Cognition mediates response to anthropogenic noise in wild Western Australian magpies (Gmynorhina tibicen dorsalis)

Anthropogenic noise is a pollutant of growing concern, with wide‐ranging effects on taxa across ecosystems. Until recently, studies investigating the effects of anthropogenic noise on animals focused primarily on population‐level consequences, rather than individual‐level impacts. Individual variation in response to anthropogenic noise may result from extrinsic or intrinsic factors. One such intrinsic factor, cognitive performance, varies between individuals and is hypothesised to aid behavioural response to novel stressors. Here, we combine cognitive testing, behavioural focals and playback experiments to investigate how anthropogenic noise affects the behaviour and anti‐predator response of Western Australian magpies (Gymnorhina tibicen dorsalis), and to determine whether this response is linked to cognitive performance. We found a significant population‐level effect of anthropogenic noise on the foraging effort, foraging efficiency, vigilance, vocalisation rate and anti‐predator response of magpies, with birds decreasing their foraging, vocalisation behaviours and anti‐predator response, and increasing vigilance when loud anthropogenic noise was present. We also found that individuals varied in their response to playbacks depending on their cognitive performance, with individuals that performed better in an associative learning task maintaining their anti‐predator response when an alarm call was played in anthropogenic noise. Our results add to the growing body of literature documenting the adverse effects of anthropogenic noise on wildlife and provide the first evidence for an association between individual cognitive performance and behavioural responses to anthropogenic noise.

cardiovascular disease and numerous other health concerns in humans (World Health Organization.Regional Office for Europe, 2011).
Only in the last two decades have the effects of noise on nonhuman animals begun to garner significant attention, with research indicating a wide range of adverse effects (Kunc & Schmidt, 2019;Shannon et al., 2016).At the ecosystem level, reductions in species abundance, distribution, and richness have been reported following exposure to anthropogenic noise (Goodwin & Shriver, 2011;McClure et al., 2013;Pandit et al., 2021;Willems et al., 2022;Wilson et al., 2021), while at the species level, disruptions to physiology, movement, behaviour and fitness have been recognised (Evans et al., 2018;Kunc & Schmidt, 2021;Schroeder et al., 2012;Shannon et al., 2016).
Such disruptions are proposed to arise via three main mechanisms.First, noise may mask important auditory cues or signals if they are spectrally similar to that of the anthropogenic noise, so cognitive processes may play a role in the adjustment of vocal parameters by callers such that vocalisations are better conveyed, or in the ability of receivers to perceive and respond to vocalisations in noise (Dooling & Blumenrath, 2013;Zhou et al., 2019).Alternatively, anthropogenic noise may act as a distraction, diverting the limited attentional resources of individuals away from vital behaviours (Purser & Radford, 2011).Finally, noise may be perceived by animals as a stressful or threatening stimulus, inducing a physiological or behavioural stress response (Luo et al., 2015;Simpson et al., 2015).
These mechanisms may act separately or in conjunction with one another to alter animal behaviour (Eastcott et al., 2020).Cognitive processes such as learning, categorisation, memory and inhibitory control may help individuals to make optimal behavioural decisions during noise, whether that be to reduce vocalising in noise when acoustic communication is disrupted or maintain behaviours such as their anti-predator response, when threatening or distracting noise occurs.
By far the most well researched consequence of anthropogenic noise is on animal communication (Kunc & Schmidt, 2021;Shannon et al., 2016).Animals use acoustic signals to convey information about reproductive status, territory ownership and predator presence, as well as many other life processes (Bradbury & Vehrencamp, 2011).
Noises that disrupt these signals may therefore affect multiple aspects of an animal's life (Evans et al., 2018).Studies across taxa have revealed species may alter the amplitude (Halfwerk et al., 2016;Lowry et al., 2012;Sementili-Cardoso & Donatelli, 2021;Templeton et al., 2016), frequency (Proppe et al., 2011;Slabbekoorn & den Boer-Visser, 2006), structure (Nemeth & Brumm, 2009) or timing (Miller et al., 2000;Pearson & Clarke, 2018;Proppe et al., 2011) of their vocalisations in noisier areas, possibly in an attempt to maintain communication during noise.One area of communication often negatively affected by noise is alarm calling and the anti-predator response (Antze & Koper, 2018;Grade & Sieving, 2016;Morris-Drake et al., 2017;Templeton et al., 2016;Zhou et al., 2019).Anthropogenic noise, either via acoustic masking or attentional distraction, can disrupt the production and perception of alarm calls and thus affect an individual's anti-predator response.Disruptions to the anti-predator response have been seen in superb fairy-wrens (Malarus cyaneus) (Zhou et al., 2019), great tits (Parus major) (Templeton et al., 2016) and Savannah sparrows (Passerculus sandwichensis) (Antze & Koper, 2018), where individuals exhibited a reduced anti-predator response after playback of an alarm call paired with anthropogenic noise, compared to an alarm call without anthropogenic noise.Such disruption to the anti-predator response of species could affect survival, and disproportionately affect mortality rates of populations residing in noise-affected areas.
Although most research on anthropogenic noise has focused on species-and population-level effects, recent research has highlighted the importance of considering factors influencing individual variation in response to anthropogenic noise (Eastcott et al., 2020;Harding et al., 2019), and has revealed that both intrinsic and extrinsic factors can influence individual response.For example, intrinsic factors such as age (Eastcott et al., 2020;Grunst et al., 2021;McClure et al., 2017), sex (Bruintjes & Radford, 2013;Grunst et al., 2021Grunst et al., , 2023)), body condition (Harding et al., 2020;Purser et al., 2016) and dominance status (Eastcott et al., 2020) can affect individual responses to anthropogenic noise.Variation in individual response to anthropogenic noise can also arise due to extrinsic factors, such as environmental context (Goldbogen et al., 2013;Neo et al., 2018), prior exposure (Conomy et al., 1998;Neo et al., 2018) and co-occurrence of other stressors (McCormick et al., 2018).
Growing evidence for individual variation in response to anthropogenic noise highlights the importance of considering such variation, as well as the factors driving it, to gain a better understanding of how animals respond to this man-made stressor.
One intrinsic factor that has yet to be considered in this context is cognition.Cognition, encompassing the ways in which animals perceive, process, store and act on environmental information, is vital for all aspects of animal life (Shettleworth, 1998).A growing body of research posits that cognition, through its underlying effect on behavioural plasticity, underpins the response of individuals to novel environments and anthropogenic stressors (Griffin et al., 2017;Lee & Thornton, 2021;Potvin, 2017;Sol et al., 2020;Szabo et al., 2020).For example, Potvin (2017) suggested cognition may help animals to cope with anthropogenic noise, either through facilitating avoidance of noise sources, enabling short-term flexible behavioural adjustments, or allowing for cross-generational adaptation to noise.Short-term behavioural adjustments to anthropogenic noise have been observed in numerous avian species and may be facilitated by cognitive processes such as learning and memory.In male great tits, individuals minimised acoustic masking from anthropogenic noise by switching to either songs with a higher minimum frequency (when exposed to low-frequency city noise) or songs with a lower maximum frequency (when exposed to high-frequency city noise; Halfwerk & Slabbekoorn, 2009).These birds may have learnt to adjust their song types based on their memory of previous experiences in which their communication using either song type has been adversely affected by city noise (Potvin, 2017).Long-term behavioural adjustments to anthropogenic noise have also been observed.For example, female house sparrows (Passer domesticus) exposed to chronic traffic noise displayed increased vigilance and were able to detect and flee from an observer earlier compared to control birds (Meillere et al., 2015).This adjustment of anti-predator behaviour may arise because individuals have learnt to increase their vigilance to compensate for the potential decreased detectability of acoustic signals during anthropogenic noise (Potvin, 2017).Thus, individuals' ability to adjust their behavioural responses to noise, or maintain their normal behavioural responses in noise, may be contingent on their ability to both learn and remember past experiences involving anthropogenic noise.In certain situations, such as in the anti-predator response, it may be more beneficial for individuals to maintain their normal response during anthropogenic noise, rather than alter their behaviour, as a reduced response may directly impact survival (Templeton et al., 2016;Zhou et al., 2019).In such situations, cognitive processes such as memory, learning and inhibitory control may help individuals to inhibit their initial instinct to respond to noise, and instead maintain their normal behavioural response.
Despite suggestions that cognition may aid in coping with anthropogenic stressors such as noise, and examples of species exhibiting behavioural adjustments in the face of anthropogenic noise, no study to date has experimentally investigated whether individual cognitive performance affects responses to anthropogenic noise.
Here, we combine behavioural focals, playback experiments and cognitive testing on a wild population of urban-living Western Australian magpies (Gmynorhina tibicen dorsalis) to investigate whether individual variation in cognition is related to variation in response to anthropogenic noise.We predict that individuals with higher cognitive performance will decrease their foraging effort and vocalisation rate during loud anthropogenic noise to avoid the energetic costs associated with these behaviours when they are less efficient due to noise.We predict these high performing individuals will also increase their vigilance in loud anthropogenic noise conditions to increase the input of environmental information when acoustic communication is hampered.With regards to our playback experiments, we predict that individuals with greater cognitive performance will respond more similarly (i.e. have more similar vigilance, latency to return to normal behaviours, and flee responses) to the playback of an alarm call alone (their 'normal' anti-predator response) and playback of an alarm call with anthropogenic noise, as these individuals may be less affected by the presence of noise, or may be more able to process multiple simultaneous acoustic stimuli.These birds will therefore be less affected by the addition of anthropogenic noise compared to individuals with low performance in our cognitive tasks.

| Study species and site
The Western Australian subspecies of magpie is a large (250-400 g) passerine inhabiting urban parklands and open grassy areas in southwest Western Australia (Johnstone, 2004).The magpie is a highly vocal group-living species that uses vocal communication to relay information about predators, defend territories, coordinate with group members and maintain group cohesion (Dutour et al., 2020;Dutour & Ridley, 2020).This subspecies breeds cooperatively and lives in stable groups of two to 12 individuals (Pike et al., 2019).Magpies are primarily terrestrial foragers that use their acute hearing to locate subterranean prey (Edwards et al., 2015;Floyd & Woodland, 1981).
Individuals are sexually dichromatic, and sex can be visually discerned from roughly 3 years of age, when individuals achieve adult plumage (Dutour & Ridley, 2020;Edwards et al., 2015).
The study population comprised 16 groups located in urban parklands of Crawley and Guildford, Perth.All groups are habituated to human presence (Edwards et al., 2015), allowing for close observation (1-10 m) of natural behaviour.Most individuals in the study population have been colour-ringed for individual identification, and unringed birds are identifiable by plumage anomalies or distinctive scarring.The territories of all study groups are in urban areas and are exposed to daily anthropogenic noise from road traffic, train lines, airplanes, construction, or machinery.Fieldwork was conducted in July 2021-October 2022 between 5 am and 11 am, when birds are most active (Edwards et al., 2015).

| Focal observations
The effect of anthropogenic noise on foraging, vigilance and vocalisation behaviours was determined by conducting repeated focal observations on 75 birds across 16 groups (N = 333 focals).Focal observations (as described by Altmann, 1974) consisted of 20-min time activity periods where all behaviours of the focal individual were recorded, including vigilance, foraging and vocalising.A DIGITECH Micro Sound Pressure Level Meter was used to measure the maximum amplitude of anthropogenic noise events (noise events typically included construction, gardening machinery, traffic, planes and trains) that occurred during focals.The start and end of each noise event was determined by the observer from when the event was first discernible, to when it was no longer discernible by the observer.Anthropogenic noise events were only included as 'loud anthropogenic noise' if the maximum amplitude exceeded 50 dB (Grade & Sieving, 2016).Focals were recorded on an ethogram programme created in CyberTracker on an android phone.
The proportion of each focal that the bird spent exhibiting foraging behaviours was referred to as the 'foraging effort' of the bird.
All prey caught by the focal bird was recorded, with the approximate size of prey items specified using prey size categories described previously by Edwards et al. (2015).Prey type (to broad taxonomic group) was recorded where possible.This allowed us to estimate the biomass (in mg) of prey consumed during each focal, which was then divided by the time spent foraging (min), to give a measure of 'foraging efficiency (mg/min)'.Foraging efficiency was only analysed for focals in which birds spent at least 5 min foraging (N = 218 focals on 70 birds).The total time spent vigilant was divided by total focal time to determine the proportion of focal time spent vigilant.All vocalisations produced by the focal bird were noted (including alarms, carols, warbles, and choruses) and were divided by total focal time to get a measure of vocalisation rate per hour.Each focal included time when anthropogenic noise events exceeding 50 dB in amplitude were occurring (classified here as 'loud anthropogenic noise'), and time when only ambient noise and quieter anthropogenic noise was present (classified here as 'background noise').This allowed for the calculation of foraging effort & efficiency, vocalisation rate, and proportion of time vigilant within each of these two noise conditions ('loud anthropogenic noise' and 'background noise').For more information regarding focal observations, see Supporting Information Section S1.

| Call collection
Alarm calls for the playback experiment were collected using presentations of a taxidermied red fox (Vulpes vulpes), the same model previously used to elicit alarm calls in our study population (Dutour et al., 2021).The predator model was covered with a blanket attached to a string and placed in the field near a foraging individual.
Once individuals were in social isolation, defined hereafter as with other magpies at least 10 m away (achievable as this species often forages >10 m apart (Ashton et al., 2018)), and approximately 10 m from the predator model, the blanket was removed by an experimenter approximately 5 m from the model by gently pulling on the string.Since magpies alter parameters of their alarm calls according to predator type and distance (Dutour et al., 2020), the same predator was presented at the same distance (10 m) during each predator presentation.Alarm calls were recorded by an observer standing 10 m from the focal bird, using a RODE NTG-2 directional microphone set within a Blimp suspension windshield system and attached to a Roland R-07 wave/MP3 recorder at a sampling rate of 44.1 kHz.
Naturally elicited warbles and airplane noise were collected using the same microphone set-up described above.Warbles are a common song of Western Australian magpies that do not elicit an anti-predator or vigilance response (Dutour et al., 2021) and therefore were used as a control in this study.All airplane noise was recorded from the same location in Guildford, Western Australia (−31.901681' S, 115.971540′E), directly under the flight path of planes departing Perth Airport, which is ca. 3 km south of the study site.

| Playback preparation
Playback tracks were prepared using Audacity version 3.0.2.Only high-quality vocalisations (high signal to background noise ratio) were used.Five types of playbacks were prepared: (1) alarm call, (2) alarm call and airplane noise, (3) warble, (4) warble and airplane noise, and (5) airplane noise.Playback tracks were 40 s in length, consisting of 5 s of silence followed by either 35 s of ambient background noise or airplane noise.This duration was chosen as 35 s is the typical duration of airplane noise from an airplane flying directly overhead (Blackburn, pers. obs.).Warbles or alarm calls occurred during the airplane or background noise, and halfway through these (22.5 s into the track; see Supporting Information Section S2 for information about vocalisations used).Each set of playback tracks featured calls from an individual of the same sex and group as the focal bird, however different alarm, warble and airplane noise exemplars were used for each playback to avoid pseudo-replication (Kroodsma et al., 2001).
A DIGITECH Micro Sound Pressure Level Meter was used to measure the amplitude of magpie warbles and alarms from 1 m from the calling bird and to measure the amplitude of airplane noise from the location it was collected.The amplitude of magpie calls at 1 m was consistent with previous measurements by Dutour et al. (2021; warbles 70.8 dB; alarm calls 68.9 dB).The amplitude of airplane noise ranged between 80 and 90 dB, the amplitude at which plane noise normally occurred at the location it was recorded for this study.
Once playback tracks were created in Audacity (see Supporting Information Section S2), they were transferred to a Fiio M6 portable high resolution music player, which was attached via aux cord to a UE boom 2 speaker.Tracks were then checked again using the DIGITECH Micro Sound Pressure Level Meter at 1 m from the speaker to ensure tracks replicated the naturally occurring magpie and airplane noise amplitude.

| Playback presentations
Playbacks began when focal birds were in social isolation (see above) and foraging on the ground in an open area, showing no signs of vigilance and with no magpie vocalisations or loud (>50 dB) anthropogenic noise having occurred in the minute prior to playback.The speaker was placed 10 m from the focal bird, and between the focal bird and calling bird (bird whose vocalisations were used in playback tracks), so that the direction the call came from did not violate the expectation of the focal bird.Playback tracks were presented to focal individuals in a randomised order, at least 1 day apart, to reduce habituation to playback.The behaviour of focal magpies was video-recorded from 1 min prior to the start of playback, throughout the playback, and for 1 min post playback, using a Panasonic HC-V180 video recorder.
The following behaviours were assessed via video analysis: whether focal individuals fled following playback, the time focal individuals spent vigilant in the 30 s post-playback (birds were considered vigilant when they were displaying an erect posture and were scanning the surrounding environment; Blackburn, Ridley, & Dutour, 2022;Zhou et al., 2019) and the time taken for focal individuals to return to normal behaviour post-playback, to a maximum of 60 s (birds were considered to have returned to normal behaviour when they spent a minimum of three consecutive seconds foraging and non-vigilant).The video analyst (GB) was blind to playback treatment during video analysis.

| Bioacoustic recorders
To determine the normal soundscape magpies were exposed to, we deployed Song Meter Mini recording units (Wildlife Acoustics Inc.) in each territory.Recorders were set to collect mono recordings continuously with a sampling rate of 44.1 kHz, with a gain of 12 dB for a period of two weekdays and two weekend-days.
Recording units were mounted at a height between 3 and 5 m in trees commonly used by magpie groups, with microphones positioned parallel to the ground.In total, we collected 96 h of recordings at each of 16 territories (1536 h of recording in total).
Recordings were then analysed using Kaleidoscope Pro software 5.4.9 (Wildlife Acoustics Inc.).For each 60-min sampling period, minimum, maximum and mean sound pressure levels (SPL) were extracted, as well as cumulative sound energy level (SEL).Within each territory, these measures were averaged across the 96 sets of 60-min samples, to get an average of hourly minimum SPL, maximum SPL, mean SPL and cumulative SEL per territory.Mean SPL of territories ranged from 40.5 to 50.8 dB.

| Cognitive testing
Cognitive testing was conducted when focal birds were in social isolation and not displaying vigilance.Individuals were tested on two ecologically relevant traits, associative learning and spatial memory.Associative learning is thought to be involved in foraging, predator avoidance, and intraspecific competition (Blackburn, Broom, et al., 2022;Shaw et al., 2015), while spatial memory may be important in remembering the location of resources, predators and territory boundaries (Ashton et al., 2018;Shaw et al., 2015).
Previous research on this species has found that individual performance is positively correlated across different cognitive tasks (including associative learning and spatial memory) (Ashton et al., 2018).Therefore, performance in the associative learning and spatial memory tasks may be seen as a measure of the general cognitive performance of individual magpies.Both cognitive tasks require individuals to peck downwards at a lid to gain a food reward (mozzarella cheese), therefore the movement required by magpies to complete these cognitive tasks mimics their natural foraging behaviour.

| Associative learning
The task used to quantify associative learning consisted of a wooden block (31 × 9 × 4 cm) with four circular wells arranged in a line, two of which were covered with plastic PVC lids suspended on elastic bands (allowing the lids to sit securely covering the wells and swivel when pecked), making the underlying food reward accessible (Figure S1).
The PVC lids were painted different shades of the same colour and therefore required individuals to learn an association between a colour shade and a food reward.
Each bird was randomly assigned a shade, either dark or light, of a colour they had not previously been tested on (Supporting Information Section S4).Testing followed the protocol of Shaw et al. (2015) and Ashton et al. (2018) whereby individuals in their first trial were allowed to search both wells to ascertain that only one colour shade was associated with a food reward.In subsequent trials, individuals were only allowed to peck at one well before the task was removed.If individuals chose the rewarded colour shade, they were allowed to eat the food reward, however if they chose the non-rewarded colour shade, the task was removed before the individual could eat the food reward under the other (rewarded) lid.The task was regularly rotated, to ensure birds were associating the colour shade of the lid with the food reward, and not the side of the array.As per Ashton et al., 2018(Ashton et al., 2018;Shaw et al., 2015), individuals were considered to have passed the associative learning task when they pecked the rewarded lid in at least 10 out of 12 consecutive trials (representing a significant deviation from random binomial probability; binomial test: p = 0.039).An individual's associative learning score was quantified as the number of trials required to achieve this criterion, with lower scores indicating better performance.

| Spatial memory
The task used to quantify spatial memory consisted of a wooden foraging grid (36 × 40 × 4.5 cm) with eight wells arranged in 3 rows and covered with light grey plastic PVC lids that swivel on an axis when pecked (Figure S1, Supporting Information Section S5).
The protocol used to quantify spatial memory was based off Shaw et al. (2015) and Ashton et al. (2018), but with one additional phase.For each bird, one of the eight wells was randomly selected to be the rewarded well.The first two phases of the spatial memory task consisted of a 'training' phase.In the first phase of the spatial memory task, individuals searched the foraging grid until they found the rewarded well and consumed the food reward.The foraging grid was then removed, and after 5 min, this was repeated.The third, fourth and fifth phases of the experiment involved presentation of the foraging grid 24, 48 and 72 h after the initial training phases.The additional fifth phase was included because previous experiments on spatial memory in birds has shown task performance to significantly improve over time or trial number (Ashton et al., 2018;Sewall et al., 2013).Spatial memory performance was quantified as the combined total number of wells searched prior to finding the food reward in the 24-, 48-and 72-h test trials.A final 'probe' trial was then conducted to ensure birds were not using olfactory cues to locate the food reward (see Supporting Information Section S5).

| Statistical analysis
The effect of anthropogenic noise on magpie behaviour (Blackburn et al., 2023) was analysed by model selection using generalised linear mixed models (GLMMs) in the lme4 package (Bates et al., 2015) in R studio 4.2.0 (RStudio Team, 2022).Group and bird identity were included as random effects in all models.

| Analysis of focal observations
Only behavioural focals with more than 30 s in each of the noise conditions (loud anthropogenic noise/background noise) were used (N = 333 focals, 75 magpies), to ensure equal sample sizes and matching of focals in each noise condition.For analysis of foraging effort and proportion of time spent vigilant, the response variable was proportional and hence a binomial distribution was used, with total time (seconds) as the binomial denominator and the time spent foraging or vigilant (seconds) as the response variable.Analysis of factors affecting foraging efficiency (mg/min) and vocalisation rate (vocalisations/hour) used zero-inflated negative binomial GLMMs.
In all models, predictors included sex (M/F), weather (cloudy/clear), adult group size, average hourly minimum sound pressure level (SPL), maximum SPL, mean SPL, cumulative sound energy level (SEL) and noise condition (loud anthropogenic noise/background noise).
To investigate whether cognitive performance affected magpie behaviour and response to anthropogenic noise, the same analysis was conducted on a subset of focals on individuals who had completed testing on both the associative learning and spatial memory tasks (N = 235 focals, 52 magpies).In this analysis, associative learning and spatial memory scores were added as predictors.

| Analysis of playback experiment
For analysis of both the proportion of time spent vigilant in the 30 s post-playback, and time taken to return to normal behaviour, binomial GLMMs were used, with total time (30 s for vigilance postplayback, or 60 s for time to return to normal behaviour) as the binomial denominator, and the time vigilant or time taken to return to normal behaviours (seconds) as the response variable.To determine factors affecting whether individuals fled to cover post-playback, a binomial GLMM with a 0.1 response variable was used (where 0 = individual did not flee, and 1 = individual fled).Models included weather (cloudy/clear), adult group size, average hourly minimum SPL, maximum SPL, mean SPL, and cumulative SEL, sex, track order (the order in which playback tracks were presented to focal birds) and time spent vigilant in the 30 s prior to playback as predictors.
The same analysis was then conducted using a subset of playbacks on birds who had undergone cognitive testing (N = 100 playbacks across 20 birds), adding spatial memory and associative learning scores as predictors.

| Model selection
Model selection was conducted using Akaike information criterion values corrected for small sample size (AICc) to determine which candidate models best explained variation in the data.Models including single terms, additive terms, and interactions were tested based on their suitability as plausible biological hypotheses (Burnham & Anderson, 2002).If the 95% confidence intervals of a term when tested alone intersected 0, the term was not included in additive models.If multiple terms were highly correlated (e.g.mean SPL and maximum SPL), the term with the lowest AICc alone was used in further additive models (Harrison et al., 2018).Models were compared to a null model containing only the intercept and random terms.A top model set was generated by considering all models within 2 AICc of the best model.If multiple models were within 2AICc of the top model, the simplest model was chosen as per Harrison et al. (2018).
Model terms with confidence interval parameters intersecting 0 were not considered good predictors of data patterns.

| Posthoc comparisons
For models with significant interactions between anthropogenic noise treatment or playback track and other explanatory factors, we used the functions emmeans and emtrends (within R package emmeans (Lenth, 2019)) to obtain contrasts between different noise treatments or playback tracks and sex, associative learning score or spatial memory score.p-values adjusted for multiple comparisons via the Tukey method were also calculated using emmeans and emtrends.

| Foraging effort
Foraging effort was significantly lower during loud anthropogenic noise conditions (Table 1 and Figure 1a), and this relationship was affected by sex (Table 1; Figure S2a).Posthoc comparisons revealed that both males and females foraged significantly less in the presence of loud anthropogenic noise (Figure 2a; Table S2), however males foraged significantly more than females during background noise conditions (Figure 4a; Table S2) and therefore experienced a larger decline in foraging effort when loud anthropogenic noise was present.Individuals from larger groups spent more time foraging (Table 1) and cloudy weather significantly decreased foraging effort (Table 1; Figure S2b).
For individuals that completed cognitive testing, neither associative learning nor spatial memory performance affected foraging effort (Table S3).

| Foraging efficiency
Individuals had higher foraging efficiency during background noise (average of 100 mg/min) compared to during loud anthropogenic noise conditions (average of 30 mg/min) (Table 2 and Figure 1b).For individuals who had completed cognitive testing there was no significant effect of associative learning or spatial memory performance on foraging efficiency (Table S5).

| Vigilance
Individuals spent significantly more time vigilant under loud anthropogenic noise conditions compared to under background noise conditions (Table 3 and Figure 1c).The interaction between minimum SPL and noise condition was significant, whereby individuals from territories with a higher minimum SPL were less vigilant during loud anthropogenic noise compared to individuals from areas with a lower minimum SPL (Table 3; Figure S3c).Weather and sex also influenced time spent vigilant, with males spending more time vigilant than females (Table 3; Figure S3a), and birds spending more time vigilant in cloudy compared to clear conditions (Table 3; Figure S3b).
The interaction between noise condition and spatial memory score was significant (Table 4 and Figure 2).Posthoc comparisons revealed that birds who performed worse in the spatial memory task spent less time vigilant under loud anthropogenic noise conditions (Table 4 and Figure 2).In addition, males spent significantly more time vigilant than females regardless of noise condition (Table 4).
There was no effect of associative learning performance on time spent vigilant (Table S7).

| Vocalisation rate
Individuals vocalised significantly less during loud anthropogenic noise (average of 19 vocalisations/hour) compared to background noise (average of 23.9 vocalisations/hour) (Table 5 and Figure 1d).In addition, females vocalised significantly more than males regardless of noise condition (Table 5; Figure S4), with an average vocalisation rate 1.7 times higher than males (females average vocalisation rate: 27 vocalisations/hours; males average vocalisation rate: 15.9 vocalisations/hour).For individuals who had completed cognitive testing there was no significant effect of associative learning or spatial memory scores on vocalisation rate (Table S9).were generated from zero-inflated negative binomial GLMMs with group and bird identity as random terms.CI, confidence interval.For a list of all models tested, see Supporting Information (Table S4).

| Time spent vigilant post-playback
Individuals spent significantly more time vigilant in the 30 s following the alarm only playback treatment compared to any other playback treatment, including the alarm + plane track (Table 6 and Figure 3a; Table S11).In contrast, the warble (control) track elicited a significantly lower vigilance response compared to any other playback track (Figure 3a; Table S11).Individuals also spent significantly more time vigilant following the alarm + plane track compared to the plane only track (Figure 3a; Table S11).The interaction between TA B L E 3 Top model set of candidate terms influencing time spent vigilant.

Note:
Data are based on 235 20-min focals across 52 magpies.Outputs were generated from binomial GLMMs with group and bird identity as random terms.CI, confidence interval.For a list of all models tested, see Supporting Information (Table S7).
sex and playback treatment was significant, with males and females responding differently to the playback tracks (Table 6; Table S12 and Figure S5).Posthoc comparisons revealed that females spent significantly more time vigilant following the alarm call playback compared to any other track, including the alarm + plane playback (Table S12 and Figure S5).In comparison, males spent more time vigilant following both the alarm call playback and alarm + plane playbacks (Table S12 and Figure S5).Time spent vigilant pre-playback significantly affected the proportion of time magpies spent vigilant post-playback (Table 6).
For individuals who had undergone cognitive testing, associative learning score was a significant predictor of individual variation in response to playback tracks (Table 7 and Figure 3b).
Posthoc comparisons revealed that following the alarm-only playback, individuals who performed better in the associative learning task (lower score) displayed lower vigilance compared to birds who performed worse in the associative learning task (higher score; Posthoc comparison; t-ratio = 2.65, df = 87, p < .01; Table S14; Figure 3b).However, following the alarm + plane playback, individuals who performed better in the associative learning task had significantly higher vigilance compared to birds who performed worse (Posthoc comparison; t-ratio = −2.73,df = 87, p < .01;Table S14; Figure 3b).This suggests that individuals that performed better in the associative learning task had a more similar vigilance response between these two track types (alarm only and alarm + plane), and thus experienced less disruption to their anti-predator response compared to individuals who performed worse in the associative learning task.Individuals who spent more time vigilant in the 30 s pre-playback also spent more time vigilant post-playback (Table 7).

Note:
Data are based on 333 20-min focals across 75 magpies.Outputs were generated from zero-inflated negative binomial GLMMs with group and bird identity as random terms.CI, confidence interval.For a list of all models tested, see Supporting Information (Table S8).

TA B L E 6
Top model set of candidate terms influencing time spent vigilant in the 30 s post-playback.Note: Data are based on 120 playbacks conducted on 24 magpies.Outputs were generated from binomial GLMMs with group and bird identity as random terms.CI, confidence interval.For a list of all models tested, see Supporting Information (Table S10).
any other track (Table 8 and Figure 4a; Table S16).Individuals also took significantly longer to return to normal behaviours after the alarm + plane track compared to the plane only track (Figure 4a, Table S16).
Weather condition also affected playback response, with individuals taking longer to return to normal behaviour following the alarm call and warble tracks during cloudy weather (Figure S7 and Table S17).Despite this, under both weather conditions, response to an alarm-only playback was significantly higher than the response to an alarm call played in the presence of plane noise (Figure S7 and Table S17).
For individuals who had undergone cognitive testing, individuals that performed better in associative learning tests (lower scores, indicating fewer trials to reach learning criterion) took less time to return to normal behaviours following all playback tracks compared to individuals that performed worse (higher scores; Figure S8; Table 9).Individuals responded differently to the playback tracks depending on their performance on the associative learning task (Table 9 and Figure 4b).Posthoc comparisons indicated that there was a significant positive relationship between associative learning score and the time taken for focal individuals to return to normal behaviour following the alarm only track (Table S19; Figure 4b).Following the alarm only playback track, individuals who performed better in the associative learning task took less time to return to normal behaviours compared to birds who performed worse (Posthoc comparison; t-ratio = 2.83, df = 87, p < .01;Table S19; Figure 4b).Response to the alarm + plane playback track did not differ significantly between birds based on cognitive performance (Posthoc comparison; t-ratio = −1.28,df = 87, P d = 0.20; Table S19).
Comparing the difference in response between alarm only and alarm + plane tracks in individuals with the best associative learning scores (score of <13, top 20% of individuals) and those with the worst associative learning score (score of >20, bottom 20% of individuals) reveals that this difference was significant (onesided Mann-Whitney test, w = 14, p = .036,n = 8), and suggests that individuals who performed better in the associative learning task were more able to maintain their anti-predator response when plane noise was present.Individuals who spent more time vigilant pre-playback also spent more time vigilant post-playback (Table 9).

| Flee response
Playback treatment significantly affected whether individuals fled to cover post-playback (Table 10 and Figure 5).No birds fled to cover following the warble and warble + plane playback (Figure 5).6. Points are raw data; fitted lines and 95% confidence intervals generated from output of the top model presented in Table 7.
playback (Table 10 and Figure 5).Following playback of an alarm call alone, 37.5% of birds fled to cover, however following playback of an alarm + plane, only 8.3% of birds fled to cover.For individuals who had completed cognitive testing there was no significant effect of associative learning or spatial memory performance on whether individuals fled to cover (Table S21).

| DISCUSS ION
We found that anthropogenic noise affected the foraging, vigilance, vocalisation and anti-predator behaviour of wild Western Australian magpies, and that the response of individuals to anthropogenic noise was related to their cognitive performance.Our study represents the first empirical evidence to date of a relationship between cognitive performance and the response of wild animals to anthropogenic noise.

| Anthropogenic noise behavioural focals
When loud anthropogenic noise was present, individuals spent less time foraging and were less efficient foragers.Negative effects of noise on foraging have been identified in numerous species (Evans et al., 2018;Floyd & Woodland, 1981;Montgomerie & Weatherhead, 1997;Senzaki et al., 2016), and may be due to acoustic masking disrupting the ability of individuals to locate subterranean prey using auditory cues (Edwards et al., 2015;Floyd & Woodland, 1981), or may be linked to a trade-off with vigilance behaviours (foraging-predation risk trade-off; Cresswell, 2008;Kern & Radford, 2016).Focal birds in our study spent significantly more time vigilant during loud anthropogenic noise conditions, suggesting they may be trading off foraging with vigilance under these conditions.Such a trade-off can potentially reduce body condition and provisioning rates to young during breeding (Cresswell, 2008;Kern & Radford, 2016;Pandit et al., 2021), with negative consequences for reproductive success, and population viability in the long-term.
Anthropogenic noise also significantly affected vocalisation rates.Reduced vocalisation during loud anthropogenic noise conditions suggests magpies are able to perform short-term temporal adjustments to the timing of their vocalisations to avoid overlap with loud anthropogenic noise, a phenomenom seen in a number of other species (Lengagne, 2008;Melcón et al., 2012;Pearson & Clarke, 2018).This may occur because individuals cannot enhance their vocal signals sufficiently to be heard above this noise, and so are avoiding the energetic costs associated with vocalising at all (Franz & Goller, 2003).By calling only when anthropogenic noise levels are low or absent, individuals may be able to more effectively maintain communication with conspecifics.However, if anthropogenic noise continues to increase as human urban populations grow (World Health Organization.Regional Office for Europe, 2011), it may lead to an overall reduction in the vocalisation rate of individuals, potentially affecting the many life processes animals rely on vocal communication for, such as territory ownership and defence, threat response, foraging and reproduction (Bradbury & Vehrencamp, 2011).

| Sex differences
Male magpies foraged significantly more, and more efficiently, than females during background noise conditions, but foraged less and were less efficient under loud anthropogenic noise conditions.Male magpies may be more sensitive to the effects of anthropogenic noise, and may prioritise vigilance over foraging in the presence of loud anthropogenic noise.These findings align with previous research identifying sex-differences in response to anthropogenic noise (Grunst et al., 2023;Mancera et al., 2017;Powolny et al., 2014).For example, in captive wild mice (Mus musculus), males, but not females, exposed to low-frequency mining noise had increased corticosterone levels compared to controls (Mancera et al., 2017).Such sex-related GLMMs with group and bird identity as random terms.CI, confidence interval.For a list of all models tested, see Supporting Information (Table S13).
TA B L E 8 Top model set of candidate terms influencing time taken to return to normal behaviour post-playback.Note: Data are based on 120 playbacks conducted on 24 magpies.Outputs were generated from binomial GLMMs with group and bird identity as random terms.CI, confidence interval.For a list of all models tested, see Supporting Information (Table S15).8. Points are raw data; fitted lines and 95% confidence intervals generated from output of the top model presented in Table 9.
differences in response to noise and predation risk may arise due to differences in risk-taking behaviour, sensitivity to anthropogenic noise or sex differences in energetic needs (Blickley et al., 2012;Grunst et al., 2023;Mancera et al., 2017;Powolny et al., 2014).

| Cognition and behavioural focals
We found evidence that individual cognitive performance is related to vigilance response to anthropogenic noise.Individuals that performed better in the spatial memory task spent more time vigilant compared to individuals with worse spatial memory performance, but this relationship was only seen under loud anthropogenic noise conditions.The mechanistic basis for this relationship between spatial memory and vigilance is unclear, but it is possible that spatial memory may aid in anti-predator vigilance by assisting individuals in creating a mental map of predation threat across their territory (Gaynor et al., 2019;Laundré et al., 2001).Appropriate perception and evaluation of risk and predator cues may rely in part on the cognitive abilities to correctly associate the predator cue with danger, and with the spatial and temporal location of the cue (Gaynor et al., 2019).When loud anthropogenic noise is present, birds may experience higher perceived risk compared to when loud anthropogenic noise is absent, or may have a compromised ability to hear and respond appropriately to conspecific cues signifying danger (Templeton et al., 2016;Zhou et al., 2019), as was seen in our playback experiment.This may lead these more cognitively adept individuals to shift reliance from group members and their alarm cues to their own vigilance and knowledge of their environment.When anthropogenic noise levels are lower and birds can detect and respond to auditory predator cues, individuals, regardless of their cognitive abilities, may rely on their groupmates for information.This could explain why the relationship between spatial memory and vigilance is only seen in noisier conditions, when disruption to the anti-predator response occurs, and individuals can no longer reliably perceive conspecific alarm cues.
Results from our behavioural focals show a significant effect of loud anthropogenic noise on the behaviour of magpies, and reveal a potential relationship between cognition and behavioural response to noise.While these findings are robust and based on a large sample size (333 focals across 75 individuals), it should be noted that blind scoring of noise condition for these observational focals was not possible, and hence it is possible that unconscious bias may have affected the findings.In addition, more research is needed to disentangle the mechanistic basis of the relationship between spatial memory and behavioural response.

| Anthropogenic noise playback experiment
Our playback experiment revealed that the addition of anthropogenic noise to a conspecific alarm call significantly altered the Warble only and warble and plane playbacks were excluded from this analysis because they did not elicit a flee response from any magpie.Outputs were generated from binomial GLMMs with group and bird identity as random terms.CI, confidence interval.For a list of all models tested, see Supporting Information (Table S20).
anti-predator response of magpies.Magpies spent less time vigilant, showed a decreased latency to return to normal behaviours, and were less likely to flee to cover following playback of an alarm call with plane noise compared to an alarm call with background noise.The increase in vigilance and latency to return to normal behaviours following an alarm call played with plane noise is unlikely to be solely a result of increased vigilance due to plane noise, as magpies still responded significantly more to the alarm call with plane noise playback compared to the plane noise only track.These findings are consistent with previous research documenting decreased anti-predator response when anthropogenic noise is played simultaneously with alarm signals (Antze & Koper, 2018;Grade & Sieving, 2016;Morris-Drake et al., 2017;Templeton et al., 2016;Zhou et al., 2019).Disruptions to the anti-predator response during anthropogenic noise may arise due to partial acoustic masking of alarm calls, an increased stress response or distracted attention due to noise, or because individuals may view anthropogenic noise as a threatening stimulus (Eastcott et al., 2020).Regardless of the mechanisms behind the observed reduction in anti-predator response during noise, results from this study add to the growing literature identifying the negative impact of anthropogenic noise on the antipredator response, an effect that could lead to survival implications for individuals.

| Sex differences
Results from our playback experiment reveal sex to be an important factor explaining individual variation in response to noise.While the addition of plane noise to warbles and alarm calls significantly affected the response of females birds to these conspecific vocalisations, there was no significant difference in the response of males to these vocalisations when they were played alone compared to when they were played with plane noise.These differences in response suggest that males may be better at both perceiving and discerning between types of conspecific calls during anthropogenic noise.The increased vigilance of males seen in our behavioural focals may assist them in perceiving acoustic signals during noise, and explain why males do not significantly differ in their response to vocalisations played alone and with anthropogenic noise.Findings from our behavioural focals and playbacks combined suggest that males may be better at maintaining their perception of acoustically relevant signals during high levels of anthropogenic noise, potentially due to their heightened vigilance, but that this may come at a cost to their foraging behaviour, and therefore may negatively affect their predationstarvation trade-off.

| Cognition and playbacks
Our playback experiment also revealed that individuals' cognitive performance predicted how well they were able to maintain their anti-predator response when an alarm call was played with anthropogenic noise.Individuals that performed better in the associative learning task exhibited a more similar response between the alarm only playback track and the alarm and plane playback track compared to individuals who performed worse, suggesting these individuals may be able to maintain their normal anti-predator response in noise.
Following the playback of an alarm without anthropogenic noise, birds who performed better on the associative learning task spent less time vigilant post-playback, and returned to normal behaviours quicker than birds who performed worse on the task.While it may seem counter-intuitive that individuals with higher cognitive scores responded to a conspecific alarm with reduced vigilance, this may be an adaptive response to minimise non-lethal fitness consequences that can arise due to sustained vigilance (Cresswell, 2008;Ydenberg & Dill, 1986).As well as having the potential to negatively affect body condition, parental care, and reproduction by affecting the predation-starvation trade-off (Cresswell, 2008;Kern & Radford, 2016), sustained vigilance may lead to vigilance decrement, and thus affect how well individuals can process information from their environment regarding predation threat (Dukas & Clark, 1995).
We also found that individuals that performed better in the associative learning task had a significantly heightened vigilance response following playback of an alarm with plane noise, compared to individuals who performed worse in this task.While cognitive performance did not significantly affect latency to return to normal behaviours in the 60s following the alarm and plane playback, this may be due to the fact that most birds had returned to normal behaviour by 30 s post-playback.Therefore the difference in response based on cognitive performance is weakened when looking at the 60 s post-playback compared to the immediate response, measured by the time spent vigilant in the 30 s post-playback, in which associative learning performance did significantly affect response.
The probability of individuals fleeing to cover also was unrelated to cognitive performance in our study; however this is likely due to the relatively small number of individuals fleeing following playback that also underwent cognitive testing (11 instances of fleeing out of 100 playbacks).Regardless of this, better performing individuals were seen to display more comparable latency to return to normal behaviours and vigilance responses between alarm calls played with and without plane noise compared to individuals who performed worse.These differences in response may be due to individuals with greater associative learning scores being able to form a stronger association between alarm call and threat that is less likely to be adversely affected by environmental perturbances such as noise.Environmental factors affect cognitive performance and functioning in a number of species, including Western Australian magpies (Blackburn, Broom, et al., 2022;Freas et al., 2012;Triki et al., 2018), and studies have identified effects of anthropogenic noise on performance in cognitive tasks in animal populations (Cheng et al., 2011;Osbrink et al., 2021).An alternative explanation for this variation in response may be that cognitive and/or sensory differences between individuals generate variation in the ability to process and correctly respond to multiple sources of acoustic information (i.e.plane noise and alarm call).It is well established that tasks involving the discrimination and simultaneous processing of multiple signals are more difficult than tasks that only involve detecting a single signal (Dukas & Clark, 1995;Dukas & Ellner, 1993;Wollerman & Wiley, 2002).Magpies with better associative learning abilities may be better able to simultaneously perceive, discriminate, and appropriately respond to multiple sources of acoustic information, either through increased cognitive abilities, or heightened sensory acuity.

| CON CLUS ION
In this study, we show that anthropogenic noise significantly affects the foraging, vigilance, and vocalisation behaviour of magpies.We also provide the first evidence for the role of cognition in individual response to anthropogenic noise.We show that birds with better associative learning scores exhibit a more similar response to an alarm call playback and an alarm call played with anthropogenic (plane) noise, suggesting they are still able to recognise the alarm call in presence of anthropogenic noise, while birds with worse scores are less able to recognise and/or appropriately respond to an alarm call during noise.This ability may have important benefits for these individuals, and could have the potential to directly affect survival.Such findings highlight the potential role of cognition in enabling wildlife to rapidly adapt to environmental change, and suggests that variation in cognitive performance is linked to variation in response to stressors such as anthropogenic noise.Our study adds to the growing body of literature documenting the adverse effects of anthropogenic noise on wildlife, while also highlighting the importance of investigating causes of individual variation in response to anthropogenic noise.

F
Boxplots show median and quartiles for the distribution of (a) foraging effort (N = 333 focals), (b) foraging efficiency (N = 218 focals), (c) proportion of time spent vigilant (N = 333 focals) and (d) vocalisation rate, in each noise condition (LAN; loud anthropogenic noise, BN; background noise).Median and quartiles generated from predictions of the terms from the top model in Tables 1-3 and 5. F I G U R E 2 The proportion of time spent vigilant in relation to the interaction between spatial memory score and noise condition (loud anthropogenic noise: red line and points; background noise: blue lines and points) (N = 235 focals).Points represent raw data; fitted lines and 95% confidence intervals are from the output of the model presented in Table 4. TA B L E 2 Top model set of candidate terms influencing foraging efficiency.Data are based on 218 20-min focals across 70 magpies.Outputs Birds were significantly more likely to flee to cover following the alarm only playback compared to the alarm + plane and plane only F I G U R E 3 The proportion of time spent vigilant in relation to (a) playback treatment (N = 120 tracks) and (b) the interaction between playback treatment and associative learning score (high score indicates worse performance; N = 100 tracks).Median and quartiles generated from predictions of the terms from the top model in Table

F
The time (s) taken to return to normal behaviour in relation to (a) playback treatment (N = 120 tracks) and (b) the interaction between playback treatment and associative learning score (N = 100 tracks).Median and quartiles generated from predictions of the terms from the top model in Table

F
and plane Warble and plane Warble only Playback treatment Proportion of individuals fleeing to cover Top model set of candidate terms influencing foraging effort.
Note: Data are based on 333 20-min focals across 75 magpies.Outputs were generated from binomial GLMMs with group and bird identity as random terms.CI, confidence interval.For a list of all models tested, see Supporting Information (TableS1).
Data are based on 333 20-min focals across 75 magpies.Outputs were generated from binomial GLMMs with group and bird identity as random terms.CI, confidence interval.For a list of all models tested, see Supporting Information (TableS6).
TA B L E 4Top model set of candidate terms influencing time spent vigilant for individuals that had undergone cognitive testing.
Top model set of candidate terms influencing time spent vigilant in the 30 s post-playback.
Top model set of candidate terms influencing time taken to return to normal behaviour post-playback.Top model set of candidate terms influencing whether individuals fled to cover post-playback.
TA B L E 9 TA B L E 1 0Note: Data are based on 72 playbacks conducted on 24 magpies.