## Abstract

It has been repeatedly reported that the collective dynamics of social insects exhibit universal emergent properties similar to other complex systems. In this note, we study a previously published data set in which the positions of thousands of honeybees in a hive are individually tracked over multiple days. The results show that the hive dynamics exhibit long-range spatial and temporal correlations in the occupancy density fluctuations, despite the characteristic short-range bees’ mutual interactions. The variations in the occupancy unveil a non-monotonic function between density and bees’ flow, reminiscent of the car traffic dynamic near a jamming transition at which the system performance is optimized to achieve the highest possible throughput. Overall, these results suggest that the beehive collective dynamics are self-adjusted towards a point near its optimal density.

### Similar content being viewed by others

## Introduction

The striking collective phenomena exhibited by social insects and animals have inspired complex systems scientists for decades. The study of the basic principles of their communication and self-organization into large ensembles can be traced back to the seminal work of Wilson on social insects^{1,2,3}. Large collective behavioral structures, such as hives, swarms, bird flocks, foraging ants trails, etc., emerge out of local interactions. The sizes of the resulting complex global structures are several orders of magnitude larger than the scale that the individuals are able to communicate. Prototypical examples are the large trails formed by foragers ants^{4}, the collective defensive reactions of entire bird flocks^{5,6}, the transition from disordered motion to ordered marching in locusts^{7} or the shimmering phenomena^{8} seen in giant honeybees as waves rippling across their open nests. These emergent behaviors have an adaptive value, facilitating the species’ survival. Often, the dynamics of these apparently disparate large-scale phenomena exhibit scale-invariant properties both in space and time which are common across the species, an observation that could be studied from the perspective of statistical physics^{9,10}.

It is well established that scale-free spatial and temporal correlation characterize systems near the second-order phase transition^{11}. In the paradigmatic case of the Ising model, when the control parameter, temperature, approaches its critical value long-range correlations emerge between distance lattice sites. The theory of complex systems implies that the information processing capacities of the system, as well as its dynamic range and capacity to generate emergent behavior patterns, are amplified at criticality. Therefore it stands to reason that living systems could be evolutionarily poised to exist in the vicinity of critical point.

The best-known example of a biological system with scale-free correlations and space and time is the vertebrate brain^{12,13}, however, evidence of scale-free correlations^{14} in animal collectives imply that they might be in a similar state. It should be noted, that the existence of statistical similarities between vertebrate brains and collective entities has wide-ranging philosophical and scientific implications, which largely lie outside the scope of this paper, but are actively explored in the literature^{15,16}.

The application of statistical physics toolbox to biological data has boomed once the data became available. Signatures of critical behavior have been uncovered in flocks of murmurating starlings^{9}, swarms of midges^{17}, mammal collectives^{18}, and fish shoals. It should be noted that honeybees, while being an archetypal example of complex group behavior^{19} have not been studied from this perspective, a deficiency we hope to address in our work. Modeling work has demonstrated, that fish shoals operate best in the parameter region close to criticality^{20}. At the same time, proximity to the critical region does come with associated costs—small-scale perturbation could be unnecessarily amplified to the detrimental results to the system. Moreover, as biological systems, even those of the same species vary in size, a single combination of preset parameters cannot always result in critical behavior. Therefore, ongoing research investigates mechanisms through which critical state could be controlled and maintained, with current theory leaning towards the view that animal collectives can tune their distance to the critical point depending on the needs of the moment^{14}.

Moreover, the aforementioned ample observations often lack the elucidation of generative mechanisms of scale-free correlations. At the same time, turning to simple models has been fruitful in explaining complex patterns of collective behavior^{21}. Invoking Occam’s razor, one should seek a sufficiently simple explanation. Exploring statistical similarities between the novel system at hand and well-studied systems for which the generative mechanism for the phase transition and its crucial properties have been established is a route that avoids unnecessary complexity. As bees in natural conditions are densely packed, the suitable analogy is the jamming transition^{22}. Such transition occurs in a system of moving agents when a certain density is reached—at this state, a non-trivial relation between the throughput (traffic) of agents and their density emerges—at low-density traffic grows linearly as a function of density, however once critical threshold is reached further increase in density only leads to a decreased throughput. At this density, critical phenomena, such as long-range correlations in time and space appear, stemming from the complex non-linear interactions of agents. Initially developed for car traffic this transition has been studied both in simulation and in the real traffic condition^{23,24}. Concepts of jamming have been applied to study motile biological agents such as cells and bacteria^{25,26}.

In this work we study the collective dynamics of a hive of thousands of bees, searching for clues of their large-scale dynamics, revisiting the data previously acquired by Gernat et al.^{27} graciously shared with the authors of the current work. The paper is organized as follows: In the next paragraphs, we describe briefly the main features of the data and the methods. Next, we study the univariate aspects of the fluctuations, including occupancy (density) fluctuations, nearest neighbors bees’ interaction distances, and insects’ speed. After that, we proceed to explore collective effects, by calculating spatial and temporal correlations for different coarse grainings of the data being considered. Finally, we investigate, in analogy to traffic dynamics, the relation between average mobility and local density. The paper concludes with a discussion on the possible origin of the scaling found in the hive collective dynamics.

## Results and discussion

### Data

The full details of the methods are described in Ref.^{27}. In brief, the experiment is conducted with a colony of 1200 1-day-old worker bees with a queen which are placed in a rectangular hive (dimensions \(348 \times 232\) mm) covered by a transparent material. The hive thickness is such that prevents the bees from climbing on top of each other while permitting full observation for recording. The dataset includes the position (the center of mass) of each bee sampled every second. It should be noted that bees could exhibit two types of behaviors that allow them to elude observation and also exclude them from interactions with the rest of the insects in the hive: they could go inside the honeycomb cells to sleep, eat, and take care of the brood or they could walk on the glass cover of the hive. For the first 2 days and two nights of the recording, insects were kept sealed and were supplied with the necessary nutrients (hereafter termed “phase I”). Afterward, bees were allowed to leave at will to scout and forage (“phase II”). During the last days of the trial, forages were not permitted to return to the hive (“phase III”). (The top 16 rows of the honeycomb cells were provisioned with 150 g of honey, and the two rows below the honey cells were provisioned with 15 g of artificial “bee bread”. Each bee was marked with barcodes attached to their thoraxes, enabling the continuous tracking of their positions. The sampling rate of the raw data is 1 s. Five different colonies were recorded, each for several days and nights. Results presented in the paper are computed using data from the second trial (dataset #2)).

### Density fluctuations

Figure 1A shows an example of the trajectories of two bees during the 7 h of the first day of the experiment, from which it can be seen that their activity covers the entire hive. This is best appreciated by computing the average density of bees discretizing the hive in \(30\times 20\) squares of side length = 11.6 mm and counting the number of bees in each grid cell at a sampling rate of 5 s. Figure 1B depicts, as a heat map, the average occupancy over the entire experiment. Typically, each bee, unless it is hiding inside one of the honeycomb cells, explores an average distance of \(\sim \)4–5 m every hour (see Panel D) displaying bursts of activity (as described before^{27}) seen in the two examples presented in Panel C.

It is well known that bees in the hive usually are highly packed. To study the typical inter-bees interaction length over time we computed (for each bee) its distance to their nearest neighbor (NN) (between their centers of masses) at intervals of 5 s. It can be noted that the mean distance is of the same order as the typical bee’s body length \( \mu = 10\;{\text{mm}} \pm 2 \)) and that exhibits daily fluctuations as shown in Fig. 2A,B. The distribution of nearest neighbor distances that a single bee experiences also fluctuates, as illustrated by the histograms shown in Panels (D) and (E) of Fig. 2, which seem to be approximated by a log-normal function (see insets showing the histograms plotted in log-linear axis). As expected the bees’ displacement speed also exhibits changes. Notice in panel C the circadian fluctuations in speed as the experiment evolves, as well as the one-fold change at the end of the experiment, when the bee density in the hive decreases after the foragers were not allowed to return to the hive.

Now we turn to explore the collective properties of the fluctuations described. The first indication is given by the fluctuations in the polarization \(\Phi \), a parameter commonly used to quantify the degree of global order in collective motion^{17}. This metric ranges from 0 to 1 and is higher when directions of individual velocity vectors \(\vec {v_{i}}\) are aligned.

By simple visual inspection of the traces in Fig. 2 it can be noticed that the increases in the nearest neighbor variance are associated with an increase in the polarization (although less obvious it seems that the increases in polarization seem to precede the spread of the bees nearest neighbor distances). Also is worth mentioning that polarization and speed seem to hold an inverse relation, an observation that will be commented later on. These results show that more packed conditions (i.e., smaller NN distances) and less polarized collective motion came together with slower speeds on average. Also during more polarized motion bees spread apart only slightly on the average (observe only a small change in the mean NN distance) but with notable non-uniformity.

Another approach to reveal emergent properties analyzes different statistics computed over increasingly larger ensembles. The results of this type of finite-size scaling analysis can distinguish collective properties when compared with a known null hypothesis in which there are no collective effects. For that, we use the time series of bee densities *x*(*t*) recorded at each of the 600 square grids. After constructing ensembles of increasing size N we computed the statistics of \(f(t)=1/N \sum _{i=1}^N x(t) \) for several different stochastic combinations. The average results on Fig. 3A demonstrate a manifest anomalous scaling of the *f*(*t*) variance \(\sigma ^2\) with increasing *N*, since it remains almost constant. This scaling is in clear contrast with the 1/*N* behavior of an independent hypothesis constructed by random scrambling of the same time series. Together with the polarization calculations already commented on, this result is an indication of collective behavior since it shows that the density fluctuations inside one grid are not independent of the other grids fluctuations. In addition, we explored the so-called fluctuation scaling^{28}, also known as Taylor law^{29}, which considers the proportionality between the mean and the variance of the fluctuations in each grid. The results shown in Fig. 3B indicate that the process can not be due to independent random processes since it shows excess variance with respect to the Poisson hypothesis, demonstrated by the exponent \(\alpha > 1/2\) in the scaling \(\sigma ^2 \propto \mu ^ {-\alpha } \).

### Density–density correlations

Information relevant to untangle collective phenomena usually can be grasped from the computation of the correlation functions in time (2) and in space (3)

between the occupancy time series. In particular, the correlation decay as a function of distance, as presented in Fig. 4A, where it can be seen that the fluctuations between two points in the hive are correlated an order of magnitude beyond the spatial scale the bees interact (\(\sim 8\) mm, see Fig. 2).

It is expected, as in other complex systems near a critical point, that the properties of the correlations in space reflect into the temporal correlations^{30,31}. Thus, in Fig. 4B we estimate the behavior of the autocorrelation function as a function of the ensemble size N. The results show that the temporal correlations of larger ensembles become asymptotically longer, i.e., 1-AC(1) vanishes, in contrast with the null hypothesis results (depicted with diamonds) which remain larger and constant. AC(1) is used as a proxy for the entire autocorrelation function^{30}, in this computation, additional information concerning the scaling of the autocorrelation function with time is contained in the supplementary materials to the paper.

In addition, the correlation results in Panels A and B show an important feature: when the density in the hive decreases in phase III (due to the removal of foragers) the correlation properties seem to change to a regime that can be considered of relatively larger independence. Indeed, on the spatial aspects, the correlations shift to a weaker and shorter range (see filled circles in Panel A) which agrees with the temporal aspects where the autocorrelations become also weaker and shorter (see filled circles and diamonds in Panel B) (Additional information concerning numerical aspects of presented results is contained in the supplementary material).

We conjecture that the presence of density-dependent correlations may shed light on the mechanism responsible for the collective dynamics. To test that we computed an observable alike to traffic by first partitioning the timeseries into 5000-s intervals and computing the average occupancy of each grid square. For the same intervals, we compute the throughput of bees by dividing each square with a midline and counting the occurrences of bees crossing this line in either direction. The results, presented in Fig. 4, unveil a pattern that is interpreted as follows: For low values, the traffic is a linear function of occupancy. This is true for all phases when the occupancy within the interval is lower, but it is especially evident if only the data collected in phase III (black dots) is considered because there is one fold reduction in density due to the foragers removal (see also Fig. 2). For higher densities, there is a maximum in the traffic, beyond which any further increase in density leads to a reduction in the bee’s traffic. Considering the typical size of a bee (\(\sim 10\) mm) and the mean distance to the nearest neighbor computed (\(\sim 7\)–9 mm, see Fig. 2A) the observed maximum traffic occurs near a regime in which bees would only be able to move as a solid-like ensemble.

The changes in the traffic and the correlation properties seen in Fig. 4, together with the changes in the nearest neighbor distances shown in Fig. 2A seem to be interrelated. As a first hypothesis, these results may be explained by a process of jamming-unjamming as described in a variety of systems from car traffic^{22} to cells^{25}. Jamming is a self-organized process found both in living and inert systems in which even a small change in the control parameter, such as density, can induce big changes in the macroscopic properties of the system or material. Slow power-law decay of correlation function with distance has been empirically demonstrated for car traffic in major cities^{24}.

We should note, that although we posit the jamming transition as the generative mechanism for the scale-free correlations, as discussed above, we do not assume that it is an exhaustive explanation for all phenomena observed in the data. Thus it is evident from Fig. 2 that rhythmic oscillations of N.N. distance, mean speed, and Polarization appear in Phase II after the hive was open permitting foragers to go in and out. These fluctuations are correlated with the day–night cycle and have been likely caused by the interaction of the beehive with the surrounding environment.

### Speed correlations

Additional evidence is provided by the results in Fig. 5, which presents the correlations in speed, computed for all bees present in the hive during different phases of the experiment. Speed is computed as \(v=||\frac{\delta \vec {r}}{\delta t}||\); \(\delta t =5\) s. Distance-binned speed correlation coefficient \(C_{sp}(r_1, r_2)\) for each bin is computed as a mean correlation coefficient between the speed time series \(v_i, v_{j}\) of all bee pairs located at mutual distance *r*: \(r_1<r<r_2\), were \(r_1\) and \(r_2\) are the bin’ edges (4).

Results demonstrate that the spatial correlations in *speed* are only long-range (i.e., decaying slowly with distance as a power law) for when the bee density in the hive is high enough (Phases I and II) and become short-range (i.e., decaying faster) when the density of bees is reduced in Phase III such that their interactions are weaker.

## Conclusion

In summary, we have shown that bees in the hive exhibit complex collective dynamics that involve scale-free spatial and temporal correlations. We conjecture possible mechanisms, through which such dynamics could emerge. Several routes are open for further study including appropriate modeling efforts concerning the dynamics of jamming of active particles in 2D. Jamming transitions are well investigated with respect to traffic phenomena, however, one would expect additional effects coming from behavior germane to eusocial insects, as suggested previously^{32}.

The correlation properties observed for the fluctuations in the hive might indicate that the system is tuning itself towards an optimal state by varying its control parameter—bee density. This critical state corresponds to a density that provides maximum throughput (i.e., peak at Fig. 4). It is striking that such throughput optimization is very similar to what car drivers experience in daily traffic, including the presence of long-range correlations in speed and occupancy, much larger than the short-range interactions. The present results call for modeling attempts to validate this conjecture.

## Data availability

The data that support the findings of this study are available from authors of Ref.^{27} but restrictions apply to the availability of this data, which was used under license for the current study, and so is not publicly available. Data are however available from the authors upon reasonable request and with permission of the authors of Ref.^{27}.

## References

Wilson, E. O. Chemical communication among workers of the fire ant Solenopsis saevissima (Fr. Smith) 1. The Organization of Mass-Foraging.

*Anim. Behav.***10**, 134–164 (1962).Wilson, E. O.

*The Insect Societies*(Belknam Press, 1971).Wilson, E. O.

*Sociobiology*(Belknam Press, 1975).Rauch, E., Millonas, M. M. & Chialvo, D. R. Pattern formation and functionality in swarm models.

*Phys. Lett. A***207**, 185–193 (1995).Cavagna, A.

*et al.*Scale-free correlations in starling flocks.*Proc. Natl. Acad. Sci.***107**, 11865 (2010).Attanasi, A.

*et al.*Finite-size scaling as a way to probe near-criticality in natural swarms.*Phys. Rev. Lett.***113**, 238102 (2014).Jerome, B.

*et al.*From disorder to order in marching locusts.*Science.***312**(5778), 1402–1406 (2006).Vijayan, S., Warrant, E. J. & Somanathan, H. Defensive shimmering responses in Apis dorsata are triggered by dark stimuli moving against a bright background.

*J. Exp. Biol.***225**(17), jeb244716 (2022).Cavagna, A., Giardina, I. & Grigera, T. S. The physics of flocking: Correlation as a compass from experiments to theory.

*Phys. Rep.***728**, 1 (2018).Shishkov, O. & Peleg, O. Social insects and beyond: The physics of soft, dense invertebrate aggregations.

*Collective Intel.***1**(2), 1–18 (2022).Sethna, J. P.

*Statistical Mechanics: Entropy, Order Parameters, and Complexity*Vol. 14 (Oxford University Press, 2021).Chialvo, D. R. Emergent complex neural dynamics.

*Nat. Phys.***6**, 744–750 (2010).Beggs, J. M. & Timme, N. Being critical of criticality in the brain.

*Front. Physiol.***3**, 163 (2012).Pawel, Romanczuk, & Daniels, Bryan C. Phase transitions and criticality in the collective behavior of animals-self-organization and biological function.

*Order, Disorder and Criticality: Advanced Problems of Phase Transition Theory*, pp. 179-208. (2023).Solé, R., Moses, M. & Forrest, S. Liquid brains, solid brains.

*Philos. Trans. R. Soc. B***374**(1774), 20190040 (2019).Couzin, I. D. Collective cognition in animal groups.

*Trends Cognit. Sci.***13**(1), 36–43 (2009).Attanasi, A.

*et al.*Collective behaviour without collective order in wild swarms of midges.*PLoS Computational Biol.***10**(7), e1003697 (2014).Daniels, B. C., Krakauer, D. C. & Flack, J. C. Control of finite critical behavior in a small-scale social system.

*Nat. Commun.***8**(1), 14301 (2017).Seeley, T. D.

*Honeybee Democracy*(Princeton University Press, 2011).Klamser, P. P. & Romanczuk, P. Collective predator evasion: Putting the criticality hypothesis to the test.

*PLoS Comput. Biol.***17**(3), e1008832 (2021).Ginelli, F.

*et al.*Intermittent collective dynamics emerge from conflicting imperatives in sheep herds.*Proc. Natl. Acad. Sci.***112**(41), 12729–12734 (2015).Nagel, K. & Schreckenberg, M. A cellular automaton model for freeway traffic.

*J. Phys. I***2**(12), 2221 (1992).Taillanter, E. & Barthelemy, M. Empirical evidence for a jamming transition in urban traffic.

*J. R. Society Interface***18**(182), 20210391 (2021).Petri, G.

*et al.*Entangled communities and spatial synchronization lead to criticality in urban traffic.*Sci. Rep.***3**(1), 1798 (2013).Chepizhko, O.

*et al.*Bursts of activity in collective cell migration.*Proc. Natl. Acad. Sci.***113**(41), 11408–11413 (2016).Van Drongelen, R. Collective dynamics and jamming in biological systems. (2018).

Gernat, T.

*et al.*Automated monitoring of behavior reveals bursty interaction patterns and rapid spreading dynamics in honeybee social networks.*Proc. Natl. Acad. Sci.***115**(7), 1433–1438 (2018).de Menezes, M. A. & Barabasi, A.-L. Fluctuations in network dynamics.

*Phys. Rev. Lett.***92**, 28701 (2004).Eisler, Z. & Kertesz, J. Scaling theory of temporal correlations and size-dependent fluctuations in the traded value of stocks.

*Phys. Rev. E***73**, 046109 (2006).Grigera, T. S. Correlation functions as a tool to study collective behaviour phenomena in biological systems.

*J. Phys. Complexity***2**(4), 045016 (2021).Camargo, S.

*et al.*Scale-free correlations in the dynamics of a small (\(N\sim 10000\)) cortical network.*Phys. Rev. E***108**(3), 034302 (2023).Gallotti, R. & Chialvo, D. R. How ants move: Individual and collective scaling properties.

*J. R. Soc. Interface***15**(143), 20180223 (2018).

## Acknowledgements

We are indebted to the authors of Ref.^{27} for sharing the data and to T. Gernart for help in clarifying data and recording details. I.S. thanks the hospitality of UNSAM and to the members of the Instituto de Ciencias Físicas, where this work was conducted. I.S. was supported by funding from ECSU Unit in OIST. This research was supported by Grant no. 1U19NS107464-01 from NIH BRAIN Initiative, by CONICET (Argentina) and Escuela de Ciencia y Tecnología, UNSAM.

## Author information

### Authors and Affiliations

### Contributions

D. C. and I.S. wrote the main manuscript text and prepared the figures. All authors participated in editing the manuscript and agreed to the publication.

### Corresponding author

## Ethics declarations

### Competing interests

The authors declare no competing interests.

## Additional information

### Publisher's note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

## Supplementary Information

## Rights and permissions

**Open Access** This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

## About this article

### Cite this article

Shpurov, I., Froese, T. & Chialvo, D.R. Beehive scale-free emergent dynamics.
*Sci Rep* **14**, 13404 (2024). https://doi.org/10.1038/s41598-024-64219-w

Received:

Accepted:

Published:

DOI: https://doi.org/10.1038/s41598-024-64219-w

## Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.