression replicates earlier findings that urban-living birds are more aggressive than rural birds (Evans et al. 2010, Davies and Sewall, 2016, Phillips and Derryberry, 2018). We also extend previous findings by showing that the effect of experimental noise on aggression was dependent on habitat: urban males showed no further increases in aggression with experimental noise while rural males showed a significant increase in aggression. The increased aggression with experimental noise in rural habitats is consistent with the idea that urban noise has a causal role in increasing aggression. In contrast, for urban males that are already living in noisy territories, additional noise may not have as much as an effect as in rural habitats. Urban birds may also be more habituated to acute sources of noise (e.g. a car passing or idling next to their territory) than rural birds, although they did show a plastic response in their singing rate. Finally, urban birds may not be able to increase their already high levels of aggression in response to noise playback.
Only a small number of studies experimentally manipulated noise levels to examine a causal role of noise in increased aggression. These studies yielded mixed results. Grabarcyzk and Gill (2019) found that house wrens (Troglodytes aedon) males attacked the simulated intruder more frequently when playback was accompanied with experimental noise than when it wasn’t, consistent with the hypothesis that noise induces higher levels of aggression. Another study in song sparrows however, found no effect of experimental noise on aggression levels, measured as time spent within one meter of the speaker, or attacks (Akçay & Beecher, 2019). It is worth noting that in the latter study, the noise playback in that study started only when subjects approached to within five meters, which all subjects did within a short period of time (< 1 minute). Thus, lack of an effect in physical proximity may be due the fact that subjects already were close to the speaker when the noise playback started.
In another study, Zwart et al. (2016) found that European robins did not show a statistically significant increase in aggressive behaviors in response to experimental wind turbine noise during simulated intrusions, although some variables like flights did show a trend consistent with higher aggression with experimental noise. Given the fact that this study had a between-subject design and a relatively small sample size per treatment group (8 birds) the non-significant differences may be primarily due to lack of power rather than lack of an effect. Finally, a recent study by Reed et al (2021) in lazuli buntings (Passerina amoena) and spotted towhees (Pipilo maculatus) found that experimental presentation of natural noise (such as noise from a river, ocean surf or cicadas) at the landscape level led to slower detection of a simulated intruder and consequently weaker approach responses (see also Kleist et al. 2016).
Together these studies point to two apparently contradictory effects of noise on territorial aggression. On one hand, noise may make localization of the simulated intruder and perception of stimulus features more difficult, leading to slower or weaker approach behaviors (Kleist et al. 2016; Reed et al. 2012; Templeton et al. 2016). On the other hand, assuming the simulated intruder is located, noise may interfere with the signaling behaviors of subjects which may induce them to resort to higher physical aggression (e.g. Grabarcyzk and Gill, 2019). The differences in the findings may be due in part to differences experimental designs, particulary with respect to the presentation of the noise stimulus (e.g. type of noise, location of noise relative to the conspecific stimulus etc.). Therefore, to investigate the causal effect of noise on territorial aggression, it would be useful to have a more approach systematic approach to varying noise stimulus presentation.
Change in multi-modal signals with noise
We also found that European robins changed their signaling behaviors in response to noise. In the acoustic modality, robins increased their song rate in both habitat types, without changing song length. Increasing song rate may mean that the robins are attempting to increase the serial redundancy of their signals, as has been found for chaffinches (Fringella coelebs) in a study by Brumm and Slater (2006). It is worth noting however that the chaffinches sing with eventual variety, repeating the same song type multiple times before switching to a new song type. Thus, the finding in Brumm and Slater that the males sang longer bouts of the same song type in noisier habitats amounts to increasing serial redundancy. Robins, in contrast, are immediate variety singers, switching song types with each song. Whether increasing song rate also increases serial redundancy of the signal is therefore unclear and depends on whether the song types are largely interchangeable.
Other studies looking at the effect of noise on singing rate found mixed results. A study on red-winged blackbirds had previously found lower song rate in response to noise (Ríos-Chelén et al., 2015), while a more recent study in chaffinches found no difference in song rates between urban and rural birds. In studies that simulated territorial intrusions with or without experimental noise, the results were also mixed: experimental noise did not lead to changes in song rate in house wrens or song sparrows during a simulated intrusion (Akçay & Beecher, 2019; Grabarczyk & Gill, 2019).
One of our main research questions was whether robins would switch to preferentially signaling in the visual modality in response to acoustic noise. We found that urban birds were more likely to use visual signals than rural birds, with six out of nine urban birds using visual threat displays compared with two out of 12 rural birds. We did not, however, detect a shift from the acoustic signals to the visual signals under experimental noise. One caveat to this conclusion is the fact that most visual displaying was confined to the first trials which prevented us from testing the individual phenotypic plasticity. Nevertheless, limiting our analysis to the first trials only, we found that visual displays were more prevalent without the experimental noise than with it, albeit the difference did not reach significance. This finding is similar to that of Zwart et al (2016), where slightly lower rates of visual signaling were observed in noise. Thus, while increased visual signaling in urban birds is compatible for a multi-modal shift, such a multi-modal shift doesn’t seem to be due to the plastic responses to acoustic noise.
These findings are similar to the earlier studies in song sparrows. Like in the robins, urban song sparrows use proportionally more wing waves (a visual threat signal) as part of their aggressive signaling during a simulated intrusion, compatible with a multi-modal shift (Akçay et al., 2019). The multi-modal shift, however, does not appear to be due to individual phenotypic plasticity, as song sparrows did not increase their visual signaling effort during experimental presentation of acoustic noise (Akçay & Beecher, 2019). Thus, an increase in visual signaling may be due to either long-term selection for different signaling strategies or plasticity due the chronic (rather than acute) effect of noise. These hypotheses need to be tested with further studies, ideally with marked individuals that can be followed for extended periods of time during exposure to chronic noise.
Conclusion
In summary, we found that urban European robins were more aggressive than rural ones and the effect of noise on the change in aggressive behaviors depended on the habitat: while urban birds did not behave more aggressively, rural birds did so under experimental presentation of noise. Both urban and rural birds increased their song rate (but not song length) under experimental noise and urban birds used visual threat displays more frequently than rural birds. These findings extend previous studies on the change in aggressive behaviors and signaling with noise. Going forward experimental analyses of the effect of noise in signaling and aggressive behaviors will prove valuable in determining the role of plasticity in how urban-living species adapt to life in a noisy world.
Disclosure of potential conflicts of interest
The authors declare that they have no conflicts of interest.
Research involving Human Participants and/or Animals
All procedures used in this study follow the ASAB/ABS guidelines for the treatment of animals in behavioral research and teaching. Subjects were not captured or handled before, during or after any of the trials. Time spent within a territory did not exceed 15 minutes per trial, and 30 minutes per day.
Informed consent
The research did not involve human participants.
Data availability statement
The raw data and the R-code to reproduce the analyses reported in the manuscript is available as a supplementary material and will be deposited at a public repository before the manuscript is published.
Acknowledgements
This study was funded by the British Ornithological Union Small Ornithological Research Grant to ÇÖ.
