Vision and vocal communication guide three-dimensional spatial coordination of zebra finches during wind-tunnel flights

Animal collective motion is a natural phenomenon readily observable in various taxa. Although theoretical models can predict the macroscopic pattern of group movements based on the relative spatial position of group members, it is poorly understood how group members exchange directional information, which enables the spatial coordination between individuals during collective motion. To test if vocalizations emitted during flocking flight are used by birds to transmit directional information between group members, we recorded vocal behaviour, head orientation and spatial position of each individual in a small flock of zebra finches (Taeniopygia guttata) flying in a wind tunnel. We found that the finches can use both visual and acoustic cues for three-dimensional flock coordination. When visual information is insufficient, birds can increasingly exploit active vocal communication to avoid collisions with flock mates. Our study furthers the mechanistic understanding of collective motion in birds and highlights the impact interindividual vocal interactions can have on group performances in these animals.


Supplementary Text
Birds' spatial preference in the flight section Although the birds occupied the entire width of the flight section, they generally preferred to fly left to its horizontal center. In contrast to the right side of the flight section, the left-side walls 5 were transparent, and the birds' home aviary was visible through the left-side walls. Both the difference in wall transparency and the view towards the home aviary could explain the bird's tendency to fly on the left side in the flight section.
Position change frequencies during solo flight 10 To test for the dependency of the birds' behavior on the social context, we reran the experiment and interspersed flocking flight sessions with flight sessions in which each bird flew solo (four sessions per bird) in the wind tunnel's flight section. While all birds maintained their preferred flight area in the flight section when flying solo (Extended Data Fig. 3), they showed an even more dynamic flight behavior than during flocking flight (Extended Data Fig. 4). The frequency 15 of rhythmic horizontal position changes during solo flight (group mean ± STD: 0.42 ± 0.09 Hz, (i.e., cycles per second), n = 24) was significantly higher (Linear mixed model (LMM), estimates ± SE: 0.07 ± 0.02, p = 0.002, t = 3.33) than in the flocking flight condition (group mean ± STD: 0.35 ± 0.12 Hz, n = 24). And the frequency of rhythmic vertical position changes during solo flight (group mean ± STD: 0.19 ± 0.21 Hz, n = 24) was also significantly higher (LMM, 20 estimates ± SE: 0.14 ± 0.04, p = 0.002, t = 3.27) than in the flocking flight condition (group mean ± STD: 0.05 ± 0.09 Hz, n = 24). A large variability in flight position was also observed in homing pigeons (Columba livia) when flying solo, and has been suggested to aid in predator avoidance when flocking is not possible 38 . 25 Alignment of movements between flock members When flying in the wind tunnel, the birds' main direction of motion was forced by the air flow and the main flight direction was therefore artificially aligned amongst all flock members throughout the flight session. Because the birds moved against the air flow, their position in wind direction roughly remained constant, which prevented the camera system to capture any 30 alignment of motion. Significant positional changes in wind direction were only rarely performed. Except for alignment in the main flight direction, we did not expect the zebra finches to align movement directions among all members of the flock. A correlational analysis of movement directions in the footage of Camera 1 revealed that for all possible pairwise combinations of birds, the degree of coordination of movement directions among two individuals 35 constantly alternated between periods of alignment (correlation coefficient > 0) and periods of misalignment (correlation coefficient < 0). Synchronous with events of alignment in one or more pairs of birds, birds of other pairwise combinations moved in directions opposite to each other  However, in each of the four flight sessions, at least one short time period was detected, during which movement directions were significantly aligned among the majority of individuals in the flock (i.e., mean correlation coefficient above 0.3, mean P value below 0.05; Extended Data Fig. 5). We did not detect specific behaviors correlated with these brief events of significant alignment and therefore suggest that they occurred by chance. To determine whether the birds 45 are able to intentionally align their movement directions, additional experiments are necessary during which birds in the flock are forced to change their movement directions collectively.
Head turning behavior during solo flight It is important to note that head turning behavior during horizontal position changes was independent of the social context, and was observed in both flocking flight and solo flight sessions (see Supplementary Video 3 for examples of horizontal position changes during solo flight). Neither the mean angle of absolute position change, nor the mean absolute turning angle 5 of the birds' heads, nor the mean delay between initiation of head turn and initiation of position change differed significantly (LME, p > 0.4, t < 0.8 for all measurements, n = 60 per condition) between conditions. This suggests that while flying, zebra finches use visual cues to avoid collisions with both moving flock members and static obstacles, such as the flight section's walls. 10 Spatial arrangement of the flock at the initiation of upwards directed movements In all three dimensions, horizontal, vertical and wind direction, the relative distance between a bird that was moving upwards without calling and its flock mates (n = 2637 for azimuth and elevation, and n = 2284 for wind direction) at the time of movement initiation was significantly 15 (LMM, estimates ± SE: 80.5 ± 19.9, p < 0.001, t = 4.04, estimates ± SE: -127.2 ± 12.5, p < 0.001, t = 10.17, and estimates ± SE: 105.5 ± 37.9, p = 0.005, t = 2.78, respectively) different from the relative distance between a bird that was moving upwards after emitting a call and its flock mates (n = 230 for azimuth and elevation, and n = 202 for wind direction). To exclude the possibility that call emission is solely triggered by the flock arrangement, we counted events 20 within the four fully tracked flight sessions during which the median position of flock mates was further left, further up and further downwind to the position of a focal bird. The average ratio between the number of Stack call emissions and the number of specified flock arrangement events was 10. 25:135 (2:67, 7:112, 14:153, 18:208). This means that on average only 1 in 14 of these particular flock arrangement events could have been associated with a Stack call, which 25 opposes the hypothesis that a bird calls whenever its flock mates are in a particular position relative to this bird.

Effect of wind noise on vocal communication
The peak frequencies of the wind noise measured at a wind speed of 10 m/sec in the flight 30 section were below 1000 Hz, and the high-frequency cut off at half amplitude was below 2000 Hz (Extended Data Fig. 9a). In the frequency range between 2000 and 5000 Hz, the mean amplitude of Stack calls measured at a distance of 50 cm in front of zebra finches perched in the calm flight section exceeded by up to 10 dB the mean of wind noise amplitudes measured at nine different spatial positions in the flight section at a wind speed of 10 m/sec (Extended Data Fig.   35 9a). Therefore, we expect the masking effect the wind noise in the flight section had on Stack calls emitted during flight at a wind speed of 10 m/s to be moderate, and comparable to the effect wind noise has on calls of zebra finches flying at a speed of 10 m/sec in free space. 24 (4 per bird) 6 (1 per bird) -19 - At the beginning of the clip, the video is displayed at normal speed, and subsequently the same sequence is displayed slowed down ten times for better visibility of the bird's behavior. Bird Green is marked by a red 20 circle at the onset of the call it emitted before moving upwards. The audio signal of bird Green's microphone transmitter has been added to the video clip as sound trace.

Lux
Supplementary Movie 5 | Collision between birds during flight in the presence of masking noise. The video clip with footage synchronously recorded by Camera 1 (left screen) and Camera 2 (right screen) shows two birds colliding during flocking flight from two different