Birds often fly in flocks ranging from very structured V-formations to unstructured clusters. Many studies have tried to prove what causes birds to flock and how it benefits them. Flocks, for example, may help birds to avoid predators and to navigate. Flying in a V-shaped formation likely also gives aerodynamic benefits that can make it easier to fly long distances.

Few studies, however, have measured how the positions of birds in a flock relate to things like flying speed or the frequency of wing flaps. This is because it was difficult to take such measurements in large flocks of moving birds. Advances in cameras and computers are now making it easier to track individual birds flying in large flocks. The technology allows scientists to measure how birds position themselves in relation to other birds, or how flock-positioning varies by bird size, species, ecology, and behaviors. Such measurements may help scientists better understand why and how birds flock.

Corcoran and Hedrick now show that four different types of shorebirds position themselves in the same way when flying in a flock. In the experiments, digital cameras recorded video of 18 cluster-like flocks of four different species of birds flying over a bird sanctuary or agricultural fields. The flocks ranged in size from a hundred to a thousand birds. Some flocks had two types of bird. The four types of birds – dunlin, short-billed dowitcher, American avocet, and marbled godwit – live in similar environments but greatly vary in size and fly at different speeds. Corcoran and Hedrick measured individual bird positions using three-dimensional computer reconstructions of the flocks.

Each bird – regardless of size or species – most commonly flew about one wingspan to the side and between a half to one-and-a-half wingspans back from the bird in front of it. Birds flying in simple V-shaped formations follow similar rules. This suggests that birds flying in clusters may also gain aerodynamic benefits.

The collective movements of animals—from schooling fish to swarming insects and flocking birds—have long excited intrigue among observers of nature. Collective motion arises as an emergent property of interactions between individuals (reviewed by Herbert-Read, 2016 and by Vicsek and Zafeiris, 2012). Thus, much attention has been placed on identifying local interaction rules (Ballerini et al., 2008a; Herbert-Read et al., 2011; Katz et al., 2011; Lukeman et al., 2010) and how those rules affect group structure and movement (Buhl et al., 2006; Hemelrijk and Hildenbrandt, 2012). However, comparative data across species are still limited, preventing us from testing hypotheses regarding the evolution and diversity of collective movement patterns.

Hundreds of bird species fly in groups, but most quantitative research has focused on starlings (Attanasi et al., 2014; Ballerini et al., 2008b; Cavagna et al., 2010), homing pigeons (Nagy et al., 2013; Nagy et al., 2010; Pettit et al., 2015; Pettit et al., 2013; Usherwood et al., 2011) and birds that fly in V-formations (Badgerow and Hainsworth, 1981; Cutts and Speakman, 1994; Hummel, 1983; Lissaman and Shollenberger, 1970; Maeng et al., 2013; Portugal et al., 2014; Weimerskirch et al., 2001). These data indicate that smaller birds fly in relatively dense cluster flocks that facilitate group cohesion and information transfer (Attanasi et al., 2014; Ballerini et al., 2008a), whereas larger migratory birds fly in highly structured V formations (also known as line or echelon formations) that provide aerodynamic and energetic benefits (Lissaman and Shollenberger, 1970; Portugal et al., 2014; Weimerskirch et al., 2001). However, descriptive accounts of flock structure over a greater range of species (Heppner, 1974; Piersma et al., 1990) cover a range of flock types, spanning the extremes of V-formation and large cluster flocks. The species whose flocking behavior have been studied quantitatively differ in many ways that could be important for flocking, including body size, ecology, the frequency of aggregation and its behavioral context. Therefore, on the basis of the available data, it is difficult to conclude what factors cause birds to adopt a specific group formation, or even what factors affect interaction rules, positioning and behavior within flocks.

We aimed to address these questions by collecting three-dimensional (3D) trajectories of the birds in flocks of four shorebird species that have similar ecologies (all forage in large groups in coastal habitats and migrate long distances) but that cover a seven-fold range of body mass and two-fold range of wingspan. Our study species include dunlin (Calidris alpina Linnaeus 1758; 56 g, 0.34 m wingspan), short-billed dowitcher (Limnodromus griseus Gmelin 1789; 110 g, 0.52 m wingspan), American avocet (Recurvirostra americana Gmelin 1789; 312 g, 0.72 m wingspan), and marbled godwit (Limosa fedoa, Linnaeus, 1758; 370 g, 0.78 m). Molecular dating indicates that these species diverged from their nearest common ancestor approximately 50 million years ago (Mya) (Baker et al., 2007), providing time for evolutionary diversification of flocking behavior. By comparing the group structure of birds across a range of body sizes and by comparing our data with those in the literature, we aimed to determine the extent to which flock structure varies across species with different body sizes and ecologies. We employ three approaches: (1) identification of local interaction rules by quantifying the relative positions of birds and their nearest neighbors; (2) quantification of the degree of spatial structure within flocks; and (3) measurement of individual speeds and wingbeat frequencies to examine how local and global position within the flock affect flights behavior.

On the basis of existing flock data, we hypothesized that flocks of larger shorebird species would be more structured than those of smaller species (recapitulating the trend of larger birds flying in highly structured V formations) and that larger species would also exhibit aerodynamic formations more frequently. Because a previous study showed that flying in a cluster flock is energetically costly in pigeons (Usherwood et al., 2011), we hypothesized that birds flying in the middle and rear of flocks and birds flying closer to their nearest neighbor would have reduced flight performance (lower speed relative to their wingbeat frequency). Surprisingly, we found that all four species studied here flew in a flock structure that we term the compound-V formation. We propose that this structure might be an adaptation for aerodynamic flocking in migratory species, and that ecology is an underappreciated driver of the evolution of avian flocking behavior.

We reconstructed the 3D trajectories from 18 bird flocks that ranged in size from 189 to 1039 individuals, which were recorded for 2.4–13.2 s at 29.97 frames per second (Figure 1, Table 1). This resulted in 1,598,169 3D position measurements that were used to examine flock structure. Sixteen of the 18 flocks were comprised entirely of a single species. The remaining two flocks were mixed-species flocks of marbled godwits and short-billed dowitchers. Computer vision techniques allowed the species of individuals in mixed-species flocks to be identified on the basis of differences in body size (see 'Materials and methods').

Figure 1

Download asset Open asset

Figure 1—source data 1

Three-dimensional trajectory data for godwit flock 0420–1.

https://doi.org/10.7554/eLife.45071.004

Download elife-45071-fig1-data1-v1.zip

Nearest-neighbor alignment

We examined flock structure by quantifying the position of each bird with respect to its nearest neighbor. We used modal values to characterize typical neighbor positions because position distributions were skewed as a result of values being cropped at zero. In all flocks, nearest neighbors flying within the same horizontal plane [an elevation slice of ±1 wingspan, a mean of 56% of nearest neighbors across all flocks (range 35–76%)] exhibited a distinctly peaked distribution, where modal neighbor position was offset both in front-back and lateral distance (Figure 2a,b). By contrast, nearest neighbor birds flying outside the horizontal elevation slice of ±1 wingspan were distributed randomly with a peak directly above or below the focal bird (Figure 2c,d). This indicates that shorebirds adopt alignment rules for neighbors flying within their same elevation slice. On average, birds flew at approximately the same height as their nearest leading neighbor (−0.01 ± 0.02 m, mean ± s.d. for the median trailing height across all 18 flocks).

Figure 2

Download asset Open asset

Figure 2—source data 1

Nearest-neighbor positioning data for flock 0420–1.

https://doi.org/10.7554/eLife.45071.008

Download elife-45071-fig2-data1-v1.zip

Both nearest-neighbor lateral distance and front-back distance differed among flocks and species (Figure 3a). Species wingspan strongly predicted modal lateral neighbor position (linear regression, slope = 0.85, R² = 0.93, F = 228.29, p<0.0001). Wingspan also predicted front-back distance (slope = 0.70, R² = 0.86, F = 99.57, p<0.0001), although less strongly than lateral distance. After scaling alignment positions to wingspan (i.e., dividing neighbor distances by species wingspan), a distinctive pattern emerges (Figure 3b). Specifically, the flocks adopted a modal lateral distance of approximately one wingspan (mean 1.04, range 0.88–1.24 wingspans). This non-dimensionalized lateral distance had a weak inverse relationship to species wingspan (linear regression, slope = −0.37, R² = 0.37, F = 9.38, p=0.007) and was not related to flock density (i.e. nearest neighbor distance, non-dimensionalized by wingspan; linear regression, R² = 0.07, F = 1.15; p=0.30). Non-dimensionalized trailing distance was inversely proportional to species wingspan (linear regression, slope = −0.40, R² = 0.33, F = 7.93, p=0.012) and increased with non-dimensional flock density (linear regression, slope = 0.13, R² = 0.58, F = 22.39; p=0.0002). In summary, across all four species, shorebirds adhere to a non-dimensional spacing rule of aligning to neighbors with a lateral offset of approximately one wingspan while allowing trailing distance to vary with flock density.

Figure 3 with 2 supplements see all

Download asset Open asset

Figure 3—source data 1

Modal neighbor positioning data for all shorebird flocks and for a flock of chimney swifts.

https://doi.org/10.7554/eLife.45071.012

Download elife-45071-fig3-data1-v1.zip

Data from mixed-species flocks of godwits and dowitchers further support the non-dimensional nature of the lateral spacing rule within individual flocks. Both dowitchers and godwits adjusted their lateral spacing depending on the species of their neighbor (Figure 4). Godwits following conspecifics had a modal lateral spacing of 0.76 m, or 0.97 godwit wingspans. When following the smaller dowitchers, godwits reduced the modal lateral distance to 0.60 m or 0.92 wingspans when calculated using the average wingspan of dowitchers and godwits (Mann-Whitney U = 525,684; n₁ = 1034; n₂ = 81; p=0.0004). Dowitchers following conspecifics flew with a modal lateral distance of 0.51 m, or 0.98 dowitcher wingspans. When following the larger godwits, dowitchers increased the modal lateral distance to 0.58 m or 0.89 average wingspans (Mann-Whitney U test; U = 66,341; n₁ = 743; n₂ = 149; p<0.0003).

Figure 4

Download asset Open asset

Figure 4—source data 1

Neighbor position data for mixed-species flocks.

https://doi.org/10.7554/eLife.45071.014

Download elife-45071-fig4-data1-v1.zip

Comparison of simple- and compound-V formations

While recording the larger cluster flocks, we also recorded four godwit simple-V formations of between 16 and 44 individuals, which were recorded for between 42 and 211 frames (Figure 5). Here we compare the positioning of godwits in simple and compound-V formations. In both cases, nearest neighbors were most commonly in the same horizontal plane (mean of 61% in godwit cluster flocks, 97.9% in godwit simple-V formations), defined as extending one wingspan above and below the focal bird, with the follower positioned over a narrow lateral range and wider range of trailing distances (Figures 3b and 5b). The modal lateral position in the simple-V formations was slightly less (mean of 0.8 wingspans) than that in the compound-V formations, where the mean modal lateral position among godwit flocks was 0.96 wingspans (Generalized Linear Model with terms for flock and simple versus compound-V formation; p<0.0001; Figure 5, Figure 5—source data 1). The modal trailing distance in simple-V formations was 0.50 wingspans; in compound-V formations of godwits, the mean modal trailing distance was 0.86 wingspans.

Figure 5

Download asset Open asset

Figure 5—source data 1

Godwit simple-V-formation position data.

https://doi.org/10.7554/eLife.45071.016

Download elife-45071-fig5-data1-v1.zip

Extended flock structure

We next examined how individual neighbor alignment rules relate to flock structure. We measured the angular distribution of neighbors at distances of two, four, six, and eight wingspans and at the maximum distance at which half of the flock remains in the flock’s core (range 5.8–24.3 wingspans). This last measure was used as a proxy for whole flock structure while avoiding edge effects (see 'Materials and methods'). For this analysis, we included all neighbors flying within a ±15 degree elevation slice relative to each focal bird. This was used instead of the ±1 wingspan slice used in other analyses (e.g., Figure 2, Figure 3) because this metric corresponds to a decreasing proportion of the volume at further distances. At a distance of two wingspans, flocks were consistently asymmetrical, with trailing birds more frequently flying to the left of their leading neighbors in 12 of 18 flocks and to the right of their nearest leading neighbors in the remaining six flocks. This asymmetry persisted at all distances within the flock (Figure 6a), including the overall flock shape (Figure 6b). The direction of asymmetry was independent of relative camera viewing direction and flock turning direction but was positively correlated with relative wind direction (see statistical results in Table 2).

Figure 6

Download asset Open asset

Figure 6—source data 1

Extended flock structure data.

https://doi.org/10.7554/eLife.45071.018

Download elife-45071-fig6-data1-v1.zip

Table 2

Variable	Test	N	R2	t/F	P
Wind direction	Circular correlation	18	0.29	2.29	0.02
Turn direction	Linear regression	18	0.00	0.04	0.83
Camera direction	Circular correlation	18	0.02	0.61	0.54

1. Tests of the relationship between flock left-right orientation (Figure 6) and environmental factors. The data usedto generate this table are available in Table 2—source data 1.

Table 2—source data 1

Flock orientation data.

https://doi.org/10.7554/eLife.45071.020

Download elife-45071-table2-data1-v1.zip

Flock biomechanics

We quantified several biomechanically relevant parameters from individual birds in flocks, including ground speed, estimated air speed, ascent or descent speed, wingbeat frequency and flapping phase. We created linear mixed effects (LME) statistical models to predict wingbeat frequency and airspeed from local and global flock positions and other flight parameters (Table 3). While speeds were measured for all individuals, flapping frequency and phase were only available from six cluster flocks in which birds were sufficiently close to the cameras to allow wingbeat measurements and for the simple-V formations. We examine only data for which wingbeat and estimated air speed data were available (N = 3306 individuals). We were also unable to measure flapping parameters from Dunlin, the smallest species recorded here.

Table 3

Wingbeat frequency predictors	Estimate	S.E.	T	d.f.	P
Intercept (dowitchers)	8.82	0.132	66.5	2817	<0.00001
Godwit	−2.19	0.035	−63.5	2817	<0.00001
Avocet	−1.91	0.040	−47.5	2817	<0.00001
Airspeed (m s−1)	−0.05	0.013	−4.3	2817	0.00002
Flock position	0.13	0.032	4.0	2817	0.00006
Nearest neighbor distance (wingspans)	−0.03	0.010	−3.2	2817	0.00153
Nearest neighbor species	−0.45	0.038	11.9	2817	<0.00001
Z-speed (m s−1)	0.29	0.019	15.1	2817	<0.00001
Airspeed predictors	Estimate	S.E.	T	d.f.	P
Intercept (dowitchers)	10.69	0.225	47.5	2832	<0.00001
Godwit	−0.25	0.067	−3.7	2832	0.00022
Avocet	−1.51	0.067	−22.6	2832	<0.00001
n.n. distance (wingspans)	−0.08	0.015	−5.2	2832	<0.00001
Edge distance (wingspans)	−0.03	0.003	−10.3	2832	<0.00001
Wingbeat frequency (H z)	−0.12	0.025	−5.0	2832	<0.00001
Flock position	−0.24	0.045	−5.4	2832	<0.00001
Aerodynamic neighbor	0.23	0.040	5.7	2832	<0.00001

1. Results from linear mixed-effects models relating wingbeat frequency and airspeed to other measured variables. Nearest neighbor only defined when this bird is leading the focal bird. Godwit and avocet are dummy variables coding species differences relative to dowitchers. Nearest neighbor species is coded −1 for a smaller neighbor, 0 for same the species, and 1 for a larger neighbor. Flock position is continuously scaled from 0.0 (front) to 1.0 (back). Aerodynamic neighbor was coded 1 for birds flying within 0.7–1.5 wingspans lateral distance and within two wingspans distance from their nearest leading neighbor, 0 otherwise. Models were selected using Bayesian information criteria. The data used to generate this table are available in Table 3—source data 1. d.f., degrees of freedom.

Table 3—source data 1

Flock biomechanical data.

https://doi.org/10.7554/eLife.45071.022

Download elife-45071-table3-data1-v1.zip

We observed several individual and flock effects on flight speed and wingbeat frequency (see Table 3 for full statistical results). As expected, different species flew with different characteristic flapping frequencies (LME, p<0.00001 for all species) and speeds (LME, p<0.00001 for dowitcher and avocet, p=0.00022 for godwit), and climbing flight was associated with an increase in flapping frequency (LME, p<0.00001). Birds flying near the front of the flock along the direction of travel (birds were given a continuous index with 0 being the frontmost and 1 the rearward-most position) flew faster (LME, p<0.00001) and with a lower flapping frequency than those near the rear (LME, p<0.00006). Birds flying near the edge of the flock also flew faster than those in the middle (LME, p<0.00001). Higher flapping frequencies were correlated with slower flight (LME, p<0.00001). Birds flying within a plausible range of locations for aerodynamic interaction (0.7–1.5 wingspans lateral distance and within two wingspans overall distance of their leading neighbor, coded as ‘Aerodynamic neighbor’ in Table 3) flew faster than expected after controling for the other effects described above (LME, p<0.00001, Figure 7). However, positioning in this aerodynamic interaction region had no effect on flapping frequency, as it was not in our best model of wingbeat frequency based on Bayesian information criteria (BIC; Table 3). Adding the aerodynamic neighbor to the best model makes the term non-significant (LME, p=0.30) and increases the BIC of the model by 6.8.

Figure 7

Download asset Open asset

Figure 7—source data 1

Neighbor position and flight speed data.

https://doi.org/10.7554/eLife.45071.024

Download elife-45071-fig7-data1-v1.zip

We examined the compound-V-flock data for evidence of flapping synchronization by examining the temporal and spatial phase offset between pairs of nearest neighbors for which synchronous wingbeat frequency data were available for at least 20 frames (see 'Materials and methods'). We found no evidence for temporal (Rayleigh test; N = 117; Z = 1.98; p=0.14) or spatial wingbeat synchronization (Rayleigh test; N = 117; Z = 1.28; p=0.27) in the compound-V-formation shorebird flocks. We performed the same tests on the simple-V formation of godwits and again found no support for phasing relationships (Rayleigh test; temporal phasing N = 39; Z = 0.09; p=0.90; spatial phasing; N = 39; Z = 0.46, p=0.63).

Here, we report on the first cross-species quantitative analysis of bird flocking behavior. On the basis of previous studies, we predicted that larger species would adopt more structured flocks and would exhibit more frequent aerodynamic positioning. Neither of these hypotheses were supported by our data. Instead, we document a flock structure that we term the compound-V formation, in which birds in cluster flocks align to nearest neighbors within their same elevation slice (±1 wingspan) with a one-wingspan lateral offset while allowing front-back distance to fluctuate with flock density (Figures 2, 3 and 4). This flock type is similar to the shorebird ‘cluster’ and ‘bunch’ formations described by Piersma et al. (1990) and the ‘front cluster’ of Heppner (1974). Here, this structure was observed in single- and mixed-species flocks of four shorebird species covering a seven-fold range in body mass. The simple alignment rule produces a flock structure that can be observed at all spatial scales within the flock, including overall flock shape (Figure 6). This is in contrast to flocks of other species, such as starlings, in which structure is only observed within each neighbor’s six nearest neighbors, equivalent to 1.2–2.7 wingspans (Ballerini et al., 2008a). Our data also show that shorebird global flock alignment is responsive to estimated local wind conditions (Table 2), and future work exploring this interaction may allow identification of the mechanism that governs the overall alignment. Wind conditions did not have discernible effects on local alignment, possibly because of the uncertainty in the measurement of the wind vector itself.

Mixed-species assemblages typically represent around 10% of migratory shorebird flocks (Piersma et al., 1990), possibly because species that have different preferred flight speeds would have to compromise their flight speed in order to remain together as a group. Our data include ~10% mixed-species flocks (2 of 18) and support the hypothesis that differences in preferred flight speed influence whether different species flock together. We documented two mixed-species flocks of godwit and dowitcher; these flocks had an airspeed of 9.02 ± 0.34 m s⁻¹ (mean ± s.d., n = 2). The single-species godwit and dowitcher flocks had airspeeds of 9.64 ± 1.53 (n = 3) and 8.99 ± 1.92 m s⁻¹ (n = 3), respectively. Dunlin, which were present at the same time as godwits and dowitchers but did not fly in mixed flocks with these species, had an airspeed of 7.58 ± 0.11 m s⁻¹ (n = 7). Similarly, Avocets not observed to mix with other species at our field site, and they flew with airspeeds of 8.04 ± 0.13 m s⁻¹ (n = 3). Thus, the similarity in preferred flight speeds among godwits and dowitchers might be important for these species to form mixed-species flocks. Dunlin and avocets also flew at similar airspeeds, but were not observed in mixed-species flocks, perhaps because of their large difference in body size, wingbeat frequency, and/or maneuverability.

The flock data presented here also include other interesting results that lack clear explanations. Flight speed varied with position from front to rear and from center to margin (Table 2), implying that the flocks were not necessarily in equilibrium. This might cause larger flocks to separate into several smaller flocks over time, consistent with the observation that the arrival group size of migratory species is typically smaller than the departure group size (Piersma et al., 1990).

Because birds in simple and compound-V formations adopt similar neighbor alignment rules, functional hypotheses for simple-V formations might also apply to compound-V formations. These include collision avoidance and information transfer (Dill et al., 1997). Collision avoidance is a plausible hypothesis to explain the formation of simple-V formations because they theoretically permit birds to keep all neighbors out of their direct path of travel. This is not the case for compound-V formations, where many birds are flying in front of and behind one another (Figure 1b; Figure 6b). The problem of collision avoidance is exacerbated in compound-V formation because birds tend to fly in the same horizontal plane. A better strategy for collision avoidance is to fly in a three-dimensional shape, such as that adopted by flocks of chimney swifts (Evangelista et al., 2017). In these flocks, the most common neighbor position is further lateral than in the shorebird flocks and with a shorter trailing distance, more completely moving those individuals out of the path of other flock members (Figure 3—figure supplement 1). Finally, even in the simple-V formation recorded here (Figure 5), birds flew with approximately 20% of wingspan overlap and so did not have an entirely clear forward path. Thus, collision avoidance appears to be an unlikely explanation for the structuring of both the compound-V and simple-V formations recorded here.

Simple and compound-V formations might also be structured to maximize the observability of neighbors, facilitating information transfer by helping birds to detect and respond to changes in neighbor speed or direction, and improving flock cohesiveness by allowing information to propagate through the flock more quickly. Dill et al. (1997) proposed that birds in V formation should maximize the measurement of neighbor movements by aligning at a 35.3 degree angle (relative to the direction of travel), or alternatively should maximize the measurement of neighbor speed by aligning at a 63.4 degree angle. The shorebird flocks examined here had modal neighbor-position alignment angles ranging from 33.7 to 51.8 with an average of 41.2 degrees. Neither this mean angle nor the nearly 20-degree range in alignment angle is consistent with Dill’s hypotheses or others calling for a single optimal alignment angle. Our finding that lateral spacing is uncorrelated with flock density, whereas trailing spacing increases with decreasing density, shows that the shorebird flocks are more organized in lateral distance than in trailing distance or alignment angle. Thus, hypotheses calling for organization based on alignment angle, whether to maximize information transfer or to keep lead birds in the visual fovea of trailing neighbors in a V formation (Badgerow and Hainsworth, 1981), are not well supported by our results.

Theoretical (Badgerow and Hainsworth, 1981; Hummel, 1983; Lissaman and Shollenberger, 1970; Maeng et al., 2013) and empirical research (Portugal et al., 2014; Weimerskirch et al., 2001) has provided support for the hypothesis that birds flying in simple-V formations gain aerodynamic and energetic benefits, and we propose that such benefits might also explain why birds adopt the compound-V formation. In both cases, birds fly with a lateral offset of approximately one wingspan while allowing trailing distance to vary (Figures 3 and 5), facilitating aerodynamic interaction. When compared to simple-V formations, compound-V formations allow greater flock densities, which should allow more rapid information transfer (Attanasi et al., 2014), larger flock sizes, and improved predator defense (Powell, 1974). Analyses of the airspeeds and wingbeat frequencies of flocking shorebirds provide some support for the aerodynamic alignment hypothesis. Birds flying in positions where beneficial aerodynamic interactions are predicted to occur flew faster than expected after controling for other factors (aerodynamic neighbor term in Table 3, linear mixed effects model for airspeed, p<0.0001). Over the entire dataset, 29.7% of nearest neighbor positions were in the ‘aerodynamic neighbor’ location (Figure 3—figure supplement 2), compared with only 3.4% of nearest neighbor in flocking chimney swifts (Figure 3—figure supplement 1). This faster flight should produce a reduced cost of transport, assuming there are no unmeasured compensating factors such as a simultaneous increase in stroke amplitude. Nevertheless, this speculative interpretation of the compound-V formation raises many new questions, such as how birds in the flock can maintain different speeds without separating and why an aerodynamic benefit would manifest as an increase in speed instead of, for example, a reduction in flapping frequency and airspeed as suggested by theoretical models (Hummel, 1983). Furthermore, despite the similarities in modal position among compound-V and simple-V flocks (Figures 3 and 5), it is not clear whether a single set of adjustment rules or responses to changes in neighbor position can produce both flock types. These questions, and a definitive explanation for why birds adopt a compound-V formation, cannot be answered with the current dataset. Progress in these areas will depend on new theoretical modeling and data collection from on-bird loggers measuring physiological, flock positioning and biomechanical data from a variety of species over a range of behavioral contexts. Further videographic flock surveys may also improve understanding of the variety of flock types, especially when collected with careful attention to behavioral context and with full measurement of local environmental conditions.

Datasets used in the analysis are included in the manuscript and supporting files. Source data files have been provided for all figures and tables.

1. 1. Attanasi A
   2. Cavagna A
   3. Del Castello L
   4. Giardina I
   5. Grigera TS
   6. Jelić A
   7. Melillo S
   8. Parisi L
   9. Pohl O
   10. Shen E
   11. Viale M

   (2014) Information transfer and behavioural inertia in Starling flocks

   Nature Physics 10:691–696.

   https://doi.org/10.1038/nphys3035

   • Google Scholar
2. 1. Badgerow JP
   2. Hainsworth FR

   (1981) Energy savings through formation flight? A re-examination of the vee formation

   Journal of Theoretical Biology 93:41–52.

   https://doi.org/10.1016/0022-5193(81)90055-2

   • Google Scholar
3. 1. Baker AJ
   2. Pereira SL
   3. Paton TA

   (2007) Phylogenetic relationships and divergence times of charadriiformes genera: multigene evidence for the cretaceous origin of at least 14 clades of shorebirds

   Biology Letters 3:205–210.

   https://doi.org/10.1098/rsbl.2006.0606

   • PubMed
   • Google Scholar
4. 1. Ballerini M
   2. Cabibbo N
   3. Candelier R
   4. Cavagna A
   5. Cisbani E
   6. Giardina I
   7. Orlandi A
   8. Parisi G
   9. Procaccini A
   10. Viale M
   11. Zdravkovic V

   (2008a) Empirical investigation of Starling flocks: a benchmark study in collective animal behaviour

   Animal Behaviour 76:201–215.

   https://doi.org/10.1016/j.anbehav.2008.02.004

   • Google Scholar
5. 1. Ballerini M
   2. Cabibbo N
   3. Candelier R
   4. Cavagna A
   5. Cisbani E
   6. Giardina I
   7. Lecomte V
   8. Orlandi A
   9. Parisi G
   10. Procaccini A
   11. Viale M
   12. Zdravkovic V

   (2008b) Interaction ruling animal collective behavior depends on topological rather than metric distance: evidence from a field study

   PNAS 105:1232–1237.

   https://doi.org/10.1073/pnas.0711437105

   • PubMed
   • Google Scholar
6. 1. Buhl J
   2. Sumpter DJ
   3. Couzin ID
   4. Hale JJ
   5. Despland E
   6. Miller ER
   7. Simpson SJ

   (2006) From disorder to order in marching locusts

   Science 312:1402–1406.

   https://doi.org/10.1126/science.1125142

   • PubMed
   • Google Scholar
7. 1. Cavagna A
   2. Cimarelli A
   3. Giardina I
   4. Parisi G
   5. Santagati R
   6. Stefanini F
   7. Viale M

   (2010) Scale-free correlations in Starling flocks

   PNAS 107:11865–11870.

   https://doi.org/10.1073/pnas.1005766107

   • PubMed
   • Google Scholar
8. 1. Cutts C
   2. Speakman J

   (1994)

   Energy savings in formation flight of pink-footed geese

   The Journal of Experimental Biology 189:251–261.

   • PubMed
   • Google Scholar
9. Book

   1. Dill LM
   2. Holling CS
   3. Palmer LH

   (1997)

   Predicting the three-dimensional structure of animal aggregations from functional considerations: the role of information

   In: Parrish J. K, Hamner W. M, editors. Animal Groups in Three Dimensions: How Species Aggregate. Cambridge, UK: Cambridge University Press. pp. 207–224.

   • Google Scholar
10. 1. Evangelista DJ
    2. Ray DD
    3. Raja SK
    4. Hedrick TL

    (2017) Three-dimensional trajectories and network analyses of group behaviour within chimney swift flocks during approaches to the roost

    Proceedings of the Royal Society B: Biological Sciences 284:20162602.

    https://doi.org/10.1098/rspb.2016.2602

    • Google Scholar
11. Book

    1. Fisher N

    (1995)

    Statistical Analysis of Circular Data

    Cambridge University Press.

    • Google Scholar
12. 1. Hedrick TL

    (2008) Software techniques for two- and three-dimensional kinematic measurements of biological and biomimetic systems

    Bioinspiration & Biomimetics 3:034001.

    https://doi.org/10.1088/1748-3182/3/3/034001

    • PubMed
    • Google Scholar
13. 1. Hedrick TL
    2. Pichot C
    3. de Margerie E

    (2018) Gliding for a free lunch: biomechanics of foraging flight in common swifts (apus apus)

    The Journal of Experimental Biology 221:jeb186270.

    https://doi.org/10.1242/jeb.186270

    • PubMed
    • Google Scholar
14. 1. Hemelrijk CK
    2. Hildenbrandt H

    (2012) Schools of fish and flocks of birds: their shape and internal structure by self-organization

    Interface Focus 2:726–737.

    https://doi.org/10.1098/rsfs.2012.0025

    • PubMed
    • Google Scholar
15. 1. Heppner FH

    (1974) Avian Flight Formations

    Bird-Banding 45:160.

    https://doi.org/10.2307/4512025

    • Google Scholar
16. 1. Herbert-Read JE
    2. Perna A
    3. Mann RP
    4. Schaerf TM
    5. Sumpter DJ
    6. Ward AJ

    (2011) Inferring the rules of interaction of shoaling fish

    PNAS 108:18726–18731.

    https://doi.org/10.1073/pnas.1109355108

    • PubMed
    • Google Scholar
17. 1. Herbert-Read JE

    (2016) Understanding how animal groups achieve coordinated movement

    The Journal of Experimental Biology 219:2971–2983.

    https://doi.org/10.1242/jeb.129411

    • PubMed
    • Google Scholar
18. 1. Hummel D

    (1983) Aerodynamic aspects of formation flight in birds

    Journal of Theoretical Biology 104:321–347.

    https://doi.org/10.1016/0022-5193(83)90110-8

    • Google Scholar
19. 1. Jackson BE
    2. Evangelista DJ
    3. Ray DD
    4. Hedrick TL

    (2016) 3D for the people: multi-camera motion capture in the field with consumer-grade cameras and open source software

    Biology Open 5:1334–1342.

    https://doi.org/10.1242/bio.018713

    • PubMed
    • Google Scholar
20. 1. Katz Y
    2. Tunstrøm K
    3. Ioannou CC
    4. Huepe C
    5. Couzin ID

    (2011) Inferring the structure and dynamics of interactions in schooling fish

    PNAS 108:18720–18725.

    https://doi.org/10.1073/pnas.1107583108

    • PubMed
    • Google Scholar
21. 1. Ling H
    2. Mclvor GE
    3. Nagy G
    4. MohaimenianPour S
    5. Vaughan RT
    6. Thornton A
    7. Ouellette NT

    (2018) Simultaneous measurements of three-dimensional trajectories and wingbeat frequencies of birds in the field

    Journal of the Royal Society Interface 15:20180653.

    https://doi.org/10.1098/rsif.2018.0653

    • Google Scholar
22. 1. Lissaman PB
    2. Shollenberger CA

    (1970) Formation flight of birds

    Science 168:1003–1005.

    https://doi.org/10.1126/science.168.3934.1003

    • PubMed
    • Google Scholar
23. 1. Lukeman R
    2. Li YX
    3. Edelstein-Keshet L

    (2010) Inferring individual rules from collective behavior

    PNAS 107:12576–12580.

    https://doi.org/10.1073/pnas.1001763107

    • PubMed
    • Google Scholar
24. 1. Maeng JS
    2. Park JH
    3. Jang SM
    4. Han SY

    (2013) A modeling approach to energy savings of flying Canada geese using computational fluid dynamics

    Journal of Theoretical Biology 320:76–85.

    https://doi.org/10.1016/j.jtbi.2012.11.032

    • PubMed
    • Google Scholar
25. 1. Nagy M
    2. Akos Z
    3. Biro D
    4. Vicsek T

    (2010) Hierarchical group dynamics in pigeon flocks

    Nature 464:890–893.

    https://doi.org/10.1038/nature08891

    • PubMed
    • Google Scholar
26. 1. Nagy M
    2. Vásárhelyi G
    3. Pettit B
    4. Roberts-Mariani I
    5. Vicsek T
    6. Biro D

    (2013) Context-dependent hierarchies in pigeons

    PNAS 110:13049–13054.

    https://doi.org/10.1073/pnas.1305552110

    • PubMed
    • Google Scholar
27. 1. Pettit B
    2. Perna A
    3. Biro D
    4. Sumpter DJT

    (2013) Interaction rules underlying group decisions in homing pigeons

    Journal of the Royal Society Interface 10:20130529.

    https://doi.org/10.1098/rsif.2013.0529

    • Google Scholar
28. 1. Pettit B
    2. Ákos Z
    3. Vicsek T
    4. Biro D

    (2015) Speed determines leadership and leadership determines learning during pigeon flocking

    Current Biology 25:3132–3137.

    https://doi.org/10.1016/j.cub.2015.10.044

    • PubMed
    • Google Scholar
29. 1. Piersma T
    2. Zwarts L
    3. Bruggemann JH

    (1990)

    Behavioural aspects of the departure of waders before long-distance flights: flocking, vocalizations, flight paths and diurnal timing

    Ardea 78:157–184.

    • Google Scholar
30. 1. Portugal SJ
    2. Hubel TY
    3. Fritz J
    4. Heese S
    5. Trobe D
    6. Voelkl B
    7. Hailes S
    8. Wilson AM
    9. Usherwood JR

    (2014) Upwash exploitation and downwash avoidance by flap phasing in ibis formation flight

    Nature 505:399–402.

    https://doi.org/10.1038/nature12939

    • PubMed
    • Google Scholar
31. 1. Powell GVN

    (1974) Experimental analysis of the social value of flocking by starlings (Sturnus vulgaris) in relation to predation and foraging

    Animal Behaviour 22:501–505.

    https://doi.org/10.1016/S0003-3472(74)80049-7

    • Google Scholar
32. 1. Shamoun-Baranes J
    2. van Loon E
    3. Liechti F
    4. Bouten W

    (2007) Analyzing the effect of wind on flight: pitfalls and solutions

    Journal of Experimental Biology 210:82–90.

    https://doi.org/10.1242/jeb.02612

    • PubMed
    • Google Scholar
33. 1. Theriault DH
    2. Fuller NW
    3. Jackson BE
    4. Bluhm E
    5. Evangelista D
    6. Wu Z
    7. Betke M
    8. Hedrick TL

    (2014) A protocol and calibration method for accurate multi-camera field videography

    Journal of Experimental Biology 217:1843–1848.

    https://doi.org/10.1242/jeb.100529

    • PubMed
    • Google Scholar
34. 1. Usherwood JR
    2. Stavrou M
    3. Lowe JC
    4. Roskilly K
    5. Wilson AM

    (2011) Flying in a flock comes at a cost in pigeons

    Nature 474:494–497.

    https://doi.org/10.1038/nature10164

    • PubMed
    • Google Scholar
35. 1. Vicsek T
    2. Zafeiris A

    (2012) Collective motion

    Physics Reports 517:71–140.

    https://doi.org/10.1016/j.physrep.2012.03.004

    • Google Scholar
36. 1. Weimerskirch H
    2. Martin J
    3. Clerquin Y
    4. Alexandre P
    5. Jiraskova S

    (2001) Energy saving in flight formation

    Nature 413:697–698.

    https://doi.org/10.1038/35099670

    • PubMed
    • Google Scholar
37. 1. Weinzierl R
    2. Bohrer G
    3. Kranstauber B
    4. Fiedler W
    5. Wikelski M
    6. Flack A

    (2016) Wind estimation based on thermal soaring of birds

    Ecology and Evolution 6:8706–8718.

    https://doi.org/10.1002/ece3.2585

    • PubMed
    • Google Scholar
38. Conference

    1. Wu Z
    2. Hristov N
    3. Hedrick TL
    4. Kunz TH
    5. Betke M

    (2009) Tracking a large number of objects from multiple views

    In 2009 IEEE 12th International Conference on Computer Vision IEEE. pp. 1546–1553.

    https://doi.org/10.1109/ICCV.2009.5459274

    • Google Scholar

Author details

1. Aaron J Corcoran

   University of North Carolina at Chapel Hill, Chapel Hill, United States

   Contribution

   Conceptualization, Software, Investigation, Visualization, Methodology, Writing—original draft, Writing—review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-1457-3689

2. Tyson L Hedrick

   University of North Carolina at Chapel Hill, Chapel Hill, United States

   Contribution

   Conceptualization, Resources, Data curation, Software, Supervision, Funding acquisition, Visualization, Methodology, Project administration, Writing—review and editing

   For correspondence

   thedrick@bio.unc.edu

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-6573-9602

• 2,124

  views

• 167

  downloads

• 14

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.