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

Dynamic in-flight flock organization

It is commonly assumed that during flocking, flock members follow three basic interaction rules: Attraction, Repulsion and Alignment, to coordinate spatial positions between each other¹⁸. To study the spatial organization of our zebra finch flock during flight, the spatial positions of all birds in the flight section were tracked in every fifth frame (sample rate: 24 Hz (that is, frames per second)) of the synchronized footage recorded by two high-speed digital video cameras (Camera 1: centred upwind view, Fig. 1a,b; Camera 2: upturned vertical view, Fig. 1a,c) for the entire duration (51.7, 58.3, 69.2 and 127 s) of four (session 2, 5, 8 and 13) out of 13 flight sessions. Flight paths were reconstructed from the tracking data for each bird in the flock, with horizontal and vertical coordinates delivered by Camera 1 and coordinates in wind direction delivered by Camera 2. The data show that each bird mainly occupied a particular area in the flight section, and that this spatial preference was stable over different flight sessions. Bird Green, for example, was preferentially flying very low above the flight section’s floor, and bird Lilac preferred to fly at upwind positions in front of the flock (Fig. 1d, Extended Data Figs. 1 and 3 and Supplementary Information).

Despite their preference in flight area, all birds constantly changed their spatial positions fast and rhythmically along the horizontal dimension of the flight section (Fig. 1e–g, Extended Data Figs. 2 and 4, Supplementary Video 1 and Supplementary Information). This behaviour is reminiscent of the flight behaviour of wild zebra finches: when being surprised in flight by a predator, zebra finches fly in a rapid zig-zag course low above the ground, heading for nearby vegetation¹⁶. Whether the sideways oscillating flight manoeuvres, which are performed by both wild birds in open space and domesticated birds in the wind tunnel’s flight section, are caused by the close proximity to the ground or are part of an escape reaction is yet unknown.

From the tracking data, we further calculated the spatial distances in all three dimensions between all pairwise combinations of birds throughout the four flight sessions (sample rate: 24 Hz). When normalized to the maximum distance detected for each bird pairing, each dimension and each flight session, mean distances of bird pairings in all dimensions were narrowly distributed within a range of 27.7–38.0% of maximum distance (Fig. 1h and Supplementary Table 1). This may indicate that during flocking flight, zebra finches actively balance Attraction and Repulsion to maintain a stable 3D distance towards all other members of the flock. Owing to the spatial limitations in the wind tunnel’s flight section, we did not expect the zebra finches to perform large-scale flight manoeuvres with movements aligned between all flock members (Extended Data Fig. 5 and Supplementary Information), as can be observed, for example, in freely flying flocks of homing pigeons (Columba livia domestica)¹⁹ and white storks (Ciconia Ciconia)²⁰.

Visually guided horizontal repositioning

When observing the dynamic spatial organization of our zebra finch flock, a question immediately arises: how do the birds prevent collisions during their frequent horizontal position changes? When considering the spatial limitation experienced by the flock of six birds during flight in the flight section and their highly dynamic flight style, collision rates seemed to be astonishingly low (median: 0.02 Hz; interquartile range (IQR): 0–0.03 Hz; n = 13 sessions) during flocking flight (in total 16 collisions in 13 min of analysed flight time). In birds, the visual system represents the main input channel for environmental information. To tackle the above question, we therefore first investigated the role of vision during flocking flight, and tested whether a bird’s viewing direction was correlated with the direction of horizontal position change. As gaze changes are governed by head movements in birds²¹, we used a bird’s head direction as an indicator for the orientation of its visual axis. We tracked (sample rate: 120 Hz) the position of a bird’s beak tip and neck in each frame of the footage during ten horizontal position changes (Fig. 2a and Supplementary Video 2) per bird, and found a strong interaction between a bird’s head angle relative to the wind direction and its direction of horizontal position change. During horizontal position changes, the birds always turned their heads in the direction of the position change (Fig. 2b). While the population’s median absolute angle of position change was 84.0° (IQR: 78.6–87.2°; n = 60) relative to 0° in wind direction, the population’s median absolute head turning angle was 36.0° (IQR: 26.4–42.5°; n = 60; see Supplementary Information for results on head movements during solo flight). The eyes of zebra finches are positioned laterally on their heads²² and each retina features a small region of highest ganglion cell density (fovea, that is, region of highest visual spatial resolution) at an area that receives visual input from horizontal positions at 60° relative to the midsagittal plane²³. By turning their heads by about 36° during horizontal position changes, the zebra finches roughly align the foveal area in the retina of one eye with their direction of position change, and in the retina of the other eye with the wind direction (Fig. 2c,d). Thus, head turns in the direction of position change may indicate that the birds use visual cues while repositioning themselves within the flock. This hypothesis is supported by a study on zebra finch head movements performed during an obstacle avoidance task. In this study, instead of fixating on the obstacle, zebra finches turned their head in the direction of movement while navigating around the obstacle²⁴.

Interestingly, birds usually turned their heads already before they initiated the position change (Fig. 2e). The delay between initiation of head turn and initiation of position change was variable within and between individuals, ranging from −16.6 ms (position change preceding head turn) to 736.6 ms (position change following head turn). A population median delay of 215.9 ms (n = 60 position changes) may provide sufficient time for a bird to visually evaluate whether the flight path is clear before initiating a horizontal movement. The large variability in the delay between head turn onset and position change onset opposes the hypothesis that head turning behaviour may only be a motoric byproduct of the position change, and may be needed to steer the bird’s body in the direction of position change.

Theoretical models that incorporate visual input to predict the pattern of collective movements generally assume fixed values for an individual’s visual field and consequently for the spatial area in which the individual is able to perceive visual information from conspecifics¹³. Our data demonstrate that a bird’s visual field during flight is not static, but visual range can be increased by head movements. A theoretical increase in perceptual range has been shown to affect the output of collective behaviour models¹³. Incorporating natural dynamics of visual ranges in these models may therefore result in even more realistic predictions of collective behavioural patterns.

In-flight vocal behaviour

Zebra finches are highly vocal birds, emitting thousands of communication calls per day²⁵. Two of the most frequent call types, the distance call and the stack call, are also uttered during flight¹⁶. While it has been suggested that distance calls are used to localize conspecifics, stack calls seem to convey information about a bird’s intention to execute a certain movement^16,26. Based on vocal signatures unique to each bird and each call type, calls can be used for individual recognition²⁷. To observe the individual vocal behaviour of birds flying in the wind tunnel, we equipped each zebra finch with a light-weight radio-telemetric microphone transmitter^28,29,30. During flocking flight sessions, the rate of vocal emissions in our zebra finches was generally low (median: 0.03 Hz; IQR: 0.02–0.07 Hz; n = 65 (13 sessions, 5 birds)). While most vocalizations were emitted during the first four seconds following take-off from the perch, vocalization count plateaued at a low level during the subsequent phase of sustained flight (Fig. 3a, top panel). We suggest that zebra finches lower their vocalization rate during flight to reduce predation risk. A high call rate in free flight makes them conspicuous to predators, whereas a high call rate while hidden in vegetation does not.

Comparable to the vocal behaviour of wild zebra finches¹⁶, at take-off from the perch our birds often emitted a stack call, which could be followed by a distance call shortly after. During sustained flocking flight, mainly stack calls (Fig. 3b) were emitted. To test for flocking flight specificity of calling behaviour, we radio-telemetrically recorded the individual vocal activity during four solo flight sessions per bird. As in flocking flight, calling activity in solo flight was maximal during the first four seconds after take-off. In contrast to flocking flight, however, birds flying solo in the flight section never vocalized during the phase of sustained flight (Fig. 3a, bottom panel). This indicates that the emission of calls during sustained flight in zebra finch flocks depends on the social context. During both flight phases, take-off and sustained flight, zebra finches called significantly less when flying solo (population mean ± standard deviation (s.d.): 0.29 ± 0.23 Hz and 0 ± 0 Hz for take-off and sustained flight, respectively; n = 19 sessions) than when flying in a flock (population mean ± s.d.: 0.47 ± 0.39 Hz and 0.04 ± 0.06 Hz for take-off and sustained flight, respectively; n = 65 sessions; Extended Data Fig. 6). Although calls were most frequently emitted during the take-off phase of the flight session, we restricted further analysis to stack calls emitted during the sustained phase of flight. The effect of take-off calls and of distance calls, which are very rarely emitted during the sustained phase of flocking flight, on flock organization still needs to be investigated.

In addition to the general social context, the spatial arrangement of birds in the flock might also affect a bird’s propensity to emit a call during flight in the flight section. We compared the spatial positions of calling birds at stack call onset with the spatial positions of their flock mates, and found that, indeed, at call onset the calling bird was located at the right, lower edge of the frontal part of the flock (Fig. 3c,d). For example, bird Green, the individual that emitted the majority of stack calls (60 out of 93) during flocking flight, was also most often located at the bottom edge of the frontal part of the flock (Fig. 1d and Extended Data Fig. 1).

Vocally guided vertical repositioning

To determine if call emissions during sustained flight are correlated with a bird’s flight behaviour, we tracked (sample rate: 24 Hz) the calling bird’s spatial position relative to its position at call onset in the synchronized footage of both cameras. Following the onset of every stack call (n = 93) emitted during the phase of sustained flight in 12 flocking flight sessions, we performed the tracking in every 5th of 25 consecutive frames, which covered a time period of 209 ms. Our analysis showed that in the calling bird, the onset of a stack call emission was followed by an upwards-directed vertical position change (Fig. 4a,b, Extended Data Fig. 7 and Supplementary Video 3). The distribution of movement directions of calling birds within 209 ms after call onset showed a significant directionality in the horizontal/vertical plane, with the mean direction of movement pointing 73.5° upwards (Fig. 4a), but not in the horizontal/wind direction plane (Fig. 4b). Interestingly, the number of upwards-directed position changes accompanied by a stack call was very low in comparison with the number of upwards-directed position changes (that is, position change of at least 24.88 arbitrary units (a.u.) in 209 ms) not accompanied by a call. On average, only 1 in 94 (n = 4 flight sessions) upwards-directed position changes was accompanied by a call, which opposes the hypothesis that call emission is only a byproduct of the motor act of moving upwards.

The propensity to move upwards was generally high in a bird that was flying at low vertical positions in the flight section (Extended Data Fig. 8). However, the propensity to emit a stack call prior to an upwards-directed movement depended on the spatial arrangement of birds in the flock during flight. While at the time of initiation of call-unaccompanied upwards movements the moving bird was evenly surrounded by its flock mates in all three dimensions (Extended Data Fig. 8c,d), at the initiation of call-accompanied upwards movements flock mates were clustered above, to the left and in the back of the calling bird (Extended Data Fig. 8a,b and Supplementary Information). Relative to the spatial position of an upwards-moving bird at the time of movement initiation, the bird’s flock mates were located significantly more left, higher and more downwind when the bird was calling than when the bird moved upwards without calling (Fig. 4c–e). We therefore conclude that the propensity of a bird to emit a call during sustained flocking flight depends on the coincidence of two factors: (1) the bird is positioned at the bottom, right edge of the frontal part of the flock; and (2) the bird intends to move upwards.

The visual field of zebra finches is characterized by a blind area of about 60° located behind the head²². In addition, zebra finch vision seems to be impaired in a large part of the dorsal visual field by an intravitreous structure, the pecten oculi²³. Furthermore, it has been shown that in some bird species, including zebra finches, vision is lateralized, with visual input from the right eye having higher behavioural relevance than visual input from the left eye^31,32. Based on this knowledge, we assume that zebra finches flying at low and upwind positions at the right side of the flock are sometimes not able to visually localize all their flock mates. When a bird in such a situation intends to change its spatial position within the flock in an upwards direction, it emits a stack call to actively communicate its intention to its flock mates.

If vocal communication between the upwards-moving bird and its flock mates indeed occurs, and the flock mates perceive and evaluate the information carried by the stack calls, they should show a behavioural response temporally correlated to the calls. To test whether flock mates show a reaction to stack calls in their flight behaviour, we compared the flight behaviour of flock mates of birds moving upwards after calling with the flight behaviour of flock mates of birds that moved upwards without calling. We found that in contrast to flock mates of birds moving upwards without calling, call emission of an upwards-moving bird reduced movement activity in its flock mates (Fig. 4f,g). Within a time period of 209 ms after call onset and initiation of upwards movements, flock mates of calling and upwards-moving birds travelled over significantly shorter distances (population median and IQR: 64.1 and 37.9–116.6 a.u., and 104.1 and 50.6–181.8 a.u. for the horizontal/vertical plane and the horizontal/wind direction plane, respectively) than flock mates of birds that moved upwards without calling (n = 2,637 and n = 2,284, respectively; population median and IQR: 87.1 and 51.1–145.2 a.u., and 145.3 and 77–264.2 a.u. for the horizontal/vertical plane and the horizontal/wind direction plane, respectively). This suggests that flock mates react to the calls by retaining their spatial position in the flight section, probably to observe the calling bird’s movement and thus to reduce collision risk.

For both cases—call-accompanied and call-unaccompanied upwards movements—the flight trajectories of the moving bird’s flock mates showed a significant directionality in both the horizontal/vertical plane and the horizontal/wind direction plane (Fig. 4h–k). Although in both conditions the main movement direction of an upwards-moving bird’s flock mates was opposite to the upwards-moving bird, when the bird called before moving upwards a considerable amount (8.3%) of the calling bird’s flock mates aligned their movements with those of the calling bird and moved towards positions between 75° and 105° upwards (Fig. 4h). Interestingly, these straight upwards-directed movements were particularly underrepresented in flock mates of birds moving upwards without calling (Fig. 4j). This suggests that in some of the upwards-moving bird’s flock mates, the combination of an upwards movement with a stack call, but not the upwards movement alone, causes an upwards-directed movement and consequently an alignment of movement direction with the calling bird.

Complementary role of vision and vocal communication

To directly test for the effect of deficient visual input on the vocal activity and flight behaviour of zebra finches in flocking flight, we reduced the illumination in the flight section (initially 200 lx) and recorded the individual vocal behaviour of zebra finches during ten flocking flight sessions at low (20 lx, comparable to illumination levels during civil twilight) and during ten flocking flight sessions at very low (0.2 lx, comparable to the illumination of a clear night sky at full moon) illumination. While wild zebra finches may occasionally fly under 20 lx illumination levels, flying under very low illumination levels (0.2 lx), which in nature are present only during the night, is a very artificial situation for the diurnal zebra finch. Our experiment showed that while during the take-off phase the call rate was not affected by the illuminance level, during sustained flight birds called significantly more often when flying under very low light levels (population mean call rate ± s.d.: 0.06 ± 0.07 Hz; n = 60) than when flying under low light levels (population mean call rate ± s.d.: 0.03 ± 0.06 Hz; n = 60; Fig. 5a,b). Interestingly, the rate of collisions between birds did not differ between illuminance levels (Fig. 5c). The elevated call emission rates at very low ambient light levels in combination with the constant collision rates suggest that zebra finches may be able to compensate for deficient visual information by increasing the usage of vocal communication to coordinate their spatial positions in flying flocks.

Finally, to investigate the importance of vocal communication for flock coordination, we manipulated the birds’ ability to communicate vocally during flight. In zebra finches, background noise can affect the call detection and discrimination ability, especially when the noise energy is within the calls’ spectral region³³. Stack and distance calls of zebra finches have their main energy roughly between 2,000 and 5,000 Hz (ref. ¹⁶). Via a loudspeaker positioned on the floor of the flight section, we introduced additional noise to the flight section during flight sessions (Supplementary Information). We measured call emission and collision rates during ten flocking flight sessions carried out in the presence of band-pass filtered (cut-off frequencies: 1,500 and 8,000 Hz) additional noise that completely masked the birds’ calls (Extended Data Fig. 9) and compared them with call emission and collision rates measured during ten flocking flight sessions carried out in the presence of low-pass filtered (cut-off frequency: 2,500 Hz) additional noise that did not mask the birds’ calls. While the additional noise did not affect the call rate during the take-off phase, during the phase of sustained flight birds called significantly less when flying in the presence of masking noise (population mean call rate ± s.d.: 0.004 ± 0.01 Hz; n = 60) than when flying in the presence of noise that did not mask their calls (population mean call rate ± s.d.: 0.024 ± 0.038 Hz; n = 60; Fig. 5d,e). Collision rates (see Supplementary Video 5 for an exemplary collision event) were significantly higher during flight in the presence of masking noise (mean collision rate ± s.d.: 0.054 ± 0.017 Hz; n = 10) than during flight in the presence of noise not masking the calls (mean collision rate ± s.d.: 0.04 ± 0.012 Hz; n = 10; Fig. 5f).

We conclude that not being able to vocally communicate with each other has a strong effect on the coordination of spatial positions of individuals in a zebra finch flock during flight. Deficits in social information influx resulting from not being able to see each other, however, can be compensated for by increasing the usage of vocal communication to reduce collision risk in the flying flock.

Source: Ecology - nature.com