Flexibility of Territorial Aggression in Urban and Rural Chaffinches

ABSTRACT

Rapid environmental change due to urbanization poses novel challenges to animals. Behavioral change and individual plasticity are generally hypothesized to be the key to adapting to these challenges. One commonly observed behavioral change is higher observed aggression levels in urban animals, perhaps because anthropogenic noise disrupts effective acoustic communication during conflicts, leading to greater use of physical aggression. We investigated the hypothesis that urban noise drives aggression by performing repeated simulated territorial intrusion experiments on rural and urban chaffinches ( Fringilla coelebs ). We expected urban chaffinches to be more aggressive, change their aggression levels more between trials, and for aggression to increase with noise levels, irrespective of the habitat. We found that while aggression did not differ between habitats in the initial trial, rural chaffinches decreased their aggression level in the second trial and thus were less aggressive than the urban chaffinches, which did not change their response. That is, urban birds were less flexible in responding to an intruder than rural birds, contrary to previous findings in other songbirds. Moreover, aggression levels correlated positively with ambient noise levels. Given our small sample size and lack of spatial replicates, our results should be interpreted with caution. Nevertheless, as a lack of flexibility in aggression is potentially costly, our results highlight the importance of studying the plasticity in aggressive behavior in human‐impacted landscapes.

Behavioral flexibility of urban animals may allow them to adapt to rapid environmental change. We tested whether urban chaffinches are more flexible in territorial aggression than their rural counterparts and whether changes in territorial aggression are related to anthropogenic noise. Contrary to our expectations, we found that urban chaffinches were less flexible in aggression; however, aggression increased with increasing ambient noise levels, irrespective of the habitat.

Urbanization presents novel and varied evolutionary challenges to animals, including pollution, habitat fragmentation, and climate change (Grimm et al. 2008). Populations facing these challenges in cities may undergo rapid evolution as a result of adaptive and nonadaptive processes (Johnson and Munshi‐South 2017; Sih et al. 2011). These urban dwellers often shift their behavioral phenotypes to survive in the human‐impacted environments, making behavioral change a precursor to urban evolution (Caspi et al. 2022; Sol et al. 2013). Indeed, differences in behavioral traits between rural and urban populations of animals are well documented (Gil and Brumm 2014; Lowry et al. 2013; Ritzel and Gallo 2020). Compared to their rural counterparts, urban‐dwelling individuals of a species are shown to be bolder (Samia et al. 2015) and less aversive to novelty (Griffin et al. 2017; Tryjanowski et al. 2016), to breed earlier (Capilla‐Lasheras et al. 2022), and to differ in their signaling behaviors (Slabbekoorn and Peet 2003). One important phenotypic change associated with urbanization is that urban animals tend to be more aggressive compared to rural animals (Abolins‐Abols et al. 2016; Colombelli‐Négrel et al. 2023; Davies and Sewall 2016; Diniz and Duca 2021; Fokidis et al. 2011; Foltz et al. 2015; Hardman and Dalesman 2018; Önsal et al. 2022; but see Hasegawa et al. 2014; Hurtado and Mabry 2017).

It is unclear why urban animals tend to be more aggressive than their rural counterparts, although several explanations have been proposed. These include increased food availability (Foltz et al. 2015), behavioral syndromes (Evans et al. 2010), metal pollution (McClelland et al. 2019), and noise pollution (Phillips and Derryberry 2018). Among these factors, the latter has attracted significant research attention (Akçay, Porsuk, et al. 2020; Diniz and Duca 2021; Grabarczyk and Gill 2019; Önsal et al. 2022). Anthropogenic noise may decrease the detectability of signals and discrimination of features relevant to opponent assessment in territorial disputes (Lohr et al. 2003; De Kort et al. 2024). Thus, it is hypothesized that anthropogenic noise contributes to increased aggression because signals become less efficient due to masking, which prevents the resolution of disputes through acoustic communication, ultimately leading to physical fights (Akçay, Beck, et al. 2020).

Behavioral flexibility can allow individuals to cope with rapid environmental changes and so is hypothesized to be a key component of urban adaptation (Caspi et al. 2022; Sih et al. 2011). In the context of the challenge of dealing with urban noise, experimental studies showed that individuals, at least sometimes, flexibly increase their physical aggressive behavior when presented with noise (De Kort et al. 2024; Grabarczyk and Gill 2019; Hohl et al. 2025; Önsal et al. 2022). Another good example of such flexibility is a study on dark‐eyed juncos ( Junco hyemalis ) living in urban habitats that took advantage of periods of decreased human activity during the COVID‐19 pandemic (the “anthropause”; Rutz et al. 2020). In this study, urban juncos showed decreased aggression towards conspecifics during the lockdown periods than before, suggesting that urban birds can flexibly adjust their behavior in response to human activity (Walters et al. 2022).

Behavioral flexibility implies increased within‐individual variation in behaviors, and conversely lower repeatability of individual differences (i.e., proportion of total variance attributable among individual variance). Therefore, if urban birds show more behavioral flexibility in response to noise compared to rural birds, they should also show lower repeatability of behaviors such as aggression. A study with great tits ( Parus major ) supported this idea. Hardman and Dalesman (2018) found significant repeatability of all five proxies of aggression for rural birds, but for only two of them for urban birds, although their results were inconclusive, as there were no differences in repeatability estimates for each variable between urban and rural populations.

Here, we studied differences in the intensity and repeatability of territorial aggression between urban and rural populations of common chaffinch ( Fringilla coelebs ), evaluating the anthropogenic noise hypothesis. Using song playback, we simulated conspecific intrusions inside male territories on two consecutive days. First, we predicted that territorial aggression would be higher in urban chaffinches, consistent with previous studies comparing urban and rural animal populations. Because we hypothesized that anthropogenic noise can disrupt acoustic communication, we also predicted that ambient noise levels would correlate positively with aggression irrespective of habitat type. Moreover, we reasoned that mostly aggressive individuals, and those that can flexibly adjust to fluctuating environmental conditions, would be inhabiting urban habitats. Hence, we predicted that urban chaffinches would have lower among‐individual variance, higher within‐individual variance, and lower repeatability with respect to their aggressiveness.

1. Methods

1.1. Study Site and Species

We chose chaffinches for investigating differences with regard to territorial aggression, as their territorial behaviors are well studied and they are frequently found in both urban and rural habitats (Brumm and Ritschard 2011; Marler 1956; Slater 1981). Chaffinches are territorial only during the breeding season; territories are established by males, which are about 0.7 ha (Marler 1956). Males defend their territories from intrusions. In the case of playback experiments, defense typically involves flights around the speaker but not singing, which suggests song is used as a keep‐out signal but not in active defense (Slater 1981).

We performed simulated territory intrusion experiments on 23 male chaffinches inhabiting one urban site (n = 11) and two rural sites (n = 12) in Sarıyer, İstanbul, Turkey. Ambient noise in the urban site was significantly higher than in the rural sites, making the locations suitable for our investigation (Yelimlieş et al. 2023; for the territories in the present study: rural mean ± SE = 38.89 ± 0.38, urban mean ± SE = 47.89 ± 1.20). In our study, each male was tested twice; however, none of the subjects were ringed in the study, so we assumed the territory holders did not change between the two consecutive trials, which were separated by 24 h. We avoided including neighboring males in our sample and did not test males located closer than 200 m to each other in order to ensure that each male is included in our sample only once. All trials were conducted 11 to 26 May 2021 between 6 and 11 a.m., except for two rural individuals. These two individuals were tested between 4 and 5 p.m. when the chaffinch singing activity was comparable to morning levels.

1.2. Playback Stimuli

For the simulated territory intrusions, we used songs of chaffinches recorded between March and May 2019 in the study sites specified above, along with songs of the subjects in this study. Stimulus songs were recorded using a Marantz PMD660 or 661 recorder with a Sennheiser ME66/K6 shotgun microphone. From our recordings, we selected 22 high‐quality songs from 22 different male chaffinches. We removed low‐frequency noise from each song recording using a high‐pass filter (threshold 1000 Hz) in Raven Pro 1.6.1 (K. Lisa Yang Center for Conservation Bioacoustics 2019). Then we added silence at the end of each song to create 10‐s‐long stimulus files. The mean ± SD song duration was 2.27 ± 0.27 s. The same stimulus file was broadcast in both trials of each subject to prevent the influences of stimulus‐related confounds.

1.3. Experimental Procedure

Chaffinches sing from multiple locations to post their territory boundaries. We observed the singing posts of chaffinches in our sites to determine a central location in the territory to carry out the playbacks. Prior to song playback, we placed a wireless speaker (Anker SoundCore, Anker Inc.) about 2 m above the ground level, and within the boundaries of the territory of the resident male for playing the conspecific stimulus. After positioning ourselves ~10 m away from the speaker and confirming the presence of the resident male within the territory (about 10–15 m from the speaker), we played the conspecific song stimulus continuously six times (1 min playback in total). During this playback period, we narrated the flights of the focal male, noting the distance from the speaker. We used either a Marantz PMD660 recorder with a Sennheiser ME66/K6 shotgun microphone or a Zoom H5 recorder with a Zoom SGH‐6 shotgun microphone to record our narrations. After each trial, we measured the average ambient noise level in dB in the target male's territory using a sound level meter (VLIKE VL6708, VLIKE Inc., A‐weighted, fast). To get an average, we obtained eight noise levels measuring twice from four directions that are perpendicular to each other from a single location in the territory (Brumm 2004).

1.4. Data Analysis

Using Raven Pro 1.6.1, we annotated our narrations of the trial and coded the proportion of time spent within 1 and 5 m of the speaker, number of flights, and closest approach to the speaker for each trial. These variables are used as proxies of aggression levels in songbirds (Brumm and Ritschard 2011), and have been shown to predict attacking on a taxidermic model (Akçay et al. 2013). We then performed a principal component analysis (PCA) to obtain a composite score for aggression. We evaluated the significance of PCA and principal components using the PCAtest package with 1000 permutations and bootstrap replication (Camargo 2022). Permutation tests showed that only PC1 was meaningful, which accounted for 71.5% (95% CI = 61.6–79.5) of the total variation in the data, and had an eigenvalue of 2.86. All variables had significant loadings on PC1 (number of flights = 0.47, closest approach = −0.55, time spent within 5 m = 0.47, and 1 m = 0.50). So, we ran the PCA using the function prcomp from the stats package with centered and scaled variables and took PC1 scores as the aggression scores.

We then modeled the relationship between aggression scores and habitat type via a linear mixed model with Gaussian error distribution using the function lmer from the package lme4 (Bates et al. 2015). As exploratory visualizations showed a discrepancy between first and second trials depending on habitat, the model included habitat, trial order, and their interaction effect as fixed factors and male ID as the random factor. We estimated the repeatability for ambient noise and aggressiveness, as well as between and within individual variance for aggressiveness, separately for each habitat, using the package rptR (Stoffel et al. 2017), with a Gaussian error distribution and calculated 84% confidence intervals using bootstrapping with 1000 iterations. We chose 84% because non‐overlapping 84% confidence intervals can be used as a proxy of the difference in estimates being different than zero at α = 0.05 (Payton et al. 2003).

Lastly, we modeled the relationship between aggression scores and ambient noise levels, while controlling for trial order and its interaction with habitat, using another linear mixed model with a Gaussian error distribution. The model did not include an interaction term between habitat and ambient noise because ambient noise levels did not overlap between habitats (see Figure 2). The model included male ID as a random effect. We built a second model for testing the influence of ambient noise on aggression because a subset of males (4 rural and 1 urban) had missing ambient noise measurements for one trial from their territory and these trials were excluded from this analysis, reducing the sample size to 41 trials from 23 males.

FIGURE 2.

Scatterplot showing the positive correlation between aggression score (PC1) in male chaffinches and the ambient noise level (dBA) in their rural or urban territory. There was no such relationship when the urban subset was inspected alone. The points show each trial, and the lines show predicted values for urban and rural sites, controlling for other fixed and random effects in the model, and the shadings show 95% CI.

Given the almost complete separation of noise levels between habitats (see Section 2), we then repeated this analysis with the urban subset (N = 11 males, 18 trials) with trial order as the fixed effect and male ID as the random effect. We reasoned that, since noise levels vary more in the urban habitat, if there is a relationship between noise and aggression, this would strengthen our conclusion that the relationship is independent of the habitat.

We checked multicollinearity for both models by calculating variance inflation factors (VIF) using the function vif from the package car (Fox and Weisberg 2019). The largest VIF value for our models was 2.2, which is lower than recommended cut‐offs (< 4). We checked residual diagnostics using the package DHARMa (Hartig 2024). We report coefficient estimates, standard errors from the output of lmer function, as well as χ ² and p values for evaluating statistical significance for the overall terms using the Anova function from the package car. To interpret the interaction effect, we performed post hoc tests with Tukey‐adjusted p values using the pairs function from the package emmeans (Lenth and Piaskowski, 2025). We plotted our data using the packages ggplot2 (Wickham 2016) and ggeffects (Lüdecke 2018).

All the statistical analyses are performed in R (R Core Team 2024). Details of package versions and dependencies can be found at the end of the R script provided in the Supporting Information.

2. Results

Aggressive behavior during simulated territory intrusions depended on the interaction between habitat and trial order (interaction effect: χ ² = 6.24, p = 0.01, see Table 1, Figure 1). In their first trials, aggression scores did not differ between urban and rural chaffinches (contrast estimate = −0.97, SE = 0.61, p = 0.39). However, chaffinches in rural habitats became less aggressive in their second trial (contrast estimate = 1.22, SE = 0.35, p = 0.01) while urban chaffinches showed no change in aggression between trials (contrast estimate = −0.06, SE = 0.37, p = 0.99). As a result, urban chaffinches had higher aggression scores in their second trial than rural chaffinches (contrast estimate = −2.25, SE = 0.61, p = 0.004).

TABLE 1.

Output from a linear mixed model assessing whether aggression scores are influenced by habitat (urban, rural) and trial order (first, second).

Predictor	Estimate	95% CI	χ 2	df	p
Intercept	−0.16	−1.02 to 0.69	0.15	1	0.70
Habitat [urban]	0.97	−0.26 to 2.21	2.54	1	0.11
Trial order [second]	−1.22	−1.93 to −0.50	11.81	1	< 0.001
Habitat [urban] × trial order [second]	1.28	0.24 to 2.31	6.24	1	0.01

Note: Bold values indicate statistical significance (p < 0.05).

FIGURE 1.

Aggression scores (PC1, higher values indicate higher aggression) in response to a simulated territory intrusion in both urban and rural chaffinches. Although rural chaffinches decreased their aggressive response on their second encounter with a simulated intruder, urban chaffinches did not. Each colored circle represents a single trial, with individuals' first and second trials connected by a line. Group means and standard errors are represented with the darker circles and vertical lines to their right.

Aggression across the two playback trials was highly repeatable in urban chaffinches (R = 0.75, 84% CI = 0.47–0.89, p = 0.006) but was not significantly repeatable in rural chaffinches, although the estimates did not differ from each other (R = 0.41, 84% CI = 0.02–0.70, p = 0.087). Urban chaffinches had significantly lower within‐individual variance than rural chaffinches (urban = 0.42, 84% CI = 0.19–0.67; rural = 1.68, 84% CI = 0.81–2.57), while among‐individual differences did not differ between habitats (urban = 1.28, 84% CI = 0.49–2.37; rural = 1.18, 84% CI = 0–2.5).

Ambient noise levels were significantly repeatable in both habitats (Urban: R = 0.78, 84% CI = 0.56–0.91, p = 0.001; Rural: R = 0.71, 84% CI = 0.36–0.89, p = 0.016). Moreover, controlling for habitat, trial order, and their interaction, the effect of ambient noise levels on aggression scores was significant (estimate = 0.11, SE = 0.05, χ ² = 4.00, p = 0.046). Yet, a separate analysis with the urban subset revealed that there was no relationship between noise and aggression (estimate = 0.09, SE = 0.33, χ ² = 3.13, p = 0.08), where noise levels were more variable (see Figure 2).

3. Discussion

In this study, we tested whether the intensity and flexibility of aggressiveness differed between urban and rural chaffinches in response to simulated territory intrusions, and also whether anthropogenic noise predicts aggression irrespective of habitat. We found that habitat differences in aggression depended on trial order. Although urban and rural chaffinches did not differ in the first trial, urban birds were more aggressive than their rural counterparts in the second trial. Rural birds reduced their aggressiveness in their second trial while urban birds did not. Consequently, urban chaffinches were less flexible in aggression. Although we found a significant positive correlation between ambient noise and aggression scores, this relationship was absent in the urban site alone. Therefore, our data are not appropriate for making strong conclusions about the effect of noise on aggression due to the almost non‐overlapping distribution of noise between the habitats.

As we predicted, urban chaffinches were more aggressive than rural chaffinches, albeit only during their second territory intrusion trial, while in the first trial there was no difference in aggression between the two populations. Many prior studies have reported higher overall aggressiveness among urban birds (Davies and Sewall 2016; Hardman and Dalesman 2018; Önsal et al. 2022; but see Bókony et al. 2010). In one other study, Beck et al. (2023) reported a similar interaction between habitat and trial number, whereby rural and urban song sparrows ( Melospiza melodia ) did not differ in aggressiveness during their first simulated territorial intrusion but did so in a second intrusion on the same subjects 2 weeks later, with rural birds showing larger decreases in aggression than urban birds. These findings are interesting because while a number of studies found urban birds to be more aggressive than rural birds, the difference may be context or season dependent.

The differences in the repeatability of aggressiveness between urban and rural chaffinches in the present study may be due to multiple different processes. In our design, we used the same stimuli across two trials for each male to avoid confounds in which specific stimulus characteristics may drive responses. However, it may also have led to reduced response in the rural birds due to habituation or reduced effort based on recognition of a previously defeated intruder. By contrast, urban birds may not show habituation or recognition for multiple reasons. First, anthropogenic noise may make recognition of signals more difficult by masking the finer details of the stimuli. Alternatively, if urban habitats have higher territory value due to more heterogeneous distribution of suitable habitat or resources (Foltz et al. 2015; Marzluff 2001; Juárez et al. 2020), reducing aggression might be too costly for urban chaffinches, leading them to keep the same level of aggression even to a repeat intruder. These hypotheses need further experimental testing.

Whatever the cause, it seems urban birds show reduced within‐individual variation in their behavior compared to rural birds (cf. Hardman and Dalesman 2018). This finding is inconsistent with the hypothesis that urban adapters should have greater behavioral flexibility (Caspi et al. 2022). Nevertheless, other studies investigating different bird species reported similar findings (e.g., Beck et al. 2023). In dark‐eyed juncos, stress causes a decrease in the aggressiveness of rural but not urban dark‐eyed juncos (Abolins‐Abols et al. 2016). Conversely, rural but not urban European robins ( Erithacus rubecula ) increased aggression when presented with experimental noise (Önsal et al. 2022). As being able to adjust aggression would prevent unnecessary injuries or energy wastage, behavioral inflexibility may have serious fitness costs for urban animals. Although our results highlight the importance of studying the flexibility of aggressive behavior in human‐impacted environments, these results should be interpreted with caution because our sample size was limited and we were only able to study chaffinches in one urban habitat.

Author Contributions

Alper Yelimlieş: conceptualization (equal), formal analysis (lead), investigation (equal), writing – original draft (lead), writing – review and editing (equal). Çağla Önsal: investigation (equal), writing – review and editing (equal). Çağlar Akçay: conceptualization (equal), investigation (equal), supervision (lead), writing – review and editing (equal).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1: ece373145‐sup‐0001‐Supinfo.zip.

Data Availability Statement

Data and R script to reproduce analyses can be found in the Supporting Information.