Microscopic origin of scale-free correlations in natural starling flocks

Field experiments that reconstructed the instantaneous three-dimensional positions and velocities of individual starlings (122 to 4,268 birds per flock) showed that spatial correlations of both the direction and the speed of velocity fluctuations are scale free: the correlation length grows linearly with flock size, and the correlation function decays as a power law with an anomalously small exponent, so that a behavioral change in one bird can in principle influence every other bird without damping. The originating study (Cavagna et al., PNAS 2010) left the microscopic origin of this state open — whether it arises from very low coordination noise, from proximity to a critical point, or from some other mechanism — and described how starlings achieve such strong correlations as a mystery. Later work resolved parts of the question: maximum-entropy inference indicates that flock parameters are tuned near a critical point; the propagation of turning information was measured to be linear and essentially unattenuated, requiring behavioral inertia (inertial spin dynamics); and a marginal speed-confinement mechanism was proposed and validated for the scale-free correlations of speed. What remains missing is a validated microscopic mechanism — the actual coordination rules, noise structure, and dynamical laws, and the precise sense in which the collective state is critical or marginal — that quantitatively reproduces the observed correlation structure across flock sizes and is consistent with the measured propagation phenomenology. The problem is to determine and validate that mechanism.

Background

Bird flocks, fish schools, and insect swarms are canonical examples of collective motion: coherent, ordered group behavior emerging from local interactions among many individuals with no leader. The standard theoretical laboratory for this phenomenon is the class of self-propelled particle models introduced by Vicsek and coworkers: each agent moves at roughly constant speed and tends to align its direction of motion with its near neighbors, subject to random noise; as the noise is lowered, the group undergoes an order–disorder transition into a globally ordered state. The degree of order is measured by the polarization, the magnitude of the group's mean velocity normalized by the individual cruising speed. In these models, as in equilibrium magnets, fluctuations about the ordered state are short-ranged except at criticality: at a critical point the correlation length — the typical distance over which fluctuations stay correlated — diverges and correlations decay as power laws rather than exponentials. Scale-free correlations, in which the correlation length is limited only by the system size, are therefore the signature either of proximity to a critical point or of Goldstone modes — the soft transverse fluctuations that any spontaneously broken continuous symmetry generically produces even far from criticality.

The question studied here originates in a field experiment by Cavagna and colleagues (2010), who used stereometric digital photogrammetry to reconstruct the instantaneous 3D positions and velocities of individual starlings (Sturnus vulgaris) in 24 flocking events of 122 to 4,268 birds over a winter roost in Rome. Defining each bird's velocity fluctuation as its velocity minus the flock's center-of-mass velocity, they measured the spatial correlation function $C(r)$, the normalized average scalar product of the fluctuations of pairs of birds at mutual distance $r$. The correlation length $\xi$, defined as the distance where $C(r)$ first vanishes, was found to scale linearly with the flock's linear size $L$ ($\xi \approx 0.35 L$, with no saturation across flocks spanning an order of magnitude in size): the correlations are scale free. Equivalently, $C(r)$ is compatible with a power-law decay $C(r)\sim r^{-\gamma}$ with an anomalously small exponent ($\gamma<1$), meaning the correlation is almost non-decaying across the whole group. Crucially, both the direction and the modulus (the speed) of the velocity fluctuations are scale free. Symmetry arguments can explain scale-free orientation fluctuations in an ordered phase as Goldstone modes, but they cannot explain scale-free speed correlations, which the authors took as the strongest indication that flocks are, in some sense, critical systems. Functionally, scale-free correlations mean that a behavioral perturbation — a sudden turn, an escape maneuver — can reach every bird without damping, a plausible physical basis for the rapid collective responses starling flocks display under predation. It was already known that starlings interact with a fixed number of neighbors (about six or seven) irrespective of their metric distance — topological rather than metric interactions (Ballerini et al., 2008) — so the interaction is short-ranged while the correlation is long-ranged.

Two theoretical tools built for this system are needed to read the literature. First, equilibrium maximum-entropy modeling: the least-structured probability distribution over flight directions consistent with the measured local alignment correlations is mathematically equivalent to a (topological) Heisenberg ferromagnet, with flight directions playing the role of spins; fitting it to data yields effective interaction strengths and a noise level, and one can ask whether the fitted parameters place the flock near the model's critical point. Second, behavioral inertia: experiments on spontaneous collective turns showed that a change of direction propagates through a flock as a linear, essentially undamped wave — impossible for the overdamped, first-order relaxation dynamics of standard alignment models, in which disturbances diffuse and attenuate — and motivated inertial spin dynamics, in which each bird carries an internal angular momentum (a spin) that turns its velocity, making the response wave-like. The dispersion relation $\omega(k)$, relating frequency to wavelength of these traveling disturbances, is the basic experimental characterization of the propagation.

Problem Statement

Identify the microscopic mechanism by which natural flocks of starlings produce scale-free, nearly non-decaying correlations of both the direction and the speed of individual flight velocities — the question the source paper left open when it asked 'Under what conditions do scale-free correlations appear? And what do scale-free correlations teach us about the interindividual coordination mechanism?' and stated that how starlings achieve such a strong correlation 'remains a mystery to us'.

The mechanism to be determined and validated against empirical 3D flock-tracking data comprises:

(i) the interindividual coordination rules: the interaction network (topological or metric), the alignment and speed-control terms, the noise structure, and the inertial terms of the birds' dynamics;

(ii) the dynamical regime: in what precise sense, if any, the flocks' collective state is critical or marginal — as opposed to a merely very-low-noise ordered phase — including which parameter(s) must be tuned, to what values, and by what mechanism the tuning is achieved or maintained;

(iii) the quantitative consequence: that the proposed mechanism reproduces the measured orientation and speed correlation functions, their scale-free scaling of correlation length with flock size, and the anomalously small decay exponent, and is consistent with the measured linear, weakly attenuated propagation of turning information and its dependence on the flock's polarization.

These components jointly answer one underlying question — what the birds actually do, dynamically, that makes every individual's behavioral state correlated with every other's, no matter how large the flock. The propagation-timescale sub-question that the source explicitly deferred has been answered by later experiments (linear dispersion, negligible attenuation); the mechanism question — why the rules and dynamics generate the observed correlation structure — remains the target.

References

1. Andrea Cavagna, Alessio Cimarelli, Irene Giardina, Giorgio Parisi, Raffaele Santagati, Fabio Stefanini, Massimiliano Viale (2010), 'Scale-free correlations in starling flocks', Proc. Natl. Acad. Sci. USA 107(26):11865–11870, doi:10.1073/pnas.1005766107, https://doi.org/10.1073/pnas.1005766107 (open-access copy PMC2900681)
2. Tamás Vicsek, András Czirók, Eshel Ben-Jacob, Inon Cohen, Ofer Shochet (1995), 'Novel type of phase transition in a system of self-driven particles', Phys. Rev. Lett. 75:1226–1229, doi:10.1103/PhysRevLett.75.1226, https://doi.org/10.1103/PhysRevLett.75.1226
3. M. Ballerini, N. Cabibbo, R. Candelier, A. Cavagna, E. Cisbani, I. Giardina, V. Lecomte, A. Orlandi, G. Parisi, A. Procaccini, M. Viale, V. Zdravkovic (2008), 'Interaction ruling animal collective behavior depends on topological rather than metric distance: Evidence from a field study', Proc. Natl. Acad. Sci. USA 105:1232–1237, doi:10.1073/pnas.0711437105, https://doi.org/10.1073/pnas.0711437105
4. Andrea Procaccini, Alberto Orlandi, Andrea Cavagna, Irene Giardina, Francesca Zoratto, Daniela Santucci, Flavia Chiarotti, Charlotte K. Hemelrijk, Enrico Alleva, Giorgio Parisi, Claudio Carere (2011), 'Propagating waves in starling, Sturnus vulgaris, flocks under predation', Animal Behaviour 82:759–765, doi:10.1016/j.anbehav.2011.07.006, https://doi.org/10.1016/j.anbehav.2011.07.006 (phenomenon characterized via Cavagna et al. 2018, Phys. Rev. E 98, 052404)
5. William Bialek, Andrea Cavagna, Irene Giardina, Thierry Mora, Edmondo Silvestri, Massimiliano Viale, Aleksandra M. Walczak (2012), 'Statistical mechanics for natural flocks of birds', Proc. Natl. Acad. Sci. USA 109:4786–4791, doi:10.1073/pnas.1118633109, https://doi.org/10.1073/pnas.1118633109 (open-access copy PMC3324025)
6. Alessandro Attanasi, Andrea Cavagna, Lorenzo Del Castello, Irene Giardina, Tomás S. Grigera, Asja Jelić, Stefania Melillo, Leonardo Parisi, Oliver Pohl, Edward Shen, Massimiliano Viale (2014), 'Information transfer and behavioural inertia in starling flocks', Nature Physics 10:691–696, doi:10.1038/nphys3035, https://doi.org/10.1038/nphys3035
7. William Bialek, Andrea Cavagna, Irene Giardina, Thierry Mora, Oliver Pohl, Edmondo Silvestri, Massimiliano Viale, Aleksandra M. Walczak (2014), 'Social interactions dominate speed control in poising natural flocks near criticality', Proc. Natl. Acad. Sci. USA 111:7212–7217, doi:10.1073/pnas.1324045111, https://doi.org/10.1073/pnas.1324045111
8. George F. Young, Luca Scardovi, Andrea Cavagna, Irene Giardina, Naomi E. Leonard (2013), 'Starling Flock Networks Manage Uncertainty in Consensus at Low Cost', PLoS Computational Biology 9:e1002894, doi:10.1371/journal.pcbi.1002894, https://doi.org/10.1371/journal.pcbi.1002894
9. Andrea Cavagna, Lorenzo Del Castello, Supravat Dey, Irene Giardina, Stefania Melillo, Leonardo Parisi, Massimiliano Viale (2015), 'Short-range interactions versus long-range correlations in bird flocks', Phys. Rev. E 92:012705, doi:10.1103/PhysRevE.92.012705, https://doi.org/10.1103/PhysRevE.92.012705
10. Andrea Cavagna, Irene Giardina, Tomás S. Grigera, Asja Jelić, Dov Levine, Sriram Ramaswamy, Massimiliano Viale (2015), 'Silent Flocks: Constraints on Signal Propagation Across Biological Groups', Phys. Rev. Lett. 114:218101, doi:10.1103/PhysRevLett.114.218101, https://doi.org/10.1103/PhysRevLett.114.218101
11. Thierry Mora, Aleksandra M. Walczak, Lorenzo Del Castello, Francesco Ginelli, Stefania Melillo, Leonardo Parisi, Massimiliano Viale, Andrea Cavagna, Irene Giardina (2016), 'Local equilibrium in bird flocks', Nature Physics 12:1153–1157, doi:10.1038/nphys3846, https://doi.org/10.1038/nphys3846
12. Charlotte K. Hemelrijk, Hanno Hildenbrandt (2014), 'Scale-Free Correlations, Influential Neighbours and Speed Control in Flocks of Birds', Journal of Statistical Physics 158:563–578, doi:10.1007/s10955-014-1154-0, https://doi.org/10.1007/s10955-014-1154-0
13. Alessandro Attanasi, Andrea Cavagna, Lorenzo Del Castello, Irene Giardina, Stefania Melillo, Leonardo Parisi, Oliver Pohl, Bruno Rossaro, Edward Shen, Edmondo Silvestri, Massimiliano Viale (2014), 'Finite-Size Scaling as a Way to Probe Near-Criticality in Natural Swarms', Phys. Rev. Lett. 113:238102, doi:10.1103/PhysRevLett.113.238102, https://doi.org/10.1103/PhysRevLett.113.238102
14. Andrea Cavagna, Daniele Conti, Chiara Creato, Lorenzo Del Castello, Irene Giardina, Tomás S. Grigera, Stefania Melillo, Leonardo Parisi, Massimiliano Viale (2017), 'Dynamic scaling in natural swarms', Nature Physics 13:914–918, doi:10.1038/nphys4153, https://doi.org/10.1038/nphys4153
15. Andrea Cavagna, Daniele Conti, Irene Giardina, Tomás S. Grigera (2018), 'Propagating speed waves in flocks: A mathematical model', Phys. Rev. E 98:052404, doi:10.1103/PhysRevE.98.052404, https://doi.org/10.1103/PhysRevE.98.052404
16. Andrea Cavagna, Antonio Culla, Luca Di Carlo, Irene Giardina, Tomás S. Grigera (2019), 'Low-temperature marginal ferromagnetism explains anomalous scale-free correlations in natural flocks', Comptes Rendus Physique 20:319–328, doi:10.1016/j.crhy.2019.05.008, https://doi.org/10.1016/j.crhy.2019.05.008 (preprint arXiv:1812.07522)
17. Charlotte K. Hemelrijk, Alessandro Costanzo, Hanno Hildenbrandt, Claudio Carere (2019), 'Damping of waves of agitation in starling flocks', Behavioral Ecology and Sociobiology 73:125, doi:10.1007/s00265-019-2734-4, https://doi.org/10.1007/s00265-019-2734-4
18. Mathias Casiulis, Marco Tarzia, Leticia F. Cugliandolo, Olivier Dauchot (2020), 'Velocity and Speed Correlations in Hamiltonian Flocks', Phys. Rev. Lett. 124:198001, doi:10.1103/PhysRevLett.124.198001, https://doi.org/10.1103/PhysRevLett.124.198001 (preprint arXiv:1911.06042)
19. Andrea Cavagna, Antonio Culla, Xiao Feng, Irene Giardina, Tomás S. Grigera, Willow Kion-Crosby, Stefania Melillo, Giulia Pisegna, Lorena Postiglione, Pablo Villegas (2022), 'Marginal speed confinement resolves the conflict between correlation and control in collective behaviour', Nature Communications 13:2315, doi:10.1038/s41467-022-29883-4, https://doi.org/10.1038/s41467-022-29883-4
20. Andrea Cavagna, Luca Di Carlo, Irene Giardina, Tomás S. Grigera, Stefania Melillo, Leonardo Parisi, Giulia Pisegna, Mattia Scandolo (2023), 'Natural swarms in 3.99 dimensions', Nature Physics 19:1043–1049, doi:10.1038/s41567-023-02028-0, https://doi.org/10.1038/s41567-023-02028-0
21. Andrea Cavagna, Guido Cimino, Javier Cristín, Matteo Fiorini, Irene Giardina, Angelo Giustiniani, Tomás S. Grigera, Stefania Melillo, Roberto A. Palombella, Leonardo Parisi, Antonio Ponno, Mattia Scandolo, Zachary S. Stamler (2026), 'Spin Waves without Spin Waves: A Case for Soliton Propagation in Starling Flocks', Phys. Rev. Lett. 137:098301, doi:10.1103/sppm-qsnb, https://doi.org/10.1103/sppm-qsnb (preprint arXiv:2505.19665, https://arxiv.org/abs/2505.19665)
22. Marina Papadopoulou, Hanno Hildenbrandt, Rolf F. Storms, Claudio Carere, Simon Verhulst, Charlotte K. Hemelrijk (2026), 'A mechanistic understanding of collective escape in starling flocks', Communications Biology 9:1055, doi:10.1038/s42003-026-10173-4, https://doi.org/10.1038/s42003-026-10173-4
23. Andrea Cavagna, Irene Giardina (2014), 'Bird Flocks as Condensed Matter', Annual Review of Condensed Matter Physics 5:183–207, doi:10.1146/annurev-conmatphys-031113-133834, https://doi.org/10.1146/annurev-conmatphys-031113-133834