Animals and housing conditions
In January 2019, we captured 12 adult Australian magpies (six birds of each sex, based on plumage) (Kaplan, 2019) in the City of Melbourne, Australia, using a walk-in trap baited with grated cheese. All magpies came from non-breeding groups in urban parklands in the Melbourne suburb of Parkville. The average sound level experienced by these groups over a 24 hour period was approximately 50 decibels (see Connelly et al., 2022 for details regarding sound level recordings). Magpies were transported to an indoor animal holding facility at La Trobe University where they were housed in two experimental rooms with similar configurations (3 females/males per room). Magpies were housed individually in aviaries 1.8 m high x 1.8 m deep x 0.9 m wide which were left uncovered, allowing the magpies to both see and hear one another. Each aviary contained three perches: two rectangular plank perches (15 cm wide), one 130 cm and the other 45 cm above the ground, and a dowel perch 45 cm above the floor. Each aviary contained two video cameras with infrared capabilities (one mounted above the 130 cm perch where the birds regularly slept; the other on the door of the aviary). The camera on the door focused on the front perch to record testing. Magpies were fed a mixture of minced meat and an insectivore mix (55 g; Wombaroo Food Products, Australia) once per day (at c. 0900 h); on test days, food was not provided until after the testing period (1130 h). Clean water was given daily in a large bowl, providing the magpies with a place to both drink and bathe. Aviary floors were covered by woodchips and, to provide enrichment, 15–20 mealworms were scattered daily throughout the woodchips, giving the magpies the opportunity to forage. Rooms were temperature-controlled (22 ± 5º C) and insulated from all external light. Room lighting (153 ± 18 lux) was set to a 12-h light (0600–1800 h), 12-h dark (1800–0600 h) photoperiod. Night lights (~ 0.1 lux at sleeping perch) were placed in each room so that the magpies could move at night without harming themselves, and because true darkness in the wild is unrealistic for most birds (Aulsebrook et al., 2022).
Magpies were habituated to human experimenters via daily interactions over their first month in captivity. At the end of this period, the birds no longer reacted evasively to our presence, and in some instances could be fed by hand. Magpies also habituated to one another, with birds in both rooms singing together (Johnsson, et al., 2022A). Weight was monitored throughout captivity to ensure good health. Magpies underwent surgery (approx. two months prior to this study) to implant sensors for recording brain activity for unrelated sleep experiments (Aulsebrook et al., 2020A; Connelly et al., 2020; Johnsson, et al., 2022A; Johnsson et al., 2022B). To determine whether the surgical procedure altered cognitive performance, we presented the birds with a simple motor test (same test board as Spatial Memory, see Test Battery below) before and after surgery; there was no difference in the birds’ ability to perform the task between the two time-points. This experiment was conducted approximately three months after initial capture.
Experimental design
We used a repeated measures design to compare performance on the cognitive test battery in the presence and absence of anthropogenic noise (Fig. 1). While birds in one room received the noise treatment, birds in the other room received the (quiet) control. During the treatment trial, the noise playback started 48 h before testing began. Birds were tested for 3 h each day (0830–1130 h), during which the playback was muted to prevent unwanted distraction due to the noise. Once all birds completed the test battery (completed all tests or ceased testing), the playback was switched off and birds had seven days to recover. Following the week-long recovery, the treatment assigned to each room was switched and the experimental protocol commenced anew. Overall, 9 out of 12 birds participated in this experiment (females: n = 4; males = 5), each bird testing in both the control (no noise) and noise treatments.
For the treatment, we broadcasted a recording of urban noise for the entire 12-h night and remnant 9-h day outside of the 3-h testing period. Noise for the playback was recorded over one week (Monday – Sunday) at a busy roadside located near the Melbourne Central Business District, providing a realistic city soundscape (Figure S1). The time signatures of the playback and the experiment were synchronized, and sunrise/sunset times mismatched by less than 15 and 50 minutes respectively, such that birds experienced realistic noises throughout each day/night period. An omni-directional speaker (Ultimate Ears, Boom 2) was placed in the centre of each room to broadcast the playback for the noise treatment. It should be noted that these speakers may not have accurately reproduced all low frequency sounds (< 90 Hz) or vibrations. A sound level logger (Sound Level Meter Data Logger NSRT mk3; Convergence Instruments) was placed in each room, 3.5 m from the speaker to measure the sound levels represented by A-weighted decibels (dBA) from a single point in the room throughout each treatment. The average dBA level (hourly) of the playback over the testing period was 62 dBA (± 6.22), with noise levels ranging between 40 and 90 dBA throughout each 24-h period (Figure S1). The average sound level (62 dBA) was well below the safe threshold for humans and within the range of noise pollution found in metropolitan areas (Brown et al., 2015). In this way, the recording was an accurate representation of what urban magpies experience in the wild (Connelly et al., 2022). The background noise in the control treatment was 44 dBA (± 2.67), ranging between 42 and 63 dBA, and composed of constant low-level noise generated by ventilation and air-conditioning systems, and high-level noise associated with magpie carols and calls (Johnsson et al., 2022A). As intended, the soundscape of the room was louder when the recording of urban noise was played compared to when it was not (t = 38.2, p < 0.01).
Cognitive testing & sample sizes
Prior to the start of the experiment, we used a motor skill task or ‘training board’ to train the magpies to interact with the test battery. The training board consisted of a wooden block (30 cm long x 9 cm wide x 4 cm high) with four equally-spaced wells (3.0 cm deep, 3.3 cm diameter; Fig. 2). Covering the wells were four coloured plastic caps (two yellow and two grey). The plastic caps were held in place by a rubber band that looped around nails in the board. This set-up provided an axis on which the magpies could rotate the lids to access the wells. A few mealworms (3–5) were placed in each well as a reward. Magpies were presented with the training board for several days, until they successful ate out of each well. The training board, as well as all other tasks, was positioned on the plank perch in the front of each aviary for testing.
We used a cognitive test battery to quantify differences in cognitive performance between noise treatments. Each magpie was tested daily during the three hour testing period and presented with a task (typically) two times per day but dependent on motivation. Birds took on average 8 ± 2 days to complete all tasks. Testing was considered finished once all birds in each treatment had either completed all tasks or had ceased testing (i.e., three days without interacting with a task). Magpies were tested in their respective aviaries with curtains hung between cages to ensure testing occurred in isolation.
The cognitive test battery consisted of four main tasks (see Test battery for details). Tasks were always presented in the same order, and each tested a well-defined cognitive function of presumed ecological relevance:
(1) Associative Learning: The ability to acquire knowledge through repeated experiences (Morand-Ferron et al., 2016);
(2) Reversal Learning: The ability to learn a new rule, a measure of cognitive flexibility (Bond et al., 2007);
(3) Inhibitory Control: A measure of self-control based on the inhibition of immediate responses (Isaksson et al., 2018; Kabadayi et al., 2016, 2018);
(4) Spatial Memory: The ability to remember the location of resources or threats (Emery, 2006).
Tasks 2 and 3 might be indicative of executive functions (top-down cognitive processes exerting control over information processing; Bobrowicz and Greiff, 2022). A new task could begin immediately after the previous task was completed, with the exception of reversal learning, which always started one day after the completion of the associative learning task. The criterion for success varied between tasks, but in all tasks the total number of trials required to fulfil the success criterion acted as the test score. Therefore, fewer trials taken to reach their task-specific success criterion indicated a bird performed ‘better’ (see Test battery for details). Importantly, magpies were never forced to test; tasks only began when a magpie was motivated enough to approach and eat a mealworm. The training board was functionally identical to the associative learning, reversal learning, and spatial memory tasks, such that birds did not need to be further trained to operate those tests. Sample sizes varied among tests – associative learning: n = 7; reversal learning: n = 7; inhibitory control: n = 9; spatial memory: n = 9 owing to some birds not interacting with, or completing, the task.
Because birds were presented with each task twice in the repeated measures design, minor aspects of the task were modified between the treatments to introduce novelty and test for genuine learning while allowing the pair-wise comparison of scores. For instance, in the associative learning task, the colour of the caps covering the wells was blue for the noise treatment and purple for the control, thus providing a comparable test, but with a novel colour.
Test battery
Associative Learning: We used a colour-discrimination task to test how quickly magpies could learn to associate a novel-coloured cap with a reward. The foraging grid (19 cm long x 9 cm wide x 4 cm high) contained two wells (3.0 cm deep, 3.3 cm diameter; Fig. 2). Each well was covered with a cap of a different shade of the same colour (i.e., navy and sky blue; violet and lilac purple). Shades of the same colour were used instead of different colours to minimize any potential effects of colour preference. One well contained a chilled (non-moving) mealworm reward; the other well was empty. The colour of the rewarded cap was randomized. In the first trial, magpies were allowed to peck both caps in order to find the reward under the correct cap. In all consecutive trials, pecking the incorrect cap cued the removal of the board for one minute. Birds had a maximum of one minute to interact with the board before it was removed from the testing area. Testing resumed only when birds showed interest in interacting with the test boards again (i.e., by returning to the ‘testing area’ or lower perch. The position of the rewarded cap (left or right) was pseudo-randomized, meaning we avoided having the reward on the same side for more than three consecutive trials. The task was considered complete when the subject pecked the correct cap (obtaining the reward) in 10 out of 12 consecutive trials; a significant deviation from random binomial probability (binomial test: p = 0.04; same criterion as in Ashton et al., 2018; Connelly et al., 2022; Johnsson et al., 2022A). If the bird did not reach the criterion in a single day the score would carry over to the next day and testing would continue. The total number of trials required to reach this criterion (including the final 12 trials) represented the associative learning score. To prevent olfactory cues, both wells were wiped with chilled mealworms prior to the start of testing each day (similar to Ashton et al., 2018). To maintain novelty, the colour of the caps was different for each treatment.
Reversal Learning
The reversal learning task tested cognitive flexibility, quantifying the number of trials required for a magpie to dissociate the previous association and learn a new rule (Fig. 2). Reversal learning used the same board as associative learning; however, the shade of the of rewarded cap was switched (i.e., if the reward cap was navy for the association learning task, it would be sky-blue for the reversal learning task). The experimental protocol and completion criterion were the same as for the associative learning task.
Inhibitory Control
There were two inhibitory control tasks, and both quantified the magpies’ ability to inhibit predominant responses, and efficiently navigate around an obstacle to achieve a goal (in this case, to obtain a food reward). Here, the predominant response would be for the bird to try to retrieve a reward by pecking at a transparent plastic barrier placed in front of a food reward. We presented magpies with two such ‘detour reaching’ tasks that were visually distinct, yet functionally similar. In the first task, a food reward (1–3 mealworms) was placed inside a transparent open-ended cylinder (13 cm long x 5 cm diameter) mounted on a small wooden block (13 cm long x 6.5 cm wide x 5 cm high) (Fig. 2). The cylinder was presented perpendicular to the subjects’ body axis (i.e., open ends were positioned out of the magpies’ eyeline), to see if they could solve the challenge of accessing the reward. In the second inhibitory task, the cylinder was replaced with a transparent plastic wall (34 cm wide x 13 cm high) mounted vertically onto a wooden block (24 cm long x 9 cm wide x 4 cm high; Fig. 2). The same reward was placed behind the wall.
For both tasks, when first presented, the subject was allowed to explore the test board and find their reward. Each trial thereafter counted towards their overall score. A trial was considered successful if the subject did not peck the closed side of the cylinder or the plastic wall, and instead walked to the open ends of the cylinder, or around the wall, to obtain the mealworm. If the bird pecked the cylinder or wall, an incorrect score was tallied, and the test was removed from the testing area for one minute. The task was considered complete when the subject correctly detoured (without pecking) around the task to obtain the reward three consecutive times. Following and to be consistent with established protocols (Ashton et al., 2018; Connelly et al., 2022) magpies were allowed a maximum of 10 trials per day on the detour tasks, and trials were conducted in one-minute intervals. Test results for individuals on the two detour tasks were not statistically different (t = 0.91, p = 0.38) and we therefore tallied the number of trials taken to pass both these tasks into a single ‘inhibitory score’ per individual for the analysis.
After completion of the inhibitory testing, we presented the magpies with a ‘control’ task to determine if the birds were truly inhibiting the urge to peck and not simply learning to take a route around the barrier (an associative learning function). These tasks consisted of the same cylinder or wall (Fig. 2), but with a hole cut out of the centre allowing the magpies to access the reward without detouring around the obstacle. A magpie successfully completed a trial if they retrieved the reward through the cut hole instead of navigating around the obstacle. The success criterion was the same as for the inhibitory tasks.
Spatial Memory: The spatial memory task was used to determine a magpie’s ability to store location-based information for later retrieval. The task consisted of a wooden board (45 cm long x 9 cm wide x 4 cm high) containing an array of 1 x 4 wells (3.0 cm deep, 3.3 cm diameter; Fig. 2). This was modified from 2 x 4 x 2 well system used by previous studies (Ashton et al., 2018; Connelly et al., 2022) to better fit the physical constraints of the captive testing environment. Each well was covered with a lime green cap and used the same flipping mechanism as the associative and reversal learning tasks. Wells were numbered (1–4) from left to right. Each individual was randomly assigned a well number and that well contained the food reward for the entirety of the test. Magpies were presented with the test board five times over three days: twice on the first day, once on the second, and twice on the third (as per Connelly et al., 2022). Each testing ‘cycle’ was deemed complete when the magpie found the rewarded well. The first cycle (or baseline) had the magpie presented with the board to find the reward. Once found, the board was removed. After five minutes the same well was rebaited, and the second cycle (or training) began, following the same protocol. The third and fourth cycles were presented in an identical way, 24 h and 48 h later, respectively. The total number of trials required to find the food reward in the third and fourth trials determined a magpie’s spatial memory score. The well assigned with the food reward was different for each bird in each treatment. A final trial (or ‘smell’ trial) was performed five minutes after the fourth trial. For the smell trial, the board was rotated 180 degrees and no reward was placed in the wells. The foraging grid appeared identical to the subject, but the previously baited well was on the opposite side of the test board. This test was conducted to assess whether magpies might be using olfactory cues to determine the location of the baited well.
Statistical analysis
All analyses were conducted in the statistical environment R version 3.5.2 (R Development Core Team 2018). Using the R package lme4, we applied generalized linear models (GLMM) with a Poisson error structure and logarithmic link to investigate the factors affecting performance on all tasks. The GLMM contained all probable explanatory values: treatment (noise or control), treatment order (noise first or second), test order (first or second trial), sex, and the interaction of the test type (associative learning, reversal learning, inhibitory control, spatial memory) and treatment (Table 1). Random effects were bird identity and housing room (one or two). Paired t-tests were used to compare performances on individual tests.