understanding how animal groups achieve coordinated movement · review understanding how animal...
TRANSCRIPT
REVIEW
Understanding how animal groups achieve coordinatedmovementJ. E. Herbert-Read1,2,*
ABSTRACTMoving animal groups display remarkable feats of coordination. Thiscoordination is largely achieved when individuals adjust theirmovement in response to their neighbours’ movements andpositions. Recent advancements in automated trackingtechnologies, including computer vision and GPS, now allowresearchers to gather large amounts of data on the movements andpositions of individuals in groups. Furthermore, analytical techniquesfrom fields such as statistical physics now allow us to identify theprecise interaction rules used by animals on the move. Theseinteraction rules differ not only between species, but also betweenindividuals in the same group. These differences have wide-rangingimplications, affecting how groups make collective decisions anddriving the evolution of collective motion. Here, I describe howtrajectory data can be used to infer how animals interact in movinggroups. I give examples of the similarities and differences in thespatial and directional organisations of animal groups betweenspecies, and discuss the rules that animals use to achieve thisorganisation. I then explore how groups of the same species canexhibit different structures, and ask whether this results fromindividuals adapting their interaction rules. I then examine how theinteraction rules between individuals in the same groups can alsodiffer, and discuss how this can affect ecological and evolutionaryprocesses. Finally, I suggest areas of future research.
KEY WORDS: Collective motion, Collective behaviour, Interactionrules, Leadership, Social responsiveness
IntroductionSchools of fish, flocks of birds and marching insects achievecoordination (seeGlossary) without choreography. Individuals oftencoordinate their movements to reduce predation risk through dilutionor confusion effects (Krause and Ruxton, 2002). In other cases,collective motion (see Glossary) may be an adaptation to improveforaging success (Bazazi et al., 2012) or to pool information aboutthe direction of new feeding, breeding or nest sites, therebyimproving migration efficiency (Codling et al., 2007; Seeley andBuhrman, 1999; Grünbaum, 1998). Collective motion also emergesthrough antagonistic contacts between individuals, such as whenMormon crickets (Anabrus simplex) or juvenile desert locusts(Schistocerca gregaria) chase and avoid conspecifics incannibalistic interactions (Simpson et al., 2006; Bazazi et al.,2008; Hansen et al., 2011). The reasons that other animals move ingroups remain unresolved; for example, at high densities, plant–
animal worms (Symsagittifera roscoffensis) form rotating mills, thefunction of which remains unclear (Franks et al., 2016). In othersystems, collectivemotionmayemerge through repeated interactionsbetween individuals in confined environments (Mann et al., 2013).
The explanations for how animals can achieve these levels ofcoordination have ranged from mere coincidence to thoughttransference among individuals (Selous, 1931). We now know thatindividuals in moving groups respond to the local movements andpositions of their neighbours, and these ‘interaction rules’ (seeGlossary) between individuals produce coordinated movement. Theterm ‘interaction rules’ is adopted from earlierworkon self-propelledparticle models (Box 1). These rules may alternatively be thought ofas ‘decisions’, which determine how information gathered frommultiple sensory inputs affects an individual’s motor response.These inputs could include direct visual or mechanical stimuli fromconspecifics, or indirect social cues such as air or water turbulencecreated by others’ movements. Thinking about these decisions as‘rules’, therefore, can provide a useful framework for understandinghow animals in groups achieve coordinated movement.
How individuals interact in moving groups ultimately affects theirsurvival and reproductive success. Hence, deciphering theseinteractions allows us to understand the selective pressures thathave shaped these social behaviours. These social interactions arealso likely to affect how individuals within and between groups mixin populations (Morales et al., 2010), potentially affecting largerevolutionary processes such as speciation. At a finer level, the hightemporal and spatial resolution data collected from moving animalgroups allows us to map individuals’ behavioural responses to theirsensory inputs. Combining detailed movement analysis withmodern techniques to monitor neural activity (Grover et al., 2016)will produce key insights into the neural mechanisms that governseemingly complex social behaviours. Furthermore, we may alsounderstand the genetic basis underlying social behaviouraldifferences by mapping detailed behavioural variation in howindividuals interact in groups to the associated genetic variationbetween individuals. This level of understanding, linking genetics,neuroscience and behaviour, can only be achieved if we canaccurately measure how individuals interact within groups.
Owing to the large amounts of trajectory data that can becollected using automated methods, we are now amassinginformation that describes how animals interact in moving groups.This Review will describe how these trajectory data can be used toinfer social interaction rules. It will then describe variation in theinteraction rules between species, between groups and betweenindividuals, before finally exploring the implications of thisvariation, suggesting possible paths for future research.
Analysing interaction rules in animal groupsModels have been useful for understanding the general principles ofcollective motion (see Box 1), but they are no substitute for datacollected from real animal groups. Methods for collecting data on
1Department of Zoology, Stockholm University, SE-10691 Stockholm, Sweden.2Department of Mathematics, Uppsala University, S-75106 Uppsala, Sweden.
*Author for correspondence ( [email protected])
This is an Open Access article distributed under the terms of the Creative Commons AttributionLicense (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use,distribution and reproduction in any medium provided that the original work is properly attributed.
2971
© 2016. Published by The Company of Biologists Ltd | Journal of Experimental Biology (2016) 219, 2971-2983 doi:10.1242/jeb.129411
Journal
ofEx
perim
entalB
iology
the positions and movements of individuals in groups have beenaround for some time (Radakov, 1973; Partridge, 1980, 1981), butautomated tracking technologies (see Glossary) based on computervision (Cavagna et al., 2008a,b; Branson et al., 2009; Straw et al.,2011; Pérez-Escudero et al., 2014; Handegard and Williams, 2008)and miniature GPS (Nagy et al., 2010) have made it easier to collectvast amounts of data from individuals in groups. Furthermore,statistical physics has provided tools to analyse these trajectories anddetermine how individuals within groups are interacting (Cavagnaet al., 2008c; Nagy et al., 2010). These methods now allow theinvestigation of interaction rules in real animal groups, withoutrelying on predictive computer simulations.Tracking tools provide information on the position (usually the
centre of mass) of animals within their environment. One of the firstpatterns to investigate with these data is the relative positions ofindividuals with respect to one another. When animals move along ahorizontal plane, the position of a neighbour (N1) relative to a focalindividual can be described by two measures: the bearing angle (θ)and the distance to that neighbour (d; Fig. 1). The proportion of timefor which neighbours are located at certain distances and bearingsrepresents the likelihood of finding neighbours at those positions.Because the spatial distribution of individuals with respect to oneanother is fundamentally linked to the rules that individuals use tointeract, these relative positions can be particularly informative forinferring not only which individuals are interacting, but also howthey are interacting.If positional readings are taken over time, then themovements of an
animal between two points can be inferred. The vector between twosuccessive positions gives information on that animal’s speed anddirection. By factoring in the movements and positions of otherindividuals in the group, the movements of individuals with respect to
one another can bemeasured (Fig. 1). This reveals how individuals areadjusting their position and orientation in response to their neighbours’movements and positions. This is done, for example, by correlating ananimal’s acceleration or turning angle with the position, orientation orother movements of its neighbour (Lukeman et al., 2010; Katz et al.,2011; Herbert-Read et al., 2011; Pettit et al., 2013). These fine-scalemovement responses can be scaled to the whole group; individuals’directional changes can be used to infer how quickly information ispropagated across groups (Attanasi et al., 2014a; Herbert-Read et al.,2015a). Changes in the speed and direction of individuals over timecan also highlight other dynamics occurring within moving groups.For example, leader–follower relationships can be inferred by seeingwhether the speed or direction changes of a focal individual areadopted (after some time delay) by neighbouring individuals. Ifindividual A’s direction (or speed) is copied by conspecific B, then wecan infer that Bwas followingA. Thus,we can build a picture ofwhichindividuals are more influential in guiding group movement (Nagyet al., 2010). Fine-scale tracking data, therefore, allow us to infer therules that individuals are using to maintain cohesion and order inmoving animal groups.
It is important to note that these rules are proxies for how ananimal perceives information about its surroundings and integrates
GlossaryCoordination (in motion)The synchronisation of individuals’ movements in time and space.
Collective motion/movementWhen individuals move together in groups through non-independentinteractions with one another.
Directional organisationThe degree to which individuals align with each other’s headings ingroups. Neighbours that have lowangular deviation in headings aremorepolarised than neighbours that have high angular deviation in headings.Lower directional organisation is characteristic of swarming states,whereas higher directional organisation in characteristic of milling andflocking states.
Interaction rulesHow individuals adjust their movements depending on informationgathered from their local neighbours’ movements and positions. Theterm ‘interaction’ describes the non-independence between individuals’movements.
Spatial organisationHow individuals position themselves with respect to one another ingroups. Neighbours can be located uniformly in all directions relative to afocal individual (isotropy) or in particular directions relative to a focalindividual (anisotropy). Individuals often attempt to maintain separationdistances between one another, resulting in density regulation withingroups.
TrackingThe process by which the position (and sometimes orientation) of ananimal is measured in space and time.
Box 1. A brief history of collective motionThe concept of interaction rules can be traced back to the 1950s, whenBreder proposed that schooling fish have attraction and repulsion forcesthat maintain the distance between neighbouring individuals (Breder,1954). With the development of general-purpose computers in the1980s, it became feasible to simulate groups of individuals interactingaccording to such simple rules. In 1982, Aoki simulated particles thatadjusted their direction according to the position of their neighbours,moving away from one another at close distances, aligning atintermediate distances and moving towards each other at greaterdistances (Aoki, 1982). Particles obeying these simple rules formedcohesive and coordinated moving groups, even though individuals werenot following specific individuals and did not know all of the otherparticles’ movements or positions. Later, computer game developerswanted to simulate the behaviour of animal groups without having tocode individual trajectories. Reynolds (1987) provided a solution,implementing a flocking model with interacting particles. Like Aoki’smodel, this included terms such as collision avoidance, velocitymatching(speed and direction) and attraction, with particles only having localinformation. Aoki and Reynolds determined that simple movement rulesbetween neighbouring individuals could generate cohesive andcoordinated motion, much like that of real animal groups.Other researchers began to implement variations in the interaction
rules that particles obeyed (Huth and Wissel, 1992; Niwa, 1994) andassess the properties of these models (Vicsek et al., 1995). Romey(1996) and Couzin et al. (2002) varied the rules of individuals, givingparticles different repulsion and attraction strengths in a group. Couzinet al. (2002) also investigated interaction rules in three dimensions;changing the size of the zones of alignment and attraction resulted ingroups displaying different configurations. Other slight variations in themodels had interesting effects on the collective patterns that emerged(e.g. altering body shape and the size and shape of interaction zonesaffects spatial sorting in simulated groups; Kunz and Hemelrijk, 2003;Hemelrijk and Kunz, 2005; Hemelrijk and Hildenbrandt, 2008; Romeyand Vidal, 2013). Bode et al. (2010) investigated how group-levelproperties such as the speed and distribution of neighbours changedwhen individuals altered how often they updated their position. Otherdetailed models captured the dynamics and spatial organisation of birdflocks (Hildenbrandt et al., 2010; Cavagna et al., 2015; Lukeman et al.,2010), fish schools (Hemelrijk and Kunz, 2005; Kunz and Hemelrijk,2003) and insect swarms (Romanczuk and Schimansky-Geier, 2012).For reviews, see Vicsek and Zafeiris (2012) and Sumpter (2010).
2972
REVIEW Journal of Experimental Biology (2016) 219, 2971-2983 doi:10.1242/jeb.129411
Journal
ofEx
perim
entalB
iology
this information to then perform a motor response. Nevertheless,correlations between certain stimuli (e.g. distance to neighbour) andoutput (e.g. change in speed) reveal basic but important principlesgoverning an individual’s decision to move. It will be important tofind more biologically meaningful sensory inputs that govern thesebehavioural responses, for example, by correlating the retinal size ofan object or neighbour with an animal’s velocity (Boeddeker et al.,2003). Indeed, newer analytical techniques that consider the bodyshape and potential visual fields of individuals in groups will lead toa deeper understanding of the sensory inputs that elicit behaviouralmovement responses (Strandburg-Peshkin et al., 2013; Rosenthalet al., 2015). Nevertheless, basic measures such as the distance toand speed of a neighbour are inextricably linked to the sensory cuesthat individuals use to detect the movement and positions ofneighbours. They can, therefore, reveal important insights into thesimilarities and differences between how individuals interact inmoving animal groups.
Differences and similarities between speciesSpatial organisationOne of the striking differences between animal groups is the spatialorganisation (see Glossary) of individuals within them. For locust
nymphs that move along the ground, the distribution of nearestneighbours around a focal individual is uniformly distributed in alldirections, with neighbours most commonly 3–10 cm apart(Fig. 2A) (Buhl et al., 2012). This type of distribution is referredto as spatially isotropic. In other species that move along ahorizontal plane, nearest neighbours are more often located inspecific directions relative to a focal individual (spatiallyanisotropic). Surf scoters (Melanitta perspicillata), which swimacross the water’s surface, have clearly defined nearest-neighbourdistributions. Birds seldom come within 1 body length of oneanother and are most commonly found 1.45 body lengths apart, infront of or behind one another (Lukeman et al., 2010). When thepositions of pairs of fish relative to one another are measured in thehorizontal plane, mosquitofish (Gambusia holbrooki), goldenshiners (Notemigonus crysoleucas) and minnows (Phoxinusphoxinus) position themselves in front of or behind theirneighbour at ∼1.5–2 body lengths apart (Herbert-Read et al.,2011; Katz et al., 2011; Partridge, 1980). Other species, such asgiant danios (Danio aequipinnatus), homing pigeons (Columbalivia) and starlings (Sturnus vulgaris), position themselves moreoften side by side with respect to their direction of motion (Fig. 2B)(Grünbaum et al., 2005; Pettit et al., 2013; Ballerini et al., 2008).Pairs of female guppies (Poecilia reticulata) most often positionthemselves in diagonally offset configurations (Fig. 2C). In somespecies, however, the distribution of nearest neighbours oftenbecomes more isotropic as group size or density increases(Partridge, 1980; Partridge et al., 1983; Katz et al., 2011).
For animals that can also move in the vertical plane, thisadditional direction of motion allows individuals to vary theirrelative elevations. Although pairs of minnows position themselvesone in front of another with little difference in their elevation, asschool size increases, individuals increase the distance between nearneighbours in the vertical plane (Partridge, 1980). In effect,individuals position themselves slightly above or below oneanother. Similar results have been found in saithe (Pollachiusvirens), herring (Clupea harengus), dunlin (Calidris alpina),starlings (Sturnus vulgaris) and mysid shrimp (Tenagomysisoculata) (Partridge et al., 1980; Major and Dill, 1978; Balleriniet al., 2008; O’Brien, 1989). Because of the difficulties of collectingtrajectory data in three dimensions, it is currently unclear how
N2
N1
d
α
θ
φ
Fig. 1. Tracking data can provide information on the relative positions andorientations of individuals in groups. From these data, the distance (d ),direction (θ) and differences in heading (φ) between a focal individual and itsneighbours (N1 or N2) can be quantified. The acceleration, speed or turningangle (α) of the focal individual can then be correlated with thesemeasurements to infer how individuals in groups are responding to eachother’s movements and positions.
0
−10
−20
20
10
0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
Distance left−right (cm)
Dis
tanc
e ba
ck−f
ront
(cm
)
Number of observations
Distance left−right (cm)
Dis
tanc
e fro
nt−b
ack
(cm
)
Distance left−right (m)
Dis
tanc
e ba
ck−f
ront
(m)
−10 0 10−10
−5
0
5
10
Frequency
0
5000
10,0006543210−1−2−3−4−5−6
A CBB
Rel
ativ
e de
nsity
of n
eigh
bour
s
−20 −10 0 10 20 −4 −3 −2 −1 0 1 2 3 4
350 300 250 200 150 100 50
Fig. 2. The relative density of neighbours around a focal individual for different species. The focal individual is represented by the white silhouette in thecentre of each plot (not to scale) and is facing in its direction of motion. (A) Locusts are commonly found 3–10 cm apart and in any direction surrounding a focalindividual. Lower densities of neighbours occur within a 1 cm region close to an individual. Adapted from Buhl et al. (2012). (B) Pigeons (Columba livia) arecommonly found side by side with respect to their direction of motion. Adapted from Pettit et al. (2013). (C) Female guppies (Poecilia reticulata) in pairs positionthemselves in front of or behind, and to the left or right of a partner (author’s unpublished data).
2973
REVIEW Journal of Experimental Biology (2016) 219, 2971-2983 doi:10.1242/jeb.129411
Journal
ofEx
perim
entalB
iology
individuals integrate information about their neighbours’ positionand movements in the vertical plane – this will be briefly returned tolater on.The spatial positions individuals adopt in groups may be partly
determined by the available sensory modalities (Box 2) or theenergetics associated with moving in groups (Box 3). In manycases, however, whether there are functional benefits for individualsadopting specific positional preferences remains unclear (Hemelrijkand Hildenbrandt, 2012). Nevertheless, the relative positions thatindividuals adopt in groups can be used to infer which individualsare interacting. Ballerini et al. (2008) measured the degree ofanisotropy between successive nearest neighbours in starling flocks.They found that the relative positions of the eighth and furthernearest neighbours did not differ from random. This occurredregardless of a flock’s density. This is consistent with individualsinteracting with a fixed number of near neighbours (termedtopological interactions) instead of reacting to individuals withinsome distance of each other (metric interactions). The dichotomybetween topological and metric interactions, however, may not beabsolute; individuals may move away from a specific individual thatis too close, but may be attracted towards multiple individualsfurther away. In a more sensory-based approach, Strandburg-Peshkin et al. (2013) examined golden shiner schools, and foundthat the likelihood of an individual responding to the movements ofothers was best predicted by assessing which neighbours could see
each other. By assessing the visual network of these fish, Rosenthalet al. (2015) went on to predict the magnitude of informationpropagation across these schools. Therefore, assessing the basicstructure and positions that individuals adopt in animal groups canreveal which individuals are interacting in groups, and highlightsthat the range and type of these interactions, perhaps unsurprisingly,differs between species.
The spatial positions individuals adopt in groups allows us toinfer not only who is interacting with whom, but also how theseindividuals are interacting. If individuals have preferred positionsthat they wish to maintain, then movements to correct deviationsfrom these positions can be thought of as interaction rules. Pernaet al. (2014) investigated how the interaction rules of simulatedgroups would appear if individuals had different spatial preferenceswith respect to their neighbours. If animals preferred to maintainfront–back configurations with some separation distance,individuals would accelerate towards neighbours far ahead ofthem, and decelerate when neighbours were directly in front of them(Perna et al., 2014). Indeed, animals that maintain front–backconfigurations adopt these predicted movement rules. Theinteraction rules of golden shiners and mosquitofish have beenanalysed in detail (Katz et al., 2011; Herbert-Read et al., 2011).Pairs of fish mainly position themselves in front of and behind oneanother. Fish are rarely observed within 1 body length of oneanother, and are most often ∼1.5–2 body lengths apart. The fishmodulate their speed depending on their neighbour’s position.When their neighbour is directly in front of them (<1.5 bodylengths), they decelerate, and when their neighbour is directlybehind them (<1.5 body lengths), they accelerate. When theirneighbour is >1.5 body lengths in front of or behind them, theyaccelerate or decelerate, respectively. The fish have close to zeroacceleration when their neighbour is∼1.5 body lengths in front of orbehind them (Fig. 3A). These changes in speed, based onneighbours’ positions, have also been observed in juvenileblacksmith (Chromis punctipinnis) and zebrafish (Danio rerio)(Parrish and Turchin, 1997; Zienkiewicz et al., 2015).
Acceleration and deceleration responses are not only important inmaintaining separation distances between fishes, but also crucial inmaintaining coordination during tandem running in Temnothoraxants (Franks and Richardson, 2006).During these runs, the followerattempts to maintain antennal contact with the lead ant. When theleader’s gaster (or tip of her abdomen) is within twice the length ofthe follower’s antennae, the follower decelerates, whereas the leader
Box 2. Sensory determinants of spatial organisationDifferences in the spatial organisation of groups may be partly linked tothe sensory modalities animals use to interact. Temnothorax ants(Temnothorax albipennis) perform tandem runs, where one ant leadsanother to a new nest site or food source (Pratt, 2008; Franks et al.,2002). Franklin et al. (2011) found that vision was not necessary tocoordinate this movement, whereas tactile interactions and pheromoneswere. Fitzgerald and Pescador-Rubio (2002) also found that tactilestimuli and pheromones were important in establishing and maintainingsingle-file processions in social caterpillars (Hylesia lineata). Theseparation distances between whirligig beetles (Dineutes discolor) arecontrolled by both antennal detection of surface waves produced bynearby conspecifics and visual cues (Romey et al., 2014). Similarly, theseparation distances between locusts are controlled by visual and tactileinteractions (Bazazi et al., 2008). Because locusts have 360 deg vision inthe horizontal direction (Rogers et al., 2010), they may use thisinformation to maintain minimal separation distances betweenneighbours, but without directional preferences for where thoseneighbours are (Buhl et al., 2012). In contrast, animals that positionthemselves at specific bearing angles with respect to one another mayhave limitations to where neighbours can be detected (Hemelrijk andHildenbrandt, 2012). For example, starlings have a minimal blind anglebehind the head of 32 deg (Martin, 1986), and may attempt to positionneighbours where they can see them (Ballerini et al., 2008). However,this does not explain why some species of fish, which also have blindangles, position themselves more often in front or behind one another(Herbert-Read et al., 2011; Katz et al., 2011). Instead, individuals mayattempt to position neighbours in regions of their field of view wherechanges to the optic flow, looming rates or other movements ofneighbours can be more easily detected (Dill et al., 1997). Otherspecialised organs, such as the lateral line in some species of fish, allowindividuals to maintain separation distances (Pitcher et al., 1976). If thelateral line organ is cut, saithe (Pollachius virens) tend to positionthemselves side by side, rather than in their regular front–behindconfiguration (Partridge and Pitcher, 1980). Many species, therefore,integratemultiple sensory inputs to determine neighbours’ locations, andwhich sensory inputs are available can affect the movements and spatialpositioning of individuals in groups.
Box 3. Energetic determinants of spatial organisationThe hydrodynamic or aerodynamic effects associated with moving ingroups are sometimes important in determining the spatial positions thatindividuals adopt (Killen et al., 2011; Marras et al., 2015). For example,model simulations predict that individuals in fish schools swim moreefficiently in ‘diamond’ configurations (Weihs, 1975; Hemelrijk et al.,2015a). In nature, bald ibises (Geronticus eremita) occupy positions in Vformations that are consistent with aerodynamically optimal positions(Portugal et al., 2014). Pelicans (Pelecanus onocrotalus) have reducedheart rates and wing-beat frequencies when flying in V formations thanwhen alone (Weimerskirch et al., 2001). Therefore, in some groups,individuals may occupy positions that reduce their energy expenditurewhen on the move. However, when the density of individuals changes,but their directional preference does not, the hydrodynamic oraerodynamic benefits of grouping cannot explain such positioningbehaviour (Ballerini et al., 2008). Indeed, for pigeons, there areenergetic costs associated with flying in flocks, suggesting otherreasons for why pigeons flock (Usherwood et al., 2011).
2974
REVIEW Journal of Experimental Biology (2016) 219, 2971-2983 doi:10.1242/jeb.129411
Journal
ofEx
perim
entalB
iology
accelerates. However, when the ants are more than this distanceapart, the leader decelerates and the follower accelerates(Fig. 4A,B). These simple behavioural rules, which can bethought of as repulsion and attraction forces, allow the pair tomove whilst remaining together. In many species, changes in speedare particularly important for maintaining order and cohesion whenindividuals move together.
Changes in speed are not the only way for an individual to adjustits position relative to its neighbours. Individuals can also adjusttheir direction. Indeed, in addition to changes in speed, goldenshiners and zebrafish adjust their direction depending on theposition of their neighbours (Katz et al., 2011; Zienkiewicz et al.,2015). An individual will turn away from a neighbour that ispositioned <1 body length and directly to the side of itself, acting to
010203040 10
5
0
0.05
A B
Dis
tanc
e fro
m n
eigh
bour
(cm
)
Distance from neighbour (cm) Distance from neighbour (m)
Acc
eler
atio
n (c
m s−2
)
Turn
ang
le (r
ad s−1
)
−10−20−30−40
0 20 40−20−40
3
2
1
0
−1
−2
−3 Dis
tanc
e fro
m n
eigh
bour
(m)
−10
−5
−10 0 10
−0.05
0
Fig. 3. The acceleration and turning responses of mosquitofish (Gambusia holbrooki) and pigeons (Columba livia). (A) Acceleration response of a fish inthe centre of the plot and facing right, as a function of distance from its neighbour. There is a region of repulsion close to the focal individual, and attraction occurs atdistances further away. Together, these act to maintain distances between individuals within the group. Adapted fromHerbert-Read et al. (2011). (B) Pigeons turntowards or away from their neighbour depending on the position and distance of that neighbour. Positive turning angles represent turns to the right, whereasnegative turning angles represent turns to the left. Pigeons will turn towards their neighbour when it is >3 m away, and away from their neighbour when it is <3 maway. Adapted from Pettit et al. (2013). The fish and bird silhouettes in the centre of the plots are not to scale.
8
9. Such pairs move three times
RHead−gaster distance (mm)
A
B
−20
0
20
40
0 100
Lead
erac
cele
ratio
n (m
m s−2
)
C
DDistance from neighbour
back−front (mm)
−40
Follo
wer
acce
lera
tion
(mm
s−2
)
−20
0
20
40
−40
0 2 4 6 8 10 12
RHead−gaster distance (mm)
0 2 4 6 8 10 12
−20
0
20
40
Acc
eler
atio
n (m
m s−2
)
−40−100
0 100Distance from neighbour
left−right (mm)
−100
−50
0
50
Turn
ing
spee
d (d
eg s−1
)
Fig. 4. Leader and followership in pairs of ants (Temnothorax alibipennis) and mosquitofish (Gambusia holbrooki). (A,B) Acceleration responses of theleader (A) or follower (B) ant, depending on the distance to their partner. When the pair are <∼1 mm apart (R=1 mm), the leader will accelerate, whilst the followerwill decelerate (blue circles left of R). When the pair are between ∼1 and 2 mm apart, the follower will accelerate and the leader will decelerate (blue circles right ofR). Both leader and follower show close to zero acceleration when >2 mm apart (red circles). Adapted by permission from Macmillan Publishers Ltd, Nature;Franks and Richardson (2006). (C) Mean change in speed and (D) mean change in angle of mosquitofish that were leaders (solid red curve) or followers (solidblue curve) as a function of their partner’s location. In C, negative x regions indicate that the focal individual (either leader or follower) was behind its partnerand positive x regions indicate that the focal individual was in front of its partner. In D, negative y regions indicate that the focal individual was to the left of itspartner, and positive y regions indicate that it was to the right of its partner. Dashed curves are plotted one standard error above and belowall means in each panel.These responses show how individuals differentially adjust their velocity as a function of their neighbours’ location. Grey regions represent the regions where themovement responses of leaders and followers are significantly different. Adapted from Schaerf et al. (2016).
2975
REVIEW Journal of Experimental Biology (2016) 219, 2971-2983 doi:10.1242/jeb.129411
Journal
ofEx
perim
entalB
iology
maintain a minimum distance between pairs. Lukeman et al. (2010)also found that the headings of birds in front of or behind oneanother deviated strongly in surf scoter flocks. This could beattributed to individuals avoiding neighbours that were directly infront of them. Army ants (Eciton burchelli) also use turningresponses along their foraging trails to avoid oncoming neighbours(Couzin and Franks, 2003). If an ant returning along the trailcollides with an outgoing ant, the outgoing ant will turn away fromthe incoming ant, giving the returner priority in its direction oftravel. This asymmetric interaction rule can produce lanes of traffic,minimising congestion along the trail. Turning responses can beused, therefore, to move away from neighbours that are too close inorder to maintain separation distances.When neighbours are far away from one another, however,
turning responses can also act to bring individuals back together.When golden shiners or zebrafish are >1 body length apart, they turntowards their neighbours (Katz et al., 2011; Zienkiewicz et al.,2015). The turning responses of mosquitofish are similarlydependent on the position of their neighbours. Fish turn left whentheir neighbour is on their left, and turn right when their neighbouris on their right (Herbert-Read et al., 2011). In another study onbarred flagtails (Kuhlia mugil), fish kept relatively smoothtrajectories and adjusted their turning angle as a function of theirtime to collision with the wall and their neighbour’s position(distance and angle towards neighbour) (Gautrais et al., 2012). Pettitet al. (2013) also found that pigeons changed their angle of headingdepending on the distance to their neighbour. A pigeon will turntowards a neighbour >3 m away, and will turn away from aneighbour <3 m away (Fig. 3B). As well as speed changes, turningresponses depending on neighbours’ positions maintain the spatialorganisation of individuals in bird flocks, fish schools and along anttrails. Which rules animals adopt is likely to depend on how easy itis to change speed or direction in the medium the organisms aretravelling through, whilst at the same time maintaining control.
Directional organisationAbove, I discussed how individuals can maintain their relativepositions in groups by changing their speed and direction dependingon their neighbour’s position. However, groups show not onlyspatial organisation, but also directional organisation (seeGlossary). That is, individuals in groups often move in the samedirection, with small angular deviations in their heading. This couldoccur if individuals changed their direction to align with that of theirneighbours. To detect alignment responses, an individual’s turningresponse (α in Fig. 1) should be correlated with the direction itsneighbour is facing (Φ in Fig. 1), and not only with that neighbour’sposition (θ and d in Fig. 1). In some species, these effects can beobserved. Barred flagtails, for example, tend to align with theirneighbours’ direction of travel (Gautrais et al., 2012). Pigeonssimilarly take into account the direction in which a neighbour isheading when deciding in which direction to turn (Pettit et al.,2013). If a neighbouring pigeon is to the immediate left of a focalindividual and is turning right, then the focal individual is moreinclined to turn right as well. In effect, partners close to one anothermatch their orientations. Perna et al. (2014) predicted that animalsthat exhibited alignment responses would often position themselvesside by side, and indeed this is observed for pigeons (Fig. 2B).Directional organisation is also observed when honeybee swarms
are guided to a new nest site (Beekman et al., 2006). Schultz et al.(2008) and Latty et al. (2009) concluded that uninformed swarmmembers are attracted towards fast-flying ‘streaker’ bees thatprovide directional information on the new nest site’s location.
However, it is unclear whether bees move towards or align withthese fast-flying neighbours. Other species, such as golden shinersand mosquitofish, form polarised groups even though individualsdo not appear to explicitly align with their neighbour’s direction oftravel (Katz et al., 2011; Herbert-Read et al., 2011). Individuals ofboth species turn towards the location of their neighbour (when thefish are not trying to move apart), but not towards the direction thisneighbour is facing. Buhl et al. (2012) also found that pairs oflocusts <13.5 cm apart travelled in the same direction; however,whether individuals explicitly aligned with their neighbours wasequivocal. Earlier models of locust marching and fish schoolingoften included alignment terms (Buhl et al., 2006; Yates et al., 2009;Couzin et al., 2002), but lack of evidence for explicit directionalmatching has inspired a new set of models that do not include theseterms (Romanczuk and Schimansky-Geier, 2012; Strömbom,2011). These models demonstrate that groups can exhibitdirectional organisation even without explicit alignment terms.Instead, directional organisation is brought about through acombination of repulsion and preferential attraction to others inlead positions (Katz et al., 2011).
Group and individual variationAlthough there are similarities and differences in the structure ofgroups and their underlying interactions between species, some ofthe most striking differences in group structure and organisation areobserved between groups of the same species. Below, I will discusshow different group states can exist with identical interaction rules. Iwill then explain how and why individuals might adjust theinteraction rules they use in groups, and highlight the consequencesthis has for a group’s structure. This will first assume that individualswithin a group are interacting using identical rules. Second, I willexplore whether individuals within the same group differ in howthey interact with one another, and discuss the implications this hasfor the ecology and evolution of moving groups.
Group variationAnimal groups can undergo rapid transformations from millingstates, with individuals circling round a central core, to polarisedstates, with individuals travelling in the same direction at the samespeed, to swarming states, with low directional alignment betweenneighbouring individuals. Golden shiners, for example, transitionbetween swarming, milling and polarised states; however, certainstates are more common for different group sizes (Tunstrøm et al.,2013). Smaller groups (30 individuals) are more often found inpolarised states, whereas larger groups (300 individuals) more oftenadopt milling states. It may be tempting to think this occurs becauseindividuals change how they interact in different group sizes. In fact,simulations show that these two states can exist with exactly thesame interaction rules (Couzin et al., 2002); indeed, there is noevidence that golden shiners change how they interact in differentlysized groups (Katz et al., 2011). Multiple stable collective states,therefore, can emerge from identical interaction rules.
Although the specific rules that individuals use to interact maynot change with group size, the frequency of interactions can(Sumpter et al., 2008). If space is limited, increasing group size willnaturally increase the group’s density. Individuals in larger schoolsof squid (Illex illecebrosus), for example, are closer together than insmaller schools (Mather and O’Dor, 1984). This will increase thefrequency of interactions between individuals because of theincreased likelihood of collisions. For locusts, the nature ofinteractions does not appear to change with group size (Buhlet al., 2006). Instead, more frequent interactions between locusts at
2976
REVIEW Journal of Experimental Biology (2016) 219, 2971-2983 doi:10.1242/jeb.129411
Journal
ofEx
perim
entalB
iology
higher densities result in larger (and denser) groups being moreordered (Buhl et al., 2006). Similar findings are reported fortadpoles (Xenopus laevis), where denser groups are also moreordered than sparser groups (Katz et al., 1981). The velocities ofindividual midges (Dasyhelea flavifrons, Corynoneura scutellataand Cladotanytarsus atridorsum) are more correlated when they arecloser together (Attanasi et al., 2014b), although this is probably notdue to individuals attempting to avoid one another, but rather thesexual interactions that occur within these swarms (Okubo andChiang, 1974). The density of individuals in groups, therefore, canaffect the frequency of interactions between individuals and therebythe group’s directional organisation.In addition to changes in the frequency of interactions, the
strength of these interactions can also change. Here, the strength ofinteractions can be interpreted as the likelihood of individualsresponding to or copying the movements or headings of nearneighbours. These changes may be thought of as adaptions toindividuals’ interaction rules. The strength of interactions is oftenlinked to the speed at which individuals are travelling in groups.Fishes, for example, will attempt to match their speed with that oftheir neighbours (Partridge, 1981), as this presumably reduces thelikelihood of collisions. At higher speeds, individuals weight theorientational information of neighbours more strongly (Gautraiset al., 2012), which is consistent with the observation that fastergroups are often more polarised (Tunstrøm et al., 2013; Viscidoet al., 2004). It is important to note here, however, that groups canalso transition from unpolarised to polarised states even if theyremain stationary (Partridge, 1980). This highlights that individualscan change how strongly they weight the directional information ofneighbours even without changes to their speed, in turn affecting thedirectional organisation of groups.Increases in speed are often a result of some external perturbation,
such as a predator’s attack. Rapid accelerations following an attackare observed in insects (Treherne and Foster, 1981), crustaceans(O’Brien, 1989), fish (Herbert-Read et al., 2015a) and birds(Beauchamp, 2012). Because increases in speed often make groupsmore polarised, in turn, polarised states may be more effective attransmitting information between group members. This occursbecause individuals can quickly copy their neighbours’ deviationsfrom current velocities (Bialek et al., 2014), resulting in ‘waves’ ofinformation travelling through groups (Procaccini et al., 2011;Hemelrijk et al., 2015b). Alternatively, individuals may adapt theirspeed according to their nutritional state, and this can affect large-scale group dynamics. Bazazi et al. (2011) found that although anindividual locust’s movement was largely unaffected by itsnutritional state when isolated, when in groups, locusts fed low-protein diets moved ∼40% faster than locusts fed a high-proteindiet. These changes in speed could be explained by individualschanging the strength of their social interactions. Because desertlocusts cannibalise conspecifics, it was hypothesised thatindividuals attempting to pursue and escape one another did somore strongly when deprived of protein. By simulating particleswith different interaction strengths, Bazazi et al. (2011) found thatthose with stronger social interactions (protein-deprivedindividuals) formed more-ordered groups at lower densities thanweakly interacting particles (protein-satiated individuals). Later,Bazazi et al. (2012) found that spadefoot toad (Spea multiplicata)tadpoles were more likely to form a vortex group structure, withtadpoles rotating in a circular movement pattern, when they hadbeen food deprived compared with when they were satiated. Theserotating vortexes may be an adaptive response to disturb thesubstrate, thereby creating new feeding patches for individuals
within the group. Different group structures can therefore arise fromslight differences in the strength of social interactions, which mayresult from external perturbations or different nutritional statesbetween individuals.
Individuals can also change the strength of their interactionsacross their development. For example, some fish form looseaggregations with limited collective movement immediately afterhatching, compared with polarised groups during their adult lives(Shaw, 1960). Groups of larger tadpoles of the clawed frog(Xenopus laevis) are more polarised than groups of smaller tadpoles(Katz et al., 1981). Adult shrimp (Paramesopodopsis rufa) tend toform denser and more polarised schools than juveniles (O’Brien,1989), and larger squid (Loligo opalescens) form more polarisedschools than smaller squid (Hurley, 1978). Coordinated behaviourcan only emerge when the sensory and motor architecture ofindividuals are sufficiently developed to allow coordinationbetween neighbours. The development of rods in the eyes of thestriped jack (Pseudocaranx dentex), for example, coincides with thedevelopment of attraction towards conspecifics (Masuda andTsukamoto, 1996). At this developmental stage, however,individuals do not form polarised groups even though they havedeveloped the manoeuvrability required to school. The transitionbetween merely aggregative behaviour to schooling behaviour,therefore, is unlikely to only be determined by the sensory andmotor capabilities of developing individuals (Masuda andTsukamoto, 1998). How individuals process and respond to theposition and orientational information of neighbours also seems tobe important. Romenskyy et al. (2015 preprint) asked whether thestrength of interactions was different between different-sized fish(Pseudomugil signifer). Groups of small fish (∼7.5 mm) formedloose aggregations with lower polarisation compared with groups ofmedium-sized (∼13 mm) or larger (∼23 mm) fish, which formedhighly polarised, compact groups. Thus, although the strength ofrepulsion was similar for different-sized fish, the attraction strengthbetween individuals was stronger for larger individuals. There isalso a suggestion that ibis (Eudocimus albus) individuals learn tocoordinate their movements with others, with birds gradually flyingmore often in V formations as they age (Petit and Bildstein, 1986;Biro et al., 2016). The interaction strengths of individuals maychange, therefore, as they grow, learn or age.
Individual variationAbove, I considered that groups show different structures andorganisation, and this is linked to the interaction rules of individuals.However, this discussion assumed that all individuals wereinteracting using identical rules. In fact, individuals within thesame group can differ in how they are interacting with one another.The best evidence for this comes from the observation that thespatial sorting of individuals within groups is non-random, withparticular individuals occupying consistent positions within groups(Pitcher et al., 1982; Partridge, 1981; Krause, 1994; Freeman et al.,2011). Spatial sorting is important to consider, because individualsthat occupy positions at the front of moving groups can ofteninfluence group movement more than others (Herbert-Read et al.,2011; Katz et al., 2011; Nagy et al., 2010). In these cases, being atthe front of the group is synonymous with leading becausedirectional changes often propagate from front to rear positions.Some of the best examples of leader–follower dynamics areobserved in homing pigeon flocks, with individuals located at thefront of these flocks having a disproportionate influence on groupdirection (Nagy et al., 2010, 2013). However, the individuals thatemerge as leaders in these flocks are not necessarily the best
2977
REVIEW Journal of Experimental Biology (2016) 219, 2971-2983 doi:10.1242/jeb.129411
Journal
ofEx
perim
entalB
iology
navigators (Flack et al., 2012; Watts et al., 2016) or more sociallydominant (Nagy et al., 2013). Instead, faster, larger birds emerge asleaders (Pettit et al., 2015, 2013; Herbert-Read, 2015). Similarresults have been found in other taxa – larger roach (Rutilus rutilus),for example, occupy positions at the front of groups (Krause et al.,1998; Krause et al., 2000). Therefore, differences in individuals’body size and self-assortment as a result of preferred speed can leadto certain individuals disproportionally influencing others’movements.In other cases, it is clear that individuals differ in their interaction
rules, as the spatial organisation of groups is not related todifferences in body size. Partridge (1981) found that the spatialsorting in shoals of saithe (Pollachius virens) was not dependent onbody size, but individuals did have preferred positions withingroups. Burns et al. (2012) also found that mosquitofish occupiedconsistent positions within groups, and this was not related to theirsex, body size or dominance. In these cases, differences in howindividuals interact with their neighbours can drive this spatialsorting. For example, we have already seen that during tandemrunning in Temnothorax ants, leaders and followers use differentinteraction rules to maintain their position in the pair (Fig. 4A,B)(Franks and Richardson, 2006). Schaerf et al. (2016 preprint) alsofound that pairs of fish differed in how they interacted with oneanother. When individuals were close together, ‘leaders’(individuals most often found at the front of the pair) had higheraccelerations when their partner was directly behind them, andlower decelerations when their partner was directly in front of them(Fig. 4C). Thus, the repulsion strength of individuals can beasymmetrical, with some individuals having stronger repulsion infront of than behind themselves. Followers (individuals most oftenfound at the back of the pair) also had higher turning speeds towardstheir partner’s position than leaders (Fig. 4D), indicating thatfollowers were more responsive to their partner’s position.Leadership roles have also been observed in other species of fish(Nakayama et al., 2012a,b). Together, these differences highlightthat individuals can differ in their likelihood of following others,producing differences in spatial sorting (Couzin et al., 2005) andleadership in a process termed ‘leadership though socialindifference’ (Conradt et al., 2009).Why might some individuals be less responsive to their
neighbour’s movements and thereby occupy positions at thefront of groups? These differences could be state dependent anddriven by motivation to feed, as non-satiated fish move faster(Hansen et al., 2015a) and have larger inter-individual distancesthan their satiated counterparts (Hansen et al., 2015b), and oftenoccupy positions at the front of groups (Krause et al., 1998).McClure et al. (2011) similarly found that hungrier caterpillarswere more likely than recently fed conspecifics to be found atthe front of groups. Alternatively, the likelihood of followinganother’s movements may relate to intrinsic differences inbehavioural phenotype, in part driven by differences in stressphysiology or metabolism (Koolhaas et al., 1999; Careau et al.,2008). The bolder of two sticklebacks (Gasterosteus aculeatus),for example, is less likely to follow their partner’s movementsand is more likely to lead the pair (Harcourt et al., 2009; Jolles et al.,2015). We shall now see that these differences can have wide-ranging implications for the ecology and evolution of collectivemotion.
Implications of having varying interaction rules in groupsGiven that natural selection acts on the variation in individuals’behaviours and phenotypes, could selection act on differences in the
way individuals interact in groups, and could this shape groupstructure and motion? This was first confirmed in computersimulations, when Wood and Ackland (2007) found that fast-moving polarised groups or slow-moving milling groups evolvedwhen particles with different interaction rules were exposed tosimulated predation pressure. However, predatory tactics are key forunderstanding which prey are targeted in groups (Morrell et al.,2015); thus, it was important to determine whether the choices ofreal predators could select for coordinated movement. Ioannou et al.(2012) achieved this by projecting videos of simulated prey onto theside of an aquarium containing a real predator – a bluegill sunfish(Lepomis macrochirus). Prey in the videos interacted with theirneighbours differently. Individuals with low attraction towardsneighbours and low alignment formed small groups with lowtortuosity. In contrast, individuals with intermediate levels ofattraction and alignment formed cohesive and coordinated groups.The sunfish preferentially targeted prey in smaller groups, andgroups that were less polarised, showing that predation can indeedselect for prey with particular interaction rules, in turn selecting forcoordinated group movement.
Varying interaction rules not only shape the general properties ofmoving groups, but can also shape how groups make decisions.Couzin et al. (2005) investigated whether a minority of informedindividuals that balanced social interactions with a desired directionof travel could lead a majority of uninformed individuals that onlyused social interactions. Only a small number of informed groupmembers were required to guide a majority of uninformedindividuals. In effect, individuals that relied less on socialinteractions could guide more socially responsive individuals;predictions from this model were later confirmed in studies ofpigeons (Biro et al., 2006), fish (Ward et al., 2008; Couzin et al.,2011) and baboons (Strandburg-Peshkin et al., 2015). Later,Ioannou et al. (2015) asked what made informed individualseffective leaders. They showed that fish with faster, straighter andless variable paths were less effective at guiding groups comparedwith fish with slower, more tortuous paths. Effective leaders,therefore, need to balance their social interaction strength with theirown goal-orientated behaviour. Indeed, individuals still have to befollowed if they are to lead (King, 2010). Therefore, by changingtheir social interaction strengths, individuals can have adisproportionate influence on group decisions.
If individuals can lead others by reducing their social interactionstrength, and travel in their own direction of preference withoutsacrificing group membership, why don’t all individuals attempt todo this? In an evolutionary framework, Guttal and Couzin (2010)proposed that there could be costs associated with ignoring socialinteractions and relying on private information. For example,private information could be costly to acquire because of theenergetic investments needed to detect environmental cues, andindividuals may neglect important information that is gatheredthrough social interactions, e.g. predator vigilance (Guttal andCouzin, 2011). Their model demonstrated that if these costs weresufficiently high, then populations evolved that consisted of someindividuals being highly socially responsive, whilst otherindividuals adopted a strategy of private information acquisition,with reduced reliance on social interactions (Guttal and Couzin,2010). Wolf and McNamara (2013) later outlined that frequency-dependent selection could also maintain the numbers of differentresponsive types in populations. If all individuals in groupscompletely ignore social information, then they cannot achievecoordination and grouping benefits are lost. Conversely, if allindividuals are extremely socially responsive, ignoring their own
2978
REVIEW Journal of Experimental Biology (2016) 219, 2971-2983 doi:10.1242/jeb.129411
Journal
ofEx
perim
entalB
iology
private information, coordination is achieved, but individualscannot exploit novel information or resources. Hence, naturalselection should favour both socially responsive and unresponsiveindividuals in populations depending on their relative frequency(Wolf et al., 2008). Whereas leaders (socially unresponsive) gainby imposing their preferences on followers (socially responsive),followers gain by only having to sample social information, andnot potentially costly private information (Webster and Laland,2008). The proportion of leaders and followers in populationsshould then be determined by the potential for conflict amonggroup members (Johnstone and Manica, 2011). Environments thatpromote conflict within groups should favour the evolution ofsocially unresponsive individuals, whereas when conflict is limitedor absent, socially responsive individuals should be selected for.Thus, frequency-dependent selection, depending on environmentalconditions, can lead to and maintain different levels of socialresponsiveness in populations (Johnstone and Manica, 2011).These models explain why it is often only some individuals thatexert their influence in group decisions, and hint at theevolutionary strategies that may surround differences inindividuals’ interaction rules.The social responsiveness of individuals is also likely to be a
plastic, context-dependent trait. This flexibility, however, maydiffer between individuals, and may be reinforced or reversed overtime. For example, leaders in homing pigeon flocks learn moreabout their environment and become better navigators thanfollowers (Pettit et al., 2015). This may reduce leaders’ relianceon using social information, exaggerating their leadership role. Inother cases, these roles can be reversed, but this depends on whichroles are adopted. Nakayama et al. (2013) found that leaders readilyadopted follower roles in pairs of sticklebacks, but the adoption ofleadership roles by followers was less flexible. Both the flexibilityand likelihood of following others, therefore, can vary betweenindividuals. This flexibility may also change over the lifetime ofindividuals. For example, 1-year-old whooping cranes (Grusamericana) have a 34% reduction in the distance they travelduring migrations if they fly with older birds (Mueller et al., 2013).Although not directly tested, this suggests that younger birds aremore responsive to the movements of older birds, and thisresponsiveness changes as the birds learn their migration route.Differences in the flexibility of interaction rules, and in particular,the likelihood of copying the movements of others, can thereforeaffect processes such as learning and leadership, and may beintricately linked to the evolution of different socially responsivetypes in populations.
Future directionsThe field of collective motion has benefited from a surge inacquisition and analysis of data from real animal groups. We shouldnow attempt to integrate this knowledge with both proximate andultimate explanations of collective motion. At the proximate end ofthe spectrum, new analytical techniques will allow us to betterunderstand the physiology and genetic basis governing differencesin individuals’ interaction rules. For example, novel behaviouralassays, including animal–robot interactions, can provide identicalsocial conditions to different individuals in different trials (Fariaet al., 2010). This will allow us to detect precise differences in howdifferent individuals interact under standardised conditions (Warket al., 2011). Artificial selection could then be used to select forindividuals with particular social interactions, for example, byscoring individuals for their degree of social responsiveness(Szorkovszky et al., 2016 preprint). This would allow us to
measure the genetic heritability of these differences (Wark et al.,2011), and by crossing individuals from different populations, thequantitative trait loci associated with different aspects of groupbehaviour could be identified (Greenwood et al., 2013, 2015;Kowalko et al., 2013). One crucial aspect of this would be tomeasure differences in levels of hormones, and the expression ofgenes responsible for hormone production, that are involved insocial behaviour. Some key candidates that could regulate socialresponsiveness are oxytocin, vasopressin and their non-mammalianhomologues. These hormones are involved in regulating socialbehaviour and social information use in animals (Insel and Young,2000; Donaldson and Young, 2008; Reddon et al., 2014), and mayprovide a proximate link to differences between individuals’interaction rules. Integrating detailed behavioural responses withgenetic, neurological and physiological measures is now key, notonly to understanding how individuals interact in moving groups,but also in other areas of evolutionary and behavioural biology(Hofmann et al., 2014).
We should endeavour to understand how animals integrateinformation on the movements and positions of neighbours in threedimensions. This will require a better understanding of the sensoryinputs received from conspecifics (Strandburg-Peshkin et al., 2013;Rosenthal et al., 2015). Tracking systems can now monitor the 3Dpositions and body orientations of animals in real time (Straw et al.,2011), giving us the opportunity to immerse animals in virtual-reality environments (Fry et al., 2008) with virtual conspecifics.These systems will allow us to control the sensory inputs receivedfrom virtual neighbours, revealing how animals perceive andrespond to each other’s movements in groups. In these systems, weshould also endeavour to detect the precise instances when ananimal decides to update its position on the move. Approaches todate have usually determined the interaction rules of individualsbetween every frame at which their positions were recorded;however, in many systems, animals are more likely to make discrete,intermittent decisions to move (Kramer and McLaughlin, 2001).Identifying these decisions will lead to a deeper understanding ofthe neurological and perceptual processes that underlie groupmovement.
These data should then be used to inform more data-drivenmodels of collective motion (Calovi et al., 2014), which may requirea different approach to the models that have been previouslyproposed. For example, the field may benefit from building modelsin a perceptual control-theory framework, instead of a rule-basedapproach. This framework explains how complex behaviour (suchas group movement) can be generated when individuals attempt tostabilise different aspects of their perception, instead of respondingto conspecifics using certain rules (Bell and Pellis, 2011). Weshould also seek new analytical techniques to assess theeffectiveness of model fitting (Mann et al., 2014, 2013; Herbert-Read et al., 2015b). These models are important, because theyinspire us to build more reliable robotic systems (Sahin, 2004) anddevelop new algorithms to solve complex collective motionproblems, such as how to effectively herd groups of agents(Strömbom et al., 2014).
Much of the data collected on moving groups to date have beengathered under laboratory conditions. However, advances inminiature GPS and sonar will now also allow us to gather high-resolution temporal and spatial data from animals in their naturalhabitats. Combining movement data with knowledge of habitatstructure, resource sites and environmental conditions will allow usto examine the importance of social and environmental cues thatallow animal groups to navigate through their environment. This
2979
REVIEW Journal of Experimental Biology (2016) 219, 2971-2983 doi:10.1242/jeb.129411
Journal
ofEx
perim
entalB
iology
will lead to a deeper understanding of the movement ecology ofanimal groups.In addition to the proximate causes of differences between
individuals’ interaction rules, the ultimate consequences of thesedifferences should be assessed. One fundamental question is howthese interaction rules evolved in natural populations. Indeed, thesimilarities in interaction rules between distantly related species offishes suggests that some rules governing collective motion havebeen conserved over time, or that there are constraints for howindividuals can interact in groups. Comparing the interaction rulesof different species in a phylogenetic framework could prove usefulfor understanding the evolution and maintenance of particularinteraction rules. This could also be investigated by comparingdifferences in the interaction rules between populations of the samespecies that have been exposed to different selection pressure overtheir evolutionary history. Different populations of Trinidadianguppies, for example, have been subject to varying degrees ofpredation pressure (Magurran, 2005). Fish from high-predationenvironments form more cohesive groups than fish from low-predation environments (Seghers, 1974). The interaction rulesunderlying these differences have yet to be determined, and thisnatural experiment may provide important insights into howpredation can select for different interaction rules in groups.We should ask how variation in the interaction rules of
individuals can be maintained in populations, informed andinspired by the predictions of evolutionary models (Wolf andMcNamara, 2013; Johnstone and Manica, 2011; Guttal and Couzin,2010). This will require testing the range of social interactionswithin and between individuals in natural populations. Importantly,this will need to consider the costs and benefits of different socialinteractions in different social and environmental contexts. Here, wemight benefit from thinking about the interaction rules ofindividuals in a strategic framework (Laland, 2004). For example,we could consider different interaction rules as strategies thatattempt to maximise individuals’ survival and reproductive successin different contexts. A functional approach to understanding whichinteraction rules are adopted in groups would greatly improve ourunderstanding of a field that has, to date, been mechanisticallyfocused. Nevertheless, the analytical techniques outlined here nowallow these ideas to be tested.
ConclusionsIn this Review, I have highlighted how animal groups can achievecoordinated motion, and discussed how the interaction rules withinand between species, groups and individuals differ. I have alsodiscussed the larger-scale evolutionary and ecologicalconsiderations surrounding these differences. With constantimprovements in tracking methods and analytical techniques, thisarea of research will continue to provide considerable insight intoother fields. Because we can now measure detailed movement andbehavioural responses of individuals in groups, we can combine thisknowledge with fields including physiology, neuroscience andgenetics to give a more comprehensive understanding ofindividuals’ social behaviour. This integration is unprecedented,and will no doubt lead to exciting breakthroughs in years to come.
AcknowledgementsI thank David Sumpter, Maksym Romenskyy, Ashley Ward and two anonymousreferees for helpful comments on this Review. I also thank Herbert Krause for hand-drawn sketches of the fish in Fig. 4.
Competing interestsThe author declares no competing or financial interests.
FundingThis work was supported by a Knut and AliceWallenberg Foundation (Knut och AliceWallenbergs Stiftelse) grant no. 2013.0072 awarded to David J. T. Sumpter.Deposited in PMC for immediate release.
ReferencesAoki, I. (1982). A simulation study on the schooling mechanism in fish. Bull. Jpn.
Soc. Sci. Fish. 48, 1081-1088.Attanasi, A., Cavagna, A., Del Castello, L., Giardina, I., Grigera, T. S., Jelic, A.,
Melillo, S., Parisi, L., Pohl, O., Shen, E. et al. (2014a). Information transfer andbehavioural inertia in starling flocks. Nat. Phys. 10, 691-696.
Attanasi, A., Cavagna, A., Del Castello, L., Giardina, I., Melillo, S., Parisi, L.,Pohl, O., Rossaro, B., Shen, E., Silvestri, E. et al. (2014b). Collective behaviourwithout collective order in wild swarms of midges. PLoS Comput. Biol. 10, p.e1003697.
Ballerini, M., Cabibbo, N., Candelier, R., Cavagna, A., Cisbani, E., Giardina, I.,Lecomte, V., Orlandi, A., Parisi, G., Procaccini, A. et al. (2008). Interactionruling animal collective behavior depends on topological rather than metricdistance: evidence from a field study. Proc. Natl. Acad. Sci. USA 105, 1232-1237.
Bazazi, S., Buhl, J., Hale, J. J., Anstey, M. L., Sword, G. A., Simpson, S. J. andCouzin, I. D. (2008). Collectivemotion and cannibalism in locust migratory bands.Curr. Biol. 18, 735-739.
Bazazi, S., Romanczuk, P., Thomas, S., Schimansky-Geier, L., Hale, J. J.,Miller, G. A., Sword, G. A., Simpson, S. J. and Couzin, I. D. (2011). Nutritionalstate and collective motion: from individuals to mass migration. Proc. R. Soc. BBiol. Sci. 278, 356-363.
Bazazi, S., Pfennig, K. S., Handegard, N. O. and Couzin, I. D. (2012). Vortexformation and foraging in polyphenic spadefoot toad tadpoles. Behav. Ecol.Sociobiol. 66, 879-889.
Beauchamp, G. (2012). Flock size and density influence speed of escape waves insemipalmated sandpipers. Anim. Behav. 83, 1125-1129.
Beekman, M., Fathke, R. L. and Seeley, T. D. (2006). How does an informedminority of scouts guide a honeybee swarm as it flies to its new home? Anim.Behav. 71, 161-171.
Bell, H. C. and Pellis, S. M. (2011). A cybernetic perspective on food protection inrats: simple rules can generate complex and adaptable behaviour. Anim. Behav.82, 659-666.
Bialek, W., Cavagna, A., Giardina, I., Mora, T., Pohl, O., Silvestri, E., Viale, M.andWalczak, A. M. (2014). Social interactions dominate speed control in poisingnatural flocks near criticality. Proc. Natl. Acad. Sci. USA 111, 7212-7217.
Biro, D., Sumpter, D. J. T., Meade, J. and Guilford, T. (2006). From compromise toleadership in pigeon homing. Curr. Biol. 16, 2123-2128.
Biro, D., Sasaki, T. and Portugal, S. J. (2016). Bringing a time–depth perspective tocollective animal behaviour. Trends. Ecol. Evol. 31, 550-562.
Bode, N. W. F., Franks, D. W. and Wood, A. J. (2010). Making noise: emergentstochasticity in collective motion. J. Theor. Biol. 267, 292-299.
Boeddeker, N., Kern, R. and Egelhaaf, M. (2003). Chasing a dummy target:smooth pursuit and velocity control in male blowflies. Proc. R. Soc. B Biol. Sci.270, 393-399.
Branson, K., Robie, A. A., Bender, J., Perona, P. and Dickinson, M. H. (2009).High-throughput ethomics in large groups of Drosophila. Nat. Methods 6,451-457.
Breder, C. (1954). Equations descriptive of fish schools and other animalaggregations. Ecology 35, 361-370.
Buhl, J., Sumpter, D. J. T., Couzin, I. D., Hale, J. J., Despland, E., Miller, E. R. andSimpson, S. J. (2006). From disorder to order in marching locusts. Science 312,1402-1406.
Buhl, J., Sword, G. A. and Simpson, S. J. (2012). Using field data to test locustmigratory band collective movement models. Interface Foc. 2, 757-764.
Burns, A., Herbert-Read, J., Morrell, L. andWard, A. J. W. (2012). Consistency ofleadership in shoals of mosquitofish (Gambusia holbrooki) in novel and in familiarenvironments. PLoS ONE 7, e36567.
Calovi, D. S., Lopez, U., Ngo, S., Sire, C., Chate, H. and Theraulaz, G. (2014).Swarming, schooling, milling: phase diagram of a data-driven fish school model.New J. Phys. 16, 015026.
Careau, V., Thomas, D., Humphries, M. M. and Reale, D. (2008). Energymetabolism and animal personality. Oikos 117, 641-653.
Cavagna, A., Giardina, I., Orlandi, A., Parisi, G., Procaccini, A., Viale, M. andZdravkovic, V. (2008a). The starflag handbook on collective animal behaviour:1. Empirical methods. Anim. Behav. 76, 217-236.
Cavagna, A., Giardina, I., Orlandi, A., Parisi, G. and Procaccini, A. (2008b). Thestarflag handbook on collective animal behaviour: 2. three-dimensional analysis.Anim. Behav. 76, 237-248.
Cavagna, A., Cimarelli, A., Giardina, I., Orlandi, A., Parisi, G., Procaccini, A.,Santagati, R. and Stefanini, F. (2008c). New statistical tools for analyzing thestructure of animal groups. Math. Biosci. 214, 32-37.
Cavagna, A., Del Castello, L., Giardina, I., Grigera, T., Jelic, A., Melillo, S., Mora,T., Parisi, L., Silvestri, E., Viale, M. et al. (2015). Flocking and turning: a newmodel for self-organized collective motion. J. Stat. Phys. 158, 601-627.
2980
REVIEW Journal of Experimental Biology (2016) 219, 2971-2983 doi:10.1242/jeb.129411
Journal
ofEx
perim
entalB
iology
Codling, E. A., Pitchford, J. W. and Simpson, S. D. (2007). Group navigation andthe “many-wrongs principle” in models of animal movement. Ecology 88,1864-1870.
Conradt, L., Krause, J., Couzin, I. D. and Roper, T. (2009). “Leading according toneed” in self-organizing groups. Am. Nat. 173, 304-312.
Couzin, I. D., Krause, J., James, R., Ruxton, G. D. and Franks, N. R. (2002).Collective memory and spatial sorting in animal groups. J. Theor. Biol. 218, 1-11.
Couzin, I. D. and Franks, N. R. (2003). Self-organized lane formation and optimizedtraffic flow in army ants. Proc. R. Soc. B Biol. Sci. 270, 139-146.
Couzin, I. D., Krause, J., Franks, N. R. and Levin, S. A. (2005). Effectiveleadership and decision-making in animal groups on the move. Nature 433,513-516.
Couzin, I. D., Ioannou, C. C., Demirel, G., Gross, T., Torney, C. J., Hartnett, A.,Conradt, L., Levin, S. A. and Leonard, N. E. (2011). Uninformed individualspromote democratic consensus in animal groups. Science 334, 1578-1580.
Dill, L. M., Holling, C. S. and Palmer, L. H. (1997). Predicting the 3-dimensionalstructure of animal aggregations from functional considerations: the role ofinformation. In Animal Groups in Three Dimensions (ed. J. K. Parrish and W. M.Hammer), pp. 207-224. Cambridge: Cambridge University Press.
Donaldson, Z. R. and Young, L. J. (2008). Oxytocin, vasopressin, and theneurogenetics of sociality. Science 322, 900-904.
Faria, J. J., Dyer, J. R., Clement, R. O., Couzin, I. D., Holt, N., Ward, A. J. W.,Waters, D. and Krause, J. (2010). A novel method for investigating the collectivebehaviour of fish: introducing ‘Robofish’. Behav. Ecol. Sociobiol. 64, 1211-1218.
Fitzgerald, T. D. and Pescador-Rubio, A. (2002). The role of tactile and chemicalstimuli in the formation and maintenance of the processions of the socialcaterpillar Hylesia lineata (Lepidoptera: Saturniidae). J. Insect Behav. 15,659-674.
Flack, A., Pettit, B., Freeman, R., Guilford, T. and Biro, D. (2012). What areleaders made of? The role of individual experience in determining leader–followerrelations in homing pigeons. Anim. Behav. 83, 703-709.
Franklin, E. L., Richardson, T. O., Sendova-Franks, A. B., Robinson, E. J. H. andFranks, N. R. (2011). Blinkered teaching: tandem running by visually impairedants. Behav. Ecol. Sociobiol. 65, 569-579.
Franks, N. R. and Richardson, T. (2006). Teaching in tandem-running ants.Nature439, 153-153.
Franks, N. R., Pratt, S. C., Mallon, E. B., Britton, N. F. and Sumpter, D. J. T.(2002). Information flow, opinion polling and collective intelligence in house-hunting social insects. Philos. Trans. R. Soc. B 357, 1567-1583.
Franks, N. R., Worley, A., Grant, K. A. J., Gorman, A. R., Vizard, V., Plackett, H.,Doran, C., Gamble, M. L., Stumpe, M. C. and Sendova-Franks, A. B. (2016).Social behaviour and collective motion in plant–animal worms. Proc. R. Soc. BBiol. Sci. 283, 20152946.
Freeman, R., Mann, R., Guilford, T. and Biro, D. (2011). Group decisions andindividual differences: route fidelity predicts flight leadership in homing pigeons(Columba livia). Biol. Lett. 7, 63-66.
Fry, S. N., Rohrseitz, N., Straw, A. D. and Dickinson, M. H. (2008). TrackFly:virtual reality for a behavioral system analysis in free-flying fruit flies. J. Neurosci.Methods 171, 110-117.
Gautrais, J., Ginelli, F., Fournier, R., Blanco, S., Soria, M., Chate, H. andTheraulaz, G. (2012). Deciphering interactions in moving animal groups. PLoSComput. Biol. 8, e1002678.
Greenwood, A. K., Wark, A. R., Yoshida, K. and Peichel, C. L. (2013). Geneticand neural modularity underlie the evolution of schooling behavior in threespinesticklebacks. Curr. Biol. 23, 1884-1888.
Greenwood, A. K., Ardekani, R., McCann, S. R., Dubin, M. E., Sullivan, A.,Bensussen, S., Tavar ´e, S. and Peichel, C. L. (2015). Genetic mapping ofnatural variation in schooling tendency in the threespine stickleback. G3 5,761-769.
Grover, D., Katsuki, T. and Greenspan, R. J. (2016). Flyception: imaging brainactivity in freely walking fruit flies. Nat. Methods 13, 569-572.
Grunbaum, D. (1998). Schooling as a strategy for taxis in a noisy environment.Evol.Ecol. 12, 503-522.
Grunbaum, D., Viscido, S. and Parrish, J. K. (2005). Extracting interactive controlalgorithms from group dynamics of schooling fish. In Cooperative Control (ed. V.Kumar, N. Leonard and S. A. Morse), pp. 103-117. Berlin: Springer.
Guttal, V. and Couzin, I. D. (2010). Social interactions, information use, and theevolution of collective migration. Proc. Natl. Acad. Sci. USA 107, 16172-16177.
Guttal, V. and Couzin, I. D. (2011). Leadership, collective motion and the evolutionof migratory strategies. Comm. Int. Biol. 4, 294-298.
Handegard, N. O. and Williams, K. (2008). Automated tracking of fish in trawlsusing the DIDSON (dual frequency identification sonar). ICES J. Mar. Sci. 65,636-644.
Hansen, M. J., Buhl, J., Bazazi, S., Simpson, S. J. and Sword, G. A. (2011).Cannibalism in the lifeboat—collective movement in Australian plague locusts.Behav. Ecol. Sociobiol. 65, 1715-1720.
Hansen, M. J., Schaerf, T. M. and Ward, A. J. (2015a). The effect of hunger on theexploratory behaviour of shoals of mosquitofish Gambusia holbrooki. Behaviour152, 1659-1677.
Hansen, M. J., Schaerf, T. M. and Ward, A. J. (2015b). The influence of nutritionalstate on individual and group movement behaviour in shoals of crimson-spottedrainbowfish (Melanotaenia duboulayi). Behav. Ecol. Sociobiol. 69, 1713-1722.
Harcourt, J. L., Ang, T. Z., Sweetman, G., Johnstone, R. A. and Manica, A.(2009). Social feedback and the emergence of leaders and followers. Curr. Biol.19, 248-252.
Hemelrijk, C. K. and Hildenbrandt, H. (2008). Self-organized shape and frontaldensity of fish schools. Ethology 114, 245-254.
Hemelrijk, C. K. and Hildenbrandt, H. (2012). Schools of fish and flocks of birds:their shape and internal structure by self-organization. Interface Focus 6, 726-737.
Hemelrijk, C. K. and Kunz, H. (2005). Density distribution and size sorting in fishschools: an individual-based model. Behav. Ecol. 16, 178-187.
Hemelrijk, C. K., Reid, D. A. P., Hildenbrandt, H. and Padding, J. T. (2015a). Theincreased efficiency of fish swimming in a school. Fish Fisheries 16, 511-521.
Hemelrijk, C. K., van Zuidam, L. and Hildenbrandt, H. (2015b). What underlieswaves of agitation in starling flocks. Behav. Ecol. Sociobiol. 69, 755-764.
Herbert-Read, J. E. (2015). Collective behaviour: leadership and learning in flocks.Curr. Biol. 25, R1127-R1129.
Herbert-Read, J., Perna, A., Mann, R., Schaerf, T., Sumpter, D. J. T. and Ward,A. J. W. (2011). Inferring the rules of interaction of shoaling fish. Proc. Natl. Acad.Sci. USA 108, 18726-18731.
Herbert-Read, J. E., Buhl, J., Hu, F., Ward, A. J. W. and Sumpter, D. J. T. (2015a).Initiation and spread of escape waves within animal groups. R. Soc. Open Sci. 2,140355.
Herbert-Read, J., Romenskyy, M. and Sumpter, D. J. T. (2015b). A turing test forcollective motion. Biol. Lett. 11, 20150674.
Hildenbrandt, H., Carere, C. and Hemelrijk, C. K. (2010). Self-organized aerialdisplays of thousands of starlings: a model. Behav. Ecol. 21, 1349-1359.
Hofmann, H. A., Beery, A. K., Blumstein, D. T., Couzin, I. D., Earley, R. L., Hayes,L. D., Hurd, P. L., Lacey, E. A., Phelps, S. M., Solomon, N. G. et al. (2014). Anevolutionary framework for studying mechanisms of social behavior. Trends Ecol.Evol. 29, 581-589.
Hurley, A. C. (1978). School structure of squid Loligo opalescens. Fish. Bull. 76,433-442.
Huth, A. and Wissel, C. (1992). The simulation of the movement of fish schools.J. Theor. Biol. 156, 365-385.
Insel, T. R. and Young, L. J. (2000). Neuropeptides and the evolution of socialbehavior. Curr. Opin. Neurobiol. 10, 784-789.
Ioannou, C. C., Guttal, V. and Couzin, I. D. (2012). Predatory fish select forcoordinated collective motion in virtual prey. Science 337, 1212-1215.
Ioannou, C. C., Singh, M. and Couzin, I. D. (2015). Potential leaders trade off goal-oriented and socially oriented behavior in mobile animal groups. Am. Nat. 186,284-293.
Johnstone, R. A. and Manica, A. (2011). Evolution of personality differences inleadership. Proc. Natl. Acad. Sci. USA 108, 8373-8378.
Jolles, J. W., Fleetwood-Wilson, A., Nakayama, S., Stumpe, M. C., Johnstone,R. A. andManica, A. (2015). The role of social attraction and its link with boldnessin the collective movements of three-spined sticklebacks. Anim. Behav. 99,147-153.
Katz, L. C., Potel, M. J. andWassersug, R. J. (1981). Structure andmechanisms ofschooling intadpoles of the clawed frog, Xenopus laevis. Anim. Behav. 29, 20-33.
Katz, Y., Tunstrom, K., Ioannou, C. C., Huepe, C. and Couzin, I. D. (2011).Inferring the structure and dynamics of interactions in schooling fish. Proc. Natl.Acad. Sci. USA 108, 18720-18725.
Killen, S. S., Marras, S., Steffensen, J. F. and McKenzie, D. J. (2011). Aerobiccapacity influences the spatial position of individuals within fish schools.Proc. R. Soc. B Biol. Sci. 279, 357-364.
King, A. (2010). Follow me! I’m a leader if you do; I’m a failed initiator if you don’t?Behav. Proc. 84, 671-674.
Koolhaas, J. M., Korte, S. M., De Boer, S. F., Van Der Vegt, B. J., Van Reenen,C. G., Hopster, H., De Jong, I. C., Ruis, M. A. W. and Blokhuis, H. J. (1999).Coping styles in animals: current status in behavior and stress-physiology.Neurosci. Biobehav. Rev. 23, 925-935.
Kowalko, J. E., Rohner, N., Rompani, S. B., Peterson, B. K., Linden, T. A.,Yoshizawa, M., Kay, E. H., Weber, J., Hoekstra, H. E., Jeffery, W. R. et al.(2013). Loss of schooling behavior in cavefish through sight-dependent and sight-independent mechanisms. Curr. Biol. 23, 1874-1883.
Kramer, D. L. and McLaughlin, R. L. (2001). The behavioral ecology of intermittentlocomotion. Am. Zool. 41, 137-153.
Krause, J. (1994). Differential fitness returns in relation to spatial position in groups.Biol. Rev. Cam. Philos. Soc. 69, 187-206.
Krause, J. and Ruxton, G. D. (2002). Living in Groups. Oxford: Oxford UniversityPress.
Krause, J., Reeves, P. andHoare, D. (1998). Positioning behaviour in roach shoals:the role of body length and nutritional state. Behaviour 135, 1031-1039.
Krause, J., Hoare, D., Krause, S., Hemelrijk, C. K. and Rubenstein, D. I. (2000).Leadership in fish shoals. Fish Fisheries 1, 82-89.
Kunz, H. and Hemelrijk, C. K. (2003). Artificial fish schools: collective effects ofschool size, body size, and body form. Art. Life 9, 237-253.
Laland, K. N. (2004). Social learning strategies. Anim. Learn. Behav. 32, 4-14.
2981
REVIEW Journal of Experimental Biology (2016) 219, 2971-2983 doi:10.1242/jeb.129411
Journal
ofEx
perim
entalB
iology
Latty, T., Duncan, M. and Beekman, M. (2009). High bee traffic disrupts transfer ofdirectional information in flying honeybee swarms. Anim. Behav. 78, 117-121.
Lukeman, R., Li, Y.-X. and Edelstein-Keshet, L. (2010). Inferring individual rulesfrom collective behavior. Proc. Natl. Acad. Sci. USA 107, 12576.
Magurran, A. E. (2005). Evolutionary Ecology: The Trinidadian Guppy. Oxford:Oxford University Press.
Major, P. F. and Dill, L. M. (1978). The three-dimensional structure of airborne birdflocks. Behav. Ecol. Sociobiol. 4, 111-122.
Mann, R. P., Perna, A., Strombom,, D., Garnett, R., Herbert-Read, J. E.,Sumpter, D. J. T. and Ward, A. J. W. (2013). Multi-scale inference of interactionrules in animal groups using Bayesian model selection. PLoS Comput. Biol. 9,e1002961.
Mann, R. P., Herbert-Read, J. E., Ma, Q., Jordan, L. A., Sumpter, D. J. T. andWard, A. J. W. (2014). A model comparison reveals dynamic social informationdrives the movements of humbug damselfish (Dascyllus aruanus). J. R. Soc.Interface 11, 20130794.
Marras, S., Killen, S. S., Lindstrom, J., McKenzie, D. J., Steffensen, J. F. andDomenici, P. (2015). Fish swimming in schools save energy regardless of theirspatial position. Behav. Ecol. Sociobiol. 69, 219-226.
Martin, G. R. (1986). The eye of a passeriform bird, the European starling (Sturnusvulgaris): eye movement amplitude, visual fields and schematic optics. J. Comp.Physiol. A 159, 545-557.
Masuda, R. and Tsukamoto, K. (1996). Morphological development in relation tophototaxis and rheotaxis in the striped jack, Pseudocaranx dentex. Mar. Freshw.Behav. Physiol. 28, 75-90.
Masuda, R. and Tsukamoto, K. (1998). The ontogeny of schooling behaviour in thestriped jack. J. Fish Biol. 52, 483-493.
Mather, J. A. and O’Dor, R. K. (1984). Spatial organization of schools of the squidIllex illecebrosus. Mar. Freshwater Behav. Physiol. 10, 259-271.
McClure, M., Ralph, M. and Despland, E. (2011). Group leadership depends onenergetic state in a nomadic collective foraging caterpillar. Behav. Ecol. Sociobiol.65, 1573-1579.
Morales, J. M., Moorcroft, P. R., Matthiopoulos, J., Frair, J. L., Kie, J. G., Powell,R. A., Merrill, E. H. andHaydon, D. T. (2010). Building the bridge between animalmovement and population dynamics. Philos. Trans. R. Soc. B 365, 2289-2301.
Morrell, L. J., Greenwood, L. and Ruxton, G. D. (2015). Consequences ofvariation in predator attack for the evolution of the selfish herd. Evol. Ecol. 29,107-121.
Mueller, T., O’Hara, R. B., Converse, S. J., Urbanek, R. P. and Fagan, W. F.(2013). Social learning of migratory performance. Science 341, 999-1002.
Nagy, M., Ákos, Z., Biro, D. and Vicsek, T. (2010). Hierarchical group dynamics inpigeon flocks. Nature 464, 890-893.
Nagy, M., Vasarhelyi, G., Pettit, B., Roberts-Mariani, I., Vicsek, T. and Biro, D.(2013). Context-dependent hierarchies in pigeons. Proc. Natl. Acad. Sci. USA110, 13049-13054.
Nakayama, S., Harcourt, J. L., Johnstone, R. A. and Manica, A. (2012a).Initiative, personality and leadership in pairs of foraging fish. PLoS ONE 7, p.e36606.
Nakayama, S., Johnstone, R. A. and Manica, A. (2012b). Temperament andhunger interact to determine the emergence of leaders in pairs of foraging fish.PLoS ONE 7, p.e43747.
Nakayama, S., Stumpe, M. C., Manica, A. and Johnstone, R. A. (2013).Experience overrides personality differences in the tendency to follow but not inthe tendency to lead. Proc. R. Soc. B Biol. Sci. 280, 20131724.
Niwa, H.-S. (1994). Self-organizing dynamic model of fish schooling. J. Theor. Biol.171, 123-136.
O’Brien, D. P. (1989). Analysis of the internal arrangement of individuals withincrustacean aggregations (Euphausiacea, Mysidacea). J. Exp. Mar. Biol. Ecol.128, 1-30.
Okubo, A. and Chiang, H. C. (1974). An analysis of the kinematics of swarming ofAnarete pritchardi Kim (Diptera: Cecidomyiidae). Res. Pop. Ecol. 16, 1-42.
Parrish, J. K. and Turchin, P. (1997). Individual decisions, traffic rules, andemergent pattern in schooling fish. In Animal Groups in Three Dimensions (ed.J. K. Parrish and W. M. Hamner), pp. 126-142. Cambridge, UK: CambridgeUniversity Press.
Partridge, B. L. (1980). The effect of school size on the structure and dynamics ofminnow schools. Anim. Behav. 28, 68-77, IN3.
Partridge, B. L. (1981). Internal dynamics and the interrelations of fish in schools.J. Comp. Physiol. 144, 313-325.
Partridge, B. L. and Pitcher, T. J. (1980). The sensory basis of fish schools: relativeroles of lateral line and vision. J. Comp. Physiol. 135, 315-325.
Partridge, B. L., Pitcher, T., Cullen, J. M. and Wilson, J. (1980). The three-dimensional structure of fish schools. Behav. Ecol. Sociobiol. 6, 277-288.
Partridge, B. L., Johansson, J. and Kalish, J. (1983). The structure of schools ofgiant bluefin tuna in Cape Cod Bay. Envir. Biol. Fishes 9, 253-262.
Perez-Escudero, A., Vicente-Page, J., Hinz, R. C., Arganda, S. and de Polavieja,G. G. (2014). idtracker: tracking individuals in a group by automatic identification ofunmarked animals. Nat. Methods 11, 743-748.
Perna, A., Gregoire, G. and Mann, R. P. (2014). On the duality between interactionresponses and mutual positions in flocking and schooling. Mov. Ecol. 2, 1-11.
Petit, D. R. and Bildstein, K. L. (1986). Development of formation flying in juvenilewhite ibises (Eudocimus albus). Auk 103, 244-246.
Pettit, B., Perna, A., Biro, D. and Sumpter, D. J. T. (2013). Interaction rulesunderlying group decisions in homing pigeons. J. R. Soc. Interface 10, 20130529.
Pettit, B., Ákos, Z., Vicsek, T. and Biro, D. (2015). Speed determines leadershipand leadership determines learning during pigeon flocking. Curr. Biol. 25,3132-3137.
Pitcher, T. J., Partridge, B. L. and Wardle, C. S. (1976). A blind fish can school.Science 194, 963-965.
Pitcher, T. J., Wyche, C. J. and Magurran, A. E. (1982). Evidence for positionpreferences in schooling mackerel. Anim. Behav. 30, 932-934.
Portugal, S. J., Hubel, T. Y., Fritz, J., Heese, S., Trobe, D., Voelkl, B., Hailes, S.,Wilson, A. M. andUsherwood, J. R. (2014). Upwash exploitation and downwashavoidance by flap phasing in ibis formation flight. Nature 505, 399-402.
Pratt, S. C. (2008). Efficiency and regulation of recruitment during colony emigrationby the ant Temnothorax curvispinosus. Behav. Ecol. Sociobiol. 62, 1369-1376.
Procaccini, A., Orlandi, A., Cavagna, A., Giardina, I., Zoratto, F., Santucci, D.,Chiarotti, F., Hemelrijk, C. K., Alleva, E., Parisi, G. et al. (2011). Propagatingwaves in starling, Sturnus vulgaris, flocks under predation. Anim. Behav. 82,759-765.
Radakov, D. V. (1973). Schooling in the Ecology of Fish. New York: John Wiley.Reynolds, C. W. (1987). Flocks, herds and schools: a distributed behavioral model.
Comput. Graph. 21, 25-34.Reddon, A. R., Voisin, M. R., O’Connor, C. M. and Balshine, S. (2014). Isotocin
and sociality in the cooperatively breeding cichlid fish, Neolamprologus pulcher.Behaviour 151, 1389-1411.
Rogers, S. M., Harston, G. W. J., Kilburn-Toppin, F., Matheson, T., Burrows, M.,Gabbiani, F. andKrapp, H. G. (2010). Spatiotemporal receptive field properties ofa looming-sensitive neuron in solitarious and gregarious phases of the desertlocust. J. Neurophys. 103, 779-792.
Romanczuk, P. and Schimansky-Geier, L. (2012). Swarming and patternformation due to selective attraction and repulsion. Inter. Focus. 2, 746-756.
Romenskyy, M., Herbert-Read, J. E., Ward, A. J. and Sumpter, D. J. T. (2015).The statistical mechanics of schooling fish captures their interactions. arXivpreprint arXiv:1508.07708.
Romey, W. L. (1996). Individual differences make a difference in the trajectories ofsimulated schools of fish. Ecol. Model. 92, 65-77.
Romey, W. L. and Vidal, J. M. (2013). Sum of heterogeneous blind zones predictmovements of simulated groups. Ecol. Model. 258, 9-15.
Romey, W. L., Miller, M. M. and Vidal, J. M. (2014). Collision avoidance duringgroup evasive manoeuvres: a comparison of real versus simulated swarms withmanipulated vision and surface wave detectors. Proc. R. Soc. B Biol. Sci. 281,20140812.
Rosenthal, S. B., Twomey, C. R., Hartnett, A. T., Wu, H. S. and Couzin, I. D.(2015). Revealing the hidden networks of interaction in mobile animal groupsallows prediction of complex behavioral contagion. Proc. Natl. Acad. Sci. USA112, 4690-4695.
Sahin, E. (2004). Swarm robotics: From sources of inspiration to domains ofapplication. In InternationalWorkshop on SwarmRobotics (ed. E. Sahin andW.M.Spears), pp. 10-20. Berlin: Springer.
Schaerf, T., Herbert-Read, J. E., Myerscough, M. R., Sumpter, D. J. T. andWard,A. J. W. (2016). Identifying differences in the rules of interaction betweenindividuals in moving animal groups. arXiv preprint arXiv:1601.08202.
Schultz, K. M., Passino, K. M. and Seeley, T. D. (2008). The mechanism of flightguidance in honeybee swarms: subtle guides or streaker bees? J. Exp. Biol. 211,3287-3295.
Seeley, T. D. and Buhrman, S. C. (1999). Group decision making in swarms ofhoney bees. Behav. Ecol. Sociobiol. 45, 19-31.
Seghers, B. H. (1974). Schooling behavior in the guppy (Poecilia reticulata): anevolutionary response to predation. Evolution 28, 486-489.
Selous, E. (1931). Thought Transference (or what?) in Birds. London: Constableand Company Ltd.
Shaw, E. (1960). The development of schooling behavior in fishes.Physiol. Zool. 33,79-86.
Simpson, S. J., Sword, G. A., Lorch, P. D. and Couzin, I. D. (2006). Cannibalcrickets on a forced march for protein and salt. Proc. Natl. Acad. Sci. USA 103,4152-4156.
Strandburg-Peshkin, A., Twomey, C. R., Bode, N. W. F., Kao, A. B., Katz, Y.,Ioannou, C. C., Rosenthal, S. B., Torney, C. J., Wu, H. S., Levin, S. A. et al.(2013). Visual sensory networks and effective information transfer in animalgroups. Curr. Biol. 23, R709-R711.
Strandburg-Peshkin, A., Farine, D. R., Couzin, I. D. and Crofoot, M. C. (2015).Shared decision-making drives collective movement in wild baboons. Science348, 1358-1361.
Straw, A. D., Branson, K., Neumann, T. R. and Dickinson, M. H. (2011). Multi-camera real-time three-dimensional tracking of multiple flying animals. J. R. Soc.Interface 8, 395-409.
Strombom, D. (2011). Collective motion from local attraction. J. Theor. Biol. 283,145-151.
2982
REVIEW Journal of Experimental Biology (2016) 219, 2971-2983 doi:10.1242/jeb.129411
Journal
ofEx
perim
entalB
iology
Strombom, D., Mann, R. P., Wilson, A. M., Hailes, S., Morton, A. J., Sumpter,D. J. T. and King, A. J. (2014). Solving the shepherding problem: heuristics forherding autonomous, interacting agents. J. R. Soc. Interface 11, 20140719.
Sumpter, D. J. T. (2010). Collective Animal Behavior. Princeton, NJ: PrincetonUniversity Press.
Sumpter, D., Buhl, J., Biro, D. and Couzin, I. D. (2008). Information transfer inmoving animal groups. Theor. Biosci. 127, 177-186.
Szorkovszky, A., Kotrschal, A., Sumpter, D. J., Kolm, N. and Pelckmans, K.(2016). An efficient method for sorting and selecting for social behaviour. arXivpreprint arXiv:1602.05833.
Treherne, J. E. and Foster, W. A. (1981). Group transmission of predator avoidancebehaviour in a marine insect: the Trafalgar effect. Anim. Behav. 29, 911-917.
Tunstrøm, K., Katz, Y., Ioannou, C. C., Huepe, C., Lutz, M. J. and Couzin, I. D.(2013). Collective states, multistability and transitional behavior in schooling fish.PLoS Comput. Biol. 9, e1002915.
Usherwood, J. R., Stavrou, M., Lowe, J. C., Roskilly, K. andWilson, A. M. (2011).Flying in a flock comes at a cost in pigeons. Nature 474, 494-497.
Vicsek, T. and Zafeiris, A. (2012). Collective motion. Phys. Rep. 517, 71-140.Vicsek, T., Cziro k, A., Ben-Jacob, E., Cohen, I. and Shochet, O. (1995). Noveltype of phase transition in a system of self-driven particles. Phys. Rev. Lett. 75,1226.
Viscido, S. V., Parrish, J. K. and Grunbaum, D. (2004). Individual behavior andemergent properties of fish schools: a comparison of observation and theory.Mar.Ecol. Prog. Ser. 273, 239-249.
Ward, A. J. W., Sumpter, D. J. T., Couzin, I. D., Hart, P. J. B. and Krause, J.(2008). Quorum decision-making facilitates information transfer in fish shoals.Proc. Natl. Acad. Sci. USA 105, 6948-6953.
Wark, A. R., Greenwood, A. K., Taylor, E. M., Yoshida, K. and Peichel, C. L.(2011). Heritable differences in schooling behavior among threespine sticklebackpopulations revealed by a novel assay. PLoS ONE 6, e18316.
Watts, I., Pettit, B., Nagy, M., de Perera, T. B. and Biro, D. (2016). Lack ofexperience-based stratification in homing pigeon leadership hierarchies. R. Soc.Open Sci. 3, 150518.
Webster, M. M. and Laland, K. N. (2008). Social learning strategies and predationrisk: minnows copy only when using private information would be costly.Proc. R. Soc. B Biol. Sci. 275, 2869-2876.
Weihs, D. (1975). Some hydrodynamical aspects of fish schooling. In Symposiumon Swimming and Flying in Nature (ed. T. Wu, C. Brokaw and C. Brennen), pp.203-218. New York: Plenum.
Weimerskirch, H., Martin, J., Clerquin, Y., Alexandre, P. and Jiraskova, S.(2001). Energy saving in flight formation. Nature 413, 697-698.
Wolf, M. and McNamara, J. M. (2013). Adaptive between-individual differences insocial competence. Trends Ecol. Evol. 28, 253-254.
Wolf, M., Van Doorn, G. S. and Weissing, F. J. (2008). Evolutionary emergence ofresponsive and unresponsive personalities. Proc. Natl. Acad. Sci. USA 105,15825-15830.
Wood, A. J. and Ackland, G. J. (2007). Evolving the selfish herd: emergence ofdistinct aggregating strategies in an individual-based model. Proc. R. Soc. B Biol.Sci. 274, 1637-1642.
Yates, C. A., Erban, R., Escudero, C., Couzin, I. D., Buhl, J., Kevrekidis, I. G.,Maini, P. K. and Sumpter, D. J. T. (2009). Inherent noise can facilitate coherencein collective swarm motion. Proc. Natl. Acad. Sci. USA 106, 5464-5469.
Zienkiewicz, A., Barton, D. A. W., Porfiri, M. and Di Bernardo, M. (2015).Leadership emergence in a data-driven model of zebrafish shoals with speedmodulation. Eur. Phys. J. Spec. Top. 224, 3343-3360.
2983
REVIEW Journal of Experimental Biology (2016) 219, 2971-2983 doi:10.1242/jeb.129411
Journal
ofEx
perim
entalB
iology