Collective anti-predatory behaviour in animals groups Gregory M. Kohn Utrecht University, Neuroscience and Cognition Masters, Track: Beahvioural Neuroscience Supervisor: Claudio Carere Dipartimento di Scienze Ambientali Università degli Studi della Tuscia Largo dell'Università s.n.c. 01100 Viterbo Italy 1 Summary The collective behaviour of animal groups serves to reduce the risk of predation to individuals. By joining a group individuals remain in close proximity to others which reduces the risk of predation. Groups can also afford benefits by allowing for early detection of predators. During predatory attacks groups may confuse or deter predators by making it difficult for single individuals to be targeted. The mechanisms underlying many aspects of collective behaviour in response to predation are not well known. Here we review the models and studies of collective responses to predation and propose some underlying mechanisms for collective detection, evasion and communication about predators. In foraging flocks collective detection benefits individuals by providing them earlier detection of predators then when alone. Vigilance behaviour of neighboring individuals is copied. The amount of individuals displaying vigilant behaviour reaches a threshold level where the group initiates escape behaviors. Therefore predators raise vigilance levels and allow individuals to detect and escape predation earlier. Predators also attack free flying or free swimming groups. Here information about the location fo the predator may be spread through agitation wave. These waves of changing density have been shown the orient individual group members towards the location of the predator. These waves may also serve to confuse a predation attempt by making it harder for a predator to single out an individual within the group. Often prey species possess highly reflective feather or scale which reflect waves of light during agitation wave propagation, which may in turn further the confusion effect. In addition to performing waves free flying / swimming groups also exhibit consistent flocking patterns. These patterns have been shown reflect the level of predation risk experienced at certain areas. It is possible that flocking patterns represent honest signals of environmental conditions and may therefore serve as communication becons to other groups. Findings that neighboring groups tend to copy flocking patterns lends some support for this, albeit its mechanisms are unknown. However all aspects of collective behaviour involve groups moving and making decisions as a coordinated whole. Collective anti-predatory behaviour could be caused through individual threshold responses leading feeding back to collective group level behavior. In all situations individual responses to the presence or 2 attack of a predator reaches a certain level when positive feedback , through coping neighboring behaviors, spreads this behaviour through the entire group. This process allows groups to make informed and accurate decision when detecting, evading and communicating about predators. 1. Introduction From the synchrony of a flying flock of starlings (Sturnus vulgaris), to the vast schools of herring (Clupea harengus) in the open ocean, to the migration of wildebeest (Connochaetes taurinus) herds across the Serengeti, the collective behaviour of animals has always had the power to captivate man’s imagination. So compelling was the movement of bird flocks that the ancient Romans once saw it as proclaiming the will of the gods (Zimmer 2007). Until very recently it was also thought that flocking and schooling patterns were the result of some underlying psychic abilities of animals, not yet comprehensible to man (Long 2005). Although long misunderstood in antiquity, today collective animal behaviour has become a serious field of scientific investigation. Researchers are now beginning to delve into the mysteries which surround the collective movement of animal groups. Studies have shown that it is inter-individual interactions, and not psychic ability, which determines and guides collective behavior. While research into collective behaviour has yielded impressive results about the mechanisms involved in collective decision making and movement, it has remained somewhat distinct from other areas of behavioral biology. The incorporation of empirical studies and models of collective behaviour with other well studied areas of behavioral ecology and ethology, such as predator avoidance, will help enrich our knowledge of how social animals respond to environmental obstacles. This paper will review the mechanisms which animals employ in order to detect, evade and communicate about predators collectively. In order for any collective behaviour to occur animals must form and maintain groups. Being part of a group itself poses certain advantages and disadvantages in contrast to being solitary. It is widely assumed that the main benefit for being in a group is protection from predation (Hamilton 1971) which is the central focus of this review. 3 There are many different theories about the advantages which groups provide in reducing the risk of predation. The first and most intuitive reason to join a group is to increase the number of potential targets surrounding you, therefore decreasing the individual probability of being taken by predator. This is know as the dilution effect, and is seen as the basic factor moving most animals into groups (Wrona and Dixon 1991). Arranging into groups may also decrease the probability of encountering predators as more individuals will inhabit a limited spatial area then if individuals were more widespread and solitary (Zoratto et al., 2009). One example of a species utilizing the dilution effect can be seen in the parental care behaviour of ostriches (Struthio camelus). Ostriches have been know to actively adopt other juveniles into there brood. For a long time this apparently altruistic act was not well understood. A study by (Bertram 1992) found that raising precotial ostrich chicks was nearly cost free for the adults, and adoption therefore benefited parents own offspring by decreasing the predation risk. It was also found that individuals who adopted more offspring enjoyed a higher reproductive success. The dilution effect can also be seen in the mass breeding swarms of certain insects. Cicadas (genus Magicicada) for instance often have a biological cycle of about 17 years where all individuals in the population will hatch and breed synchronously. Other species such as wildebeest form huge annual aggregations when breeding and migrating. Therefore, the dilution effect can also be applied temporally as well as spatially to avoid predation. In order for the dilution effect to be functional individuals must maintain group cohesion in response to a predator in many situations. While simply being part of a group confers some advantages in reducing the predation risk not all individuals share these benefits equally. Spatial position within the group may also determine the level of risk individuals experience. Individuals at the periphery of the group are likely to experience a higher predation risk then individuals who are within the center of the group. The “selfish heard” theory posits that there will be individual competition within the group to gain a favorable position. Therefore as Hamilton (1971) put it is the, “selfish avoidance of a predator that can lead to aggregation”. Many predators will try to single out an individual at the periphery of a group to direct their attack. Groups arise as individual try to obtain a more central position in response to an attack. Thus in the “selfish heard” theory individuals try to 4 maximize the risk-dilution effect with larger groups offering bigger advantages However large groups have also been shown to increase the attention of predators (Pulliam 1973). One potential difficulty of the selfish herd theory is that individuals at the margins of the group may experience greater risk while in a group (because of the increased attention from predators) then when outside the group (Pulliam 1973). Therefore groups may be very unstable and may break apart easily. While the selfish heard theory may be a large factor in motivating individuals to form groups other benefits may be necessary in order to maintain group cohesion. Models of the selfish herd theory often do not include these other potential benefits such as the early detection of predators. Living in a group may confer an advantage in detecting predators. The “manyeyes” theory posits that the presence of many individuals increases the probability of detecting a predator earlier. Many studies have provided examples that groups tend to detect and evade predators sooner then singular individuals. For instance Kenward (1978) used trained goshawks (Accipiter gentilis) to attack groups of pigeons (Columba livia) (of differing sizes) as well as single individuals. It was shown that risk of mortality was greatest in singular doves and increased as group size decreased. Within larger groups the probability of an individual detecting a predator in close proximity is much higher then in smaller groups or when single. Therefore by joining a group, individuals will decrease the chances of becoming a victim of predation, as they are more likely to respond to the presence of a predator earlier. The ability for groups to detect predators earlier could be factor maintaining group integrity even when a predator is not immediately present. In order for the many-eyes theory to be effective however individuals in the group need to share information and coordinate escape response together. Later we look at the mechanisms which animal groups utilize in collective detection, mainly how information of the predator spreads throughout the group. The collective behaviour of animal groups may also serve as a means of actively deterring predatory attacks. Many studies have shown that predators are attracted to larger groups but actually experience a lower success rate in comparison to attacking singular individuals. This apparent contradiction may be explained by the groups deterring predation attempts through coordinated movements. The ability of groups to deter or “confuse” predators when attacking has been called the confusion effect. The 5 confusion effect is based on the assumption that it is easier for a predator to target a single individual when it is not surrounded by conspecifics. Cresswell (1994) looked at the success of predatory attacks after the predator had targeted a specific individual within the group (called “open attacks). These open attacks predators were seen as more vulnerable to the confusion effect. Surprise attacks also occurred where predators did not target a single individual within the group and would be less vulnerable to the confusion effect. As group size increased though both attacks elicited less successful captures then when attacking smaller flocks. This suggests that the presence of more individuals can somehow defect or deter predatory attacks. Krause and Godin looked at the preferences of Acara cichlid fish (Aequiden pulcher) when attacking shoals of guppies (Poecilia reticulate) of different sizes. Groups of different size were presented to the Acara cichlid which demonstrated a preference for attacking larger groups, but only if they exhibited more movement then smaller groups. Consequently, it seems that larger groups are more conspicuous, and it was this that caused the increased attraction. If smaller groups were experimentally manipulated to increase movement they were more also attractive to predator. Despite the increased attraction to larger groups predation success decreased as the size of the shoal increased. This also suggests that fish have some collective mechanism for deterring predation attempts. When large flocks of birds or schools of fish initiate escape maneuvers in response to a predator it can be difficult for a predator to maintain its focus on a single individual. Later we considers the mechanisms which groups use in order to deter predatory attacks through coordinated group movements (especially through cascading waves of changing density within a group). Information about the presence of a predator must be transferred throughout the group in order to initiate evasive maneuvers. The mechanisms which information about predators spreads in free flying/swimming groups, and its relation to deterring predation success, is discussed in detail later. Here we outlined the basic theories looking at the benefits groups confer in order to protect individuals from predation. For every theory presented groups must maintain cohesion and behave collectively in order to garner the benefits of group living. This means that individuals must be able to spread information between individuals and coordinate behaviour appropriately as a group. Here we review the mechanisms which 6 groups utilize in detecting, evading and communicating about predation. In order to get a bottom up picture of these mechanisms you must be able to measure and predict the movement of the individual group members. Previously this has been constrained to computer models which create hypothetical groups. These models have attempted to uncover the mechanisms behind collective behaviour from the level of the individual interactions (Conradt and Roper 2005). Here individuals within a group are given (programmed) with certain rules that determine their interactions with other individuals. These inter-individual interactions then have cascading influences into the collective behaviour of the group. Changing the rules of social interaction may lead to changes in the overall group behavior. While few of these models have specifically dealt with group response to predation their assumptions on group behaviour are may help in understanding collective responses to predation. Bird flocks and fish schools are some of the most studied model system in collective and behaviour and exhibit a coordinated and synchronous response to the behaviour of a predatory agent. The difficulty in capturing films of animal groups in three dimensions was a large impediment to looking at group behaviour in the field. Recently though techniques have become available that allow for tracking whole groups in three dimensions, as well as each individual within the group (Cavagna et al 2008). These techniques involve taking fast sequential images of a moving group in three dimensions, and using computer analysis to look at interactions between individuals. These procedures allow for a real time reconstruction of actual flocking patterns within a virtual environment where they can be explored in greater detail for the causal inter-individual mechanisms behind flocking behavior. Ballernini et al (2008) was able to capture the movement of starlings engaging in aerial displays and then analyze the basic movement of the flock, along with each individual. Here we seek to review the empirical studies and theoretical models of collective behavior, and integrate it within the frame work of predator detection, avoidance and communication. We also clarify and uncover some mechanisms operating within collective anti-predatory behaviour. This review will follow by investigating: (1) The mechanisms of collective predator detection, or how information about the presence of a predator spreads throughout the group initiating evasive action. (2) The mechanisms of predator evasion, or how groups actively avoid or deter a predator from attacking. (3) The 7 mechanisms of collective communication between animal groups about the presence of a predator. (4) The common mechanisms which groups utilize to be able to make decisions together in response to a predator. Here we will integrate empirical research looking at the actual behaviour of animal groups with computer models to infer for common mechanism in collective detection, response and communication of predators. Collective predator detection: Vigilance and social information Many studies have tried to quantify the amount of time which individuals spend looking out for predators, or being vigilant, in contrast to time spent in other activities such as foraging. Individuals within a group will likely face a tradeoff between maximizing foraging benefits and maintaining vigilance (Bednekoff and Lima 1998). Time spent feeding may leave an individual more vulnerable to predation, yet time spent looking out for predators will take away from time spent feeding. In field studies on bird flocks individuals are said to be foraging when their head is oriented towards the ground and vigilant (scanning) when the head is raised and the bird in an upright posture. Vigilance has been much harder to quantify in fish schools (Lima and Bednekoff 1999), but some studies have shown that fish tend to feed more when the perceived risk of predation is lower (Ryer & Olla 1991). This suggests that vigilance, in the form of scanning, may also be present in fish but much harder to quantify. Some of the earliest efforts to look at group protection from predation assumed that individuals joined groups to place individuals physically in-between themselves and a predatory attack, creating a selfish herd (Williams 1966). Models that tried to document the mechanisms of this selfish herd found that they were not stable (Pulliam 1973). Individuals at the group margins tended to have higher predation levels and were more inclined to depart. A large group would also likely attract the attention of more predators, and these marginal individuals would face less predation by foraging alone then on the periphery of the group. In order for groups to be stable other factors needed to be considered then just obstructing a predatory attack, like the early detection of predators. Pulliam (1973) created a model that looked at the dynamics of predator detection in bird flocks. His 8 model incorporated ideas from the “many-eyes” theory and assumed that when a predator is detected this information would spread throughout the group. This would in turn cause individuals to coordinate evasive action with larger groups better at detecting predators then smaller groups. Within the model only a single individual was needed to spot the predator in order to move the group to collective evasion. Pulliam did not consider the costs of foraging together and the dilution effect in his model which led it to some criticism. The collective vigilance behaviour of groups is highly unpredictable, and likely influenced by a multitude of confounding factors (Bedenkoff & Lima 1998). For instance Pulliams model assumes that the spread of information about a predator and evasive action is instantaneous and perfect. However, it has been shown that the individual who detects the predator may gain an advantage in avoiding predation over neighboring conspecifics (Lima 1995). While detectors can initiate evasive behaviour almost instantaneously, others have to interpret and respond to the behaviour of neighbors before taking evasive action. This leaves them at a disadvantage because it takes longer for them to initiate evasive action in contrast to neighbors who originally detected the predator. The dilution effect is thus likely to interact (and potentially conflict) with collective detection and influence the vigilance levels in the group (Bednekoff & Lima 1998). Group living also carries costs as individual may compete for resources which will in turn influence the levels of vigilance. Scramble competition with neighboring conspecifics will be intensified if food levels on a foraging patch are low. The levels of vigilance are likely to be a result of the relative differences in the costs and benefits. They will also be influenced by environmental conditions, group composition and individual dispositions. Despite the potential costs it is certainly advantageous to form foraging flocks in many situations as they are relatively common in numerous species. In these situations the benefits from the dilution effect and collective predator detection likely outweigh the costs of foraging alone in many circumstances. In order to understand the benefits of early predator detection there must be some understanding of the mechanisms underlying it. One common assumption of antipredatory vigilance models is that once a predator has been detected this information will spread throughout the group with other individuals responding appropriately. Social information, or information gathered by observing others interacting with the 9 environment, is thus key to understanding predator detection in groups of animals. There have been numerous theoretical attempts to investigate the mechanisms and factors influencing the amount of time spent scanning. Sirot (2006) for instance modeled the influence of social information on vigilance behaviour in flocks. She found that the perception of predatory risk was determined by the levels of vigilance within the group. Initially individual vigilance is a rather stochastic process with birds raising there heads and scanning at seemingly random intervals. Flocks with initially low levels of vigilance will tend to move to a more relaxed feeding behaviour whereas groups with higher initial vigilance levels will slowly rise in vigilance until a threshold is reached where the group takes off. A potential cost of this strategy is that flocks may create “imaginary” predators when vigilance levels reach a threshold, causing birds to exhibit evasive behaviour in absence of any real threat. These models show a high similarity to the quorum decision making seen in many species of fish (Ward et al. 2008). Vigilance levels tends to reach a threshold capacity which then creates a positive feedback response where the group takes evasive action by leaving the forging patch. Application of models of self-organization could further research on predatory vigilance. For instance, studies on the emergence of leadership often document that knowledgeable individuals, or ‘leaders’ often coordinate there behaviour more then other and thus could lower the threshold needed for evasive action. Birds with more knowledge about a certain area could thus serve as initiations of appropriate responses to predation threat. In areas where the predation threat is high knowledgeable individuals may display more scanning making the group more likely to take evasive behaviour then in areas where predation is low and individuals are more relaxed. Attention to the behaviour of others is an import determinant of group level vigilance. Empirical studies have also explored the social influences and transmission of vigilance within a group. Bekoff (1995) investigated flocks of evening grosbeaks (Coccothraustes uespertinus) and found that they maintained flock positions which facilitated information gathering between individuals. When groups were arranged in a line, where neighboring individual blocks visual inspection of other conspecifics, they were characterized by being more vigilant and less coordination in scanning behaviour then when arranged in a circular pattern. The assumption is that a circular pattern 10 facilitates social information transfer within the group because there is less visual obstruction then in a linear group patter. Fernández-Juricic and Kacelnik (2008) investigated the mechanisms of information transfer in small flock of three starlings. Three birds were lined up in a field with each bird placed in a bottomless cage (to be allowed to forage in the grass). The middle bird was designated as the ‘receiver’ which gained social information from the two distal birds or ‘senders’. It was shown that receiving birds tended to copy the scanning and foraging behaviour of conspecifics when they were in close proximity. Receivers also directed there scanning toward the senders more often when their behaviour differed from previously determined baseline levels. Increasing the distance between individuals decreased the fidelity of social information transfer. ‘Receivers’ coordinated their behaviour less with the ‘senders’ with increased distance All these studies suggest that information regarding the presence of a predator could be spread throughout a group by monitoring the vigilance of conspecifics. Higher levels of vigilance also tend to be reinforcing moving neighboring individuals to become more vigilant as well. When a predator is spotted information initially flows to near neighbors from the detector and quickly expands from the epicenter until threshold is reached and group flies away. By integrating and testing the assumptions of vigilance models in relation to the transmission of social information we can better understand the mechanisms of collective detection (Lima 1995). This may also help explain some ambiguities seen in some models. For instance some models predict that detection of predator by single individual may increase the predation risk of neighboring individuals, who must interpret the cues and respond to ‘detectors’, before initializing evasive action (Bedenkoff & Lima 1998). Through investigating the benefits of collective detection and social information use in contrast to the costs of increased predation risk associated with ‘detectors’, you may better predict the circumstances when social foraging is advantageous. However predators can also pose risks to groups of animals while they are in flight. In the next section we look at how flocks and school avoid predatory attacks once the decision to take evasion action is implemented. 11 Collective predator evasion: Waves of Agitation Tinbergen (1951) was among the first to describe the twisting and turning movement of starling flocks above their roosting areas. Since then many studies have looked at both the mechanisms involved in creating these intriguing collective displays (Couzin and Krause 2003). Comparatively fewer studies have investigated collective behaviour when groups are under direct predation However the minority of studies investigating collective responses to predation have documented certain group behaviors which may serve to inform groups on the position of the predator and deter predation through utilizing the confusion effect. For instance Michaelsen & Byrkjedal (2001) looked at the flocking patterns in groups of shorebirds in response to different predatory attacks. Predatory attacks were classifed as dives, pursuits or surprise-attacks. Flocks of Dunlins (Calidris alpina) and Common Ringed Plovers (Charadrius hiaticula) formed highly cohesive flocks when a predator was detected. The shape, pattern, and group composition were measured from video recordings to create a “collective” ethogram. Flocks exhibited a characteristic structure of flying low over the water in response to the presence of a predator in a cohesive ‘magic carpet’ like fashion. Their shape tended to be elongate with only a single layer of birds, thus giving the appearance of a carpet when in flight. The main predatory species of these flocks were Peregrine Falcons (Falco peregrinus), which often engage in diving attacks. The group formation served to prevent these attacks as they would not allow a falcon enough stalling time in order to avoid contact with the water. Because the falcon risks becoming waterlogged it often had to employ different hunting techniques, such as pursuing prey horizontally, which have a lower success rate then dives. The density of the group was higher in the anterior then in the posterior where a large amount of attacks were directed. Descriptions of group flight maneuvers under predation are sparse, despite their apparent relevance to evading predation attempts. The descriptions that exist though suggest some common mechanisms which groups use to coordinate their movement in order confuse the predator.. 12 While many studies and models have addressed predation threat in foraging aggregations comparatively fewer studies have addressed interactions with predators in free flying or swimming groups. In the previous section we documented that flocks initiate evasive action once a certain number of individuals exhibit vigilant behavior. While foraging on the ground it has been shown that certain group configurations lower the amount of visual obstruction that neighboring individuals impose and facilitate information flow between conspecifics (Bekoff 1995). Being able to interact with multiple individuals in three dimensions could help better coordinate movements and collective decisions then when on ground where more visual obstructions and limited interactions (Ballerini, et al. 2008). Evasive maneuvers in response to predation also often occur in three dimensions. Nonetheless, this makes it quantitatively much harder to look at the local interactions governing evasive behaviour in large flocks and schools. In three dimensions there are more avenues for interaction to occur then when foraging on the ground. This also means that different mechanisms may apply when social information is transferred throughout the group. For instance in foraging groups the quality of information received is a function of the metric distance of conspecifics (Fernández-Juricic and Kacelnik 2008). Increased distance from neighbors decreases the quality of information being transferred between individuals. In free flying groups the relationship may be much different. Ballerini et al (2008) filmed and reconstructed flocks in three dimensions to look at the local rules governing collective behavior. They found that individual birds tend to consistently align themselves with around six to seven birds at a time [other experiments looking at numerical abilities of birds often suggests that many species can accurately distinguish up to seven units (Pepperberg 2006) ]. This alignment was also not entirely dependent on the metric distance of neighboring individuals. A focal individual would orient him/herself in a way that the closest nearest neighbors would be located at the sides of the individuals. This could be to free up flying space ahead of the bird, or it could be that birds have anisotrophic vision (visual system with eyes located on the sides of the head). Therefore topological orientation, or how many individuals you keep within your visual field, may be more important then the distance between neighboring individual. While it is not been explicitly demonstrated, this orientation style could have many implications for the flow of information in free 13 flying and swimming groups. Flocks often change shape, size and density very rapidly in the presence of a predator. Topological orientation, which is not dependent on metric distance, allows for changes of density to flow throughout the flock while still maintaining flock integrity. Changes in density and metric distance between individuals may serve as a means which information is transmitted thorough the group. Waves of changing densities often propagate throughout a group (Carere et al. 2009) in a similar fashion to ripples in a puddle, and may carry with them information about a predator, such as its location. Within bird flocks and fish schools waves are exemplified as propagating an epicenter of a few individuals [usually in closest proximity to the predator (Happer, 1997)]. Carere et al (2009) quantified the flocking behaviour of starlings in relation to the predation risk in certain roosting areas. They observed consistent agitation waves within flocks engaging in aerial displays. Waves in response to a predator have been observed and investigated more extensively within fish schools. Sonar is utilized to measure the changes in shape and density within large schools of fish in response to predation. For instance Gerlotto (2006) looked at the ‘waves of agitation’ in fish schools disturbed by predatory sea lions (Callorhinus ursinus). Waves perpetuated throughout a group just as a sea lion attacked. Before an attack group internal structure was characterized by patterns of ‘vacuoles’ or empty spaces and ‘nucleus’ or areas of high densities within the shoal. Following an attack and progression of wave of agitation the group exhibited a more homogenous density with individuals tending to orient themselves in a certain direction. Radkov (1973) also documented changes in internal organization of captive schools in response to a wave. At the distal end of the school a few individuals were given a fright stimulus which initiated a wave of agitation across the group. After the progression of the wave individuals in the school consistently oriented themselves in the direction of the fright stimulus (get figure). This suggests that waves could relay information about the location of the predator quickly throughout the whole group. These ‘social waves’ may also be a means which flocks and schools can deter attacks by utilizing the confusion effect (Kastberger 2008). The collective movement of group members may make it hard for a predator to target a specific individual within the group. Predators are also preferentially attracted to larger groups but enjoy a lower 14 suiccess then when attacking smaller groups. Zheng et al (2005) modeled the interactions between the confusion effect and predation. In his model fish could respond to predation using three component behaviour patterns: (1) move in the coordination with neighbors, (2) avoid collisions and (3) escape singly without regard for the movement of near neighbors. Manipulating the weight of these three heuristics when making decision in virtual predator-prey scenarios influenced the group behaviour and success rate of predators. By coordinating movement with neighbors schools were able to use the confusion effect to deter virtual predators from successful attacking the group. Agitation waves, while potentially serving as a means to spread information about the location of the predator (Gerlotto 2006) may even deter predation. Giant honeybees (Apis dorsata) swarms form coordinated waves across their hive in response to the threat of predation. By flipping there abdomen over and displaying their shinny underside individuals collectively propagate a shimmering wave of reflected light across the group. Giant honeybees are sting-less and live in open hives that are not surrounded by a protective layer as in many temperate species. Kastberger (2008) found that these shimmering waves effectively deterred predatory wasps from attacking within a set area around the nest site. Wave strength and rate corresponded with the strength and closeness of the predatory hornet. Hornets responded to these waves by avoiding areas within close proximity to the hive, and foraging on isolated bees away from the hive. Consequently, the shimmering waves served to create a “safe haven” around the nest where predation was prevented. Cohesive flocks of starlings engaging in aerial displays were more common in areas of high predation (Carere et al 2008). Similarly, predation attempts directed at these cohesive flocks were less successful then attempts directed at less cohesive flocks. Starlings possess a glossy black plumage which is excellent at reflecting light, and waves often correspond with bright flashes of light. Other flocking species such as dunlins also exhibit bight flashes of light as waves disseminate throughout the group (Parrish and Hamner 1997). Wave propagation through fish schools is also often associated with flashes of light (Gerlotto et al. 2006). Fish scales can be highly reflective and wave propagation in schools often coincides with a rolling motion by the fish which reflects sunlight outward. These shimmering flashes may aid the confusion effect by visually making the group a coordinated whole, and therefore also making it harder for 15 predators to single out individuals. Agitation waves which cause flashes of light may therefore aid in deterring predation thorough the confusion effect. Because data regarding the effects of agitation waves in response to predation is sparse follow up experiments need to be conducted in order to validate these assumptions. Although it warrants significantly more study, agitation waves could be an important component in the spread of social information about the presence of the predators and a potential mechanism for predatory deterrence. Due to the increasingly sophisticated techniques used to measure animal groups in three dimensions it is now possible to look at the dissemination of waves from the level of the individual interactions. These studies could help elucidate the anti-predatory function of waves. For instance experiments could investigate if predators have lower success rate when attacking a group engaging in a wave then they are not. Fish tend to orient towards the attack stimulus after wave propagation (Radakov 1973). This suggests that waves carry basic information about the position of the predation in relation the group as a whole. Other evasive techniques such as the twist and turns of flocks and schools may rely on information provided by waves as a framework. Therefore in order to understand the complex flocking or schooling patterns in response to predation you may need to understand first the interactions within and subsequent reactions to an agitation wave. The next section investigates if the shapes which flocks form can also be utilized as “communication beacons” to other flocks or single individuals about predation. 16 From (Heppner 1997). A propagation of an agitation wave in a flock of birds. This photo also shows the reflected flash of light seen also seen in wave propagation in both fish schools and bees. Information Transfer between groups The behaviour of groups themselves may serve as a form of communication to other groups about the presence of a predator. Feare (1984) hypothesized that the shapes formed during flocking displays served to inform other individuals about the location of the roosting site. Species typical flocking behaviour has been documented in some species, such as starlings, and has been used to identify species from afar where individuals cannot be discerned. Animal aggregations have been known to be important cues used by individuals when assessing and choosing between foraging patches, but there is no evidence that individuals utilize specific patterns of aggregation when making decisions. Empirical investigations of information transfer between flocks have yet to be 17 conducted and therefore its mechanisms (if it occurs) are unknown. However flocks do often exhibit consistent patterns in certain situations. If these patterns are consistent enough to convey accurate information to other flocks evolution might favor inter-flock communication. Therefore one of the first steps to investigating inter-flock or school communication is to see if flocking / schooling patterns represent honest signals about environmental conditions which others can utilize. Starlings often forage during the day in the countryside and return to roosting areas during the evening. When arriving at roosting areas starlings will form vast flocks with hundreds or thousands of individuals. Flocks often engage in complex aerial displays above the roosting site before settling in for the evening. The purpose of these aerial displays is not entirely known (as they often occur without predation as well). Carere at al (2008) classified different starling flocking patterns in relation to the level of predator threat in certain areas. Two roosting sites in Rome were monitored and the flocking patterns recorded. From these recordings a flocking ethogram was developed with twelve distinct flocking patterns. It was also noted how abundant the main predator , the peregrine falcon (Falco peregrinus) was at each location and the number of falcon attacks upon starling flocks (either successful or unsuccessful). This information could be used to deduce the predation risk at each site. Large cohesive flocks were found at higher proportions in areas with higher predation risk, whereas in areas with lower predation risk smaller and less cohesive flocks were more apparent. Consequently different flocking patterns accurately correlated with the level of predation risk. This suggests that specific patterns could potentially function as communication beacons by transmitting accurate information about predation risk to other flocks. Moreover, Carere (2008) found that antipredatory displays by flocks outside but still in view of the roosting area influenced the probability of observing similar flocking patterns on the roosting site. This suggests that flocking patterns may served as a form of communication between groups with neighboring groups copying patterns from each other. Predation tactics are often restricted and can influence the shape of fish schools in predictable ways (Parrish 1992). Pitcher and Wyche (1983) were able to classify different schooling patterns based on the predation tactics used by predatory sand eels. Nottestad and Axelsen (1999) looked at the response of schools of herring to predation by killer whales (Orcinus orca) and found 18 that groups increased in density and displayed consistent evasion patterns when under attack. These examples show that flocking and schooling formation often correlate with risks of predation, types of predators, and types of predatory attack behavior. As far as we know there are no studies which have explicitly documented or provided direct evidence for inter-flock communication. The pre-requisites for inter-group communication, mainly that flocking patterns represent honest signals regarding predation are well founded. It is possible that instances of inter-flock communication individuals may simply be simply reacting to similar or even the same stimuli and therefore matching flocking patterns. Artificial or trained predators could be used to elicit a response a single flocks and then quickly removed from view. Changes in flocking patterns of neighboring flocks could then be observed to see if they tend to match each other. This study may suggest that flocks do indeed match other flocks patterns and do not simply respond to the same predatory agent. Further research should now be conduced to (1) develop large scale ethogams for flocking and schooling patterns in relation to contextual and environmental variables (2) document copying of flocking patterns between groups.. In order for inter-group communication to take place it individuals should be able to discriminate between different flocking patterns. Numerous studies have shown that birds can categorize different visual stimuli. Pigeons have been a classic model for visual categorization and can learn to recognize pictures based on the presence of human figures, shapes and even weather conditions (Dittrich et al 1998). Similar studies could be conducted by exposing individual birds to videos of different flocking patterns. Birds can first learn to peck at a button or light to receive a food reward when videos of flocking behaviour are played. Over training periods only certain flocking patterns come to deliver rewards. Individuals could then be tested to see if they can differentiate between patterns by presenting them with novel videos of different flocking patterns and record there response. Inter-group communication is a potentially exciting field of research that deserves in depth investigation and documentation. 19 Figure 2 From (Carere et al. 2009), a collective ethogram of different starling flocking patterns. Self-organization from foraging flocks to intra-group communication Group living allows individuals to reduce their risk of being taken by a predator. However in order for individuals to garner protection from group living they must coordinate their behaviour with surrounding individuals. The evolution of social behavior, or being able to respond appropriately to the behaviour of conspecifics, laid the ground work for the development of collective group level behaviors. For example Alberts (2007) found that rat (Rattus rattus) pups began coordinating group movements only when they developed the ability to respond to conspecifics (sociality), beforehand individual movements were random. Moreover, rats also developed coordinated responses to the presence of the mother which were absent before sociality developed. One way of investigating the mechanisms within collective behaviour is through the lens 20 of self organization (Sumpter 2006). Self organization is when the group behaviour is defined by the interaction of lower level components within the system. Interactions between these components create larger scale structures, which can then feedback and influence the behaviour of the components accordingly. Within animal groups this means that the collective group behaviour is determined by the simple interaction rules between individuals. Self organizing systems can be seen in a wide scale of natural phenomenon from the movement of people crossing a city street to the migration of birds and the fluctuation of the stock market (Couzin and Krause 2003). Traditionally self organizing phenomenon have been investigated using mathematical models, although techniques are becoming available that allow for accurate measurement of self organizing groups in the laboratory and field. The key to understanding self-organization in groups is positive feedback. A positive feedback system is where the behaviour of the group feeds-back into the interactions of the components in order to maintain and amplify the group behaviour (Couzin and Krause 2003). In order for this to occur individuals must be able to copy the behaviour of neighboring individuals. An example of a positive feedback system can be seen in the creation of game trails by many ungulate species (Couzin and Krause 2003). Considering a uniform environment many models have shown that individuals will initially choose to travel at random leaving a slight path behind them. Once a path is created though it is easier for other individuals to use this already partially made path. Therefore, the initial creation of a path feeds-back into other individuals using this path until a permanent trail system is created. Trail systems can thus be said to be a selforganizing system. Within predator detection, evasion and communication positive feedback systems may be operating to ensure that individuals collectively coordinate their behavior. During predator detection positive feedback enables vigilant behaviour (and thus information about the presence of a predator) to spread throughout the group. During predator evasion positive feedback propagate and agiation wave throughout the group. Utilizing positive feedback is one mechanism which groups may use to collectively decide on a course of action, or reach a consensus. Reaching a consensus is crucial for collective predator detection and evasion. There are many different routes 21 which individuals within a group can utilize in order to reach consensus between mutually exclusive decisions. Some of these include individual group members averaging there information together in what is called “information pooling” (Conradt and Roper 2005). A classic example of this is the popular fair game where individuals must guess the weight of the ox or other livestock and then receive prizes for the best guess. In 1906 Francis Galton obtained all the individual guesses from a local fair and averaged them to see if they matches the weight of the ox. While individual guesses were rarely close to the average weight of the ox the average of all the guesses was within one pound of the actual weight. Information pooling thus generates accurate predictions about environmental conditions but has no mechanism for coordinating collective action. Animals groups are often not capable of global communication. For instance an individual herring at the periphery of a large herring school may not be able to directly communicate with an individual at the center of the school. Starlings for instance are only able to maintain contact with seven individuals while flying (Fernández-Juricic and Kacelnik 2004).This potentially makes averaging information over large groups quite difficult and slow/ because interactions are limited to neighboring individuals. In flocks and schools that need to respond to predatory attacks information transfer and consensus need to be reached quickly and accurately. By utilizing reaction thresholds and positivefeedback systems animal groups may be able to respond to predation without “information pooling”. Here a certain number of individuals within a group display a specific behaviour together (threshold value) which subsequently initiates a positive feedback response that spreads and amplifies this behaviour (and associated information) throughout the group. Once the behaviour and information has spread throughout the group it can collectively move towards a decision together (consensus). This is a parsimonious hypothesis for how groups can reach a consensus without resorting to cogntivley sophisticated behaviours. Understanding the mechanisms within antipredatory behaviour from on the ground foraging flocks to inter-flock communication may be furthered by looking at threshold values and positive feedback responses. Being in a group confers benefits to individuals by diluting the predation risk. Therefore, maintaining group integrity and coordinating group behaviour might confer benefits upon individuals when responding to predation. In order to respond appropriately 22 to the presence of a predator it must first be detected. Individuals within foraging groups alternate between being vigilant and foraging. The key to understanding collective predator detection and response is threshold and feedback of vigilant behaviour. Once vigilance levels reach a threshold value all individuals become aware of the danger and collectively respond by flying away from the foraging area. This requires rapid spread of information throughout the group which initiates a collective response (Sumpter et al 2008). Feedback ensured that vigilance levels rise to the point of initiating evasive action. Once the group is airborne the threat of predation dose not always vanish, and often the group must respond to pursuit or further attacks. Information transfer throughout the flock here must be very fast and efficient. Agitation waves may serve to transfer information about the location of the predator quickly throughout the group in order to initiate appropriate evasive techniques (Radakov 1973). As in collective detection information must be spread rapidly throughout free flying flocks and schools in order to initiate collective maneuvers. Threshold values may also be important in initiating an agitation wave. A certain number of individuals may respond to the presence /attack of a predator by moving in towards the center of the group This movement may then be copied by surrounding individuals. When a certain threshold number of individuals responds to a predator by moving in unison an agitation wave is instigated. The threshold values that initiate wave propagation will likely be low because the speed which groups must respond to a predatory attack. Waves then advance as individuals copy the responses of neighboring individuals by moving closer to each other with positive feedback perpetuating the wave throughout the group. One common way for groups to deal with the threat of predation is to form large, dense flocks. In flocks of starlings denser flocking arrangements were observed in areas that had a higher risk of predation (Carere et al. 2009). Forming dense arrangements may facilitate information transfer through positive feedback. For instance models looking at the collective behaviour of self-propelled particles have showcased the influence of density in determining group coordination (Sumpter et al 2008, Vicsek et al 1989). Particles within a virtual environment can move in two ways, either by coordinating the movement in alignment with another particle or by moving randomly. At low densities particles behave stochastically, often randomly exploring the environment with little 23 coordination with other individuals. At medium densities particles tend to coordinate their movement with others or employ random exploration. The group as a whole exhibits a fluctuating pattern of mass coordination with mass un-coordination. Particles at high densities tend to exhibit consistent coordinated motion in a single direction on both the individual and group level. To summarize, once a threshold density is reached it becomes easier for the group to coordinate actions together. In studies of collective behaviour you can observe that denser groups are be able to coordinate evasive maneuvers better then groups with lower densities. Denser flock arrangements may also increase the utility of positive feedback by allowing for quicker copying of neighboring conspecifics. Waves of agitation for instance may propagate information throughout a group by locally increasing the density of individuals. This can be seen most clearly when groups adopt a more homogenous and oriented internal structure after wave propagation. The significance of forming dense group patterns in response to predation may therefore increase the utility of positive feedback and the spread of information throughout the group. A largely unexplored question is how the self organizing characteristics of flocks and schools influence inter-group communication. The presence of a predator is often a very palpable stimuli which individuals need to respond. Flocking and schooling patterns represent a more ambiguous stimulus because of there dynamic and continually changing structures. However flocks and schools do exhibit clear patterns in response to predation (such as creating dense flocking and schooling arrangements). This means that specific patterns could be co-opted as signals to other groups about the risk of predation in a specific area. If it is documented that individual birds can recognize and classify flocking patterns we can then start to look at the mechanisms by which inter-group communication proceeds. Hypothetically individuals could classify different flocking patterns and respond to patterns indicative of predation as they would if a predator was present. When a threshold number of individuals copies the patterns of a neighboring group positive feedback is initiated and the behaviour spreads throughout the entire group Therefore through threshold mediated positive feedback interactions groups may adopt similar flocking patterns. Inter-group communication is an unknown area of study with many potentially exciting findings for how animal groups respond to predation 24 Models of collective behaviour have focused either on the benefits of forming and maintaining groups (in response to predation) or on the self-organized mechanisms underlying group behavior. Integrating self–organization models with theories of the antipredatory advantages of groups will help further our understanding of how animal groups reduce the risk of predation. Here we saw how the self organizing process of threshold and positive feedback could help uncover potential mechanisms that groups use when detecting, evading and communicating about predators. However this still needs to be validated experimentally. Confounding environmental factors not included in our discussion may influence the grouping arrangements and flocking patterns seen in the wild. For instance similar environmental conditions, such as the size of roosting areas,, may induce similar flocking patterns and also attract predators. This might lead to a correlation between specific patterns and predation risk. The mechanisms underlying group behaviour could also be confounded by individual differences and propensities. Further studies that take into account confounding factors will give us a better picture of the advantages and mechanisms underlying collective anti-predatory behaviours. Index Acara cichlid fish, 6 Cicadas, 4 Common Ringed Plovers, 12 confusion effect, 2, 5, 12, 14 dilution effect, 3, 4, 9 Dunlins, 12 flocking patterns, 2, 7, 12, 17, 18, 19, 20, 24 Giant honeybees, 15 goshawks, 5 herring, 2, 18, 21 information pooling, 21 ostriches, 3 Pigeons, 19 positive feedback, 2, 10, 20, 21, 23, 24 predatory sand eels, 18 rat, 20 Romans, 3 Self organization, 20 selfish heard, 4 self-organization, 10, 20 starling, 11, 17, 20 starlings, 2, 7, 11, 14, 15, 17, 23 vigilance, 2, 8, 9, 10, 22 wildebeest, 2, 4 References 25 Alberts JR (2007) Huddling by rat pups: Ontogeny of individual and group behavior. Developmental Psychobiology 49:22-32. Ballerini M, Cabibbo N, Candelier R, Cavagna A, Cisbani E, Giardina I, Lecomte V, Orlandi A, Parisi G, Procaccini A, Viale M, Zdravkovic V (2008) Interaction ruling animal collective behavior depends on topological rather than metric distance: Evidence from a field study. Proceedings of the National Academy of Sciences 105:1232-1237. Ballerini M, Cabibbo N, Candelier R, Cavagna A, Cisbani E, Giardina I, Orlandi A, Parisi G, Procaccini A, Viale M, Zdravkovic V (2008) Empirical investigation of starling flocks: a benchmark study in collective animal behaviour. Animal Behaviour 76:201-215. Bednekoff PA, Lima SL (1998) Randomness, chaos and confusion in the study of antipredator vigilance. Trends in Ecology & Evolution 13:284-287. Bekoff M (1995) Cognitive ethology, vigilance, information gathering, and representation: Who might know what and why? Behavioural Processes 35:225237. Bertram BCR (1992) The Ostrich Communal Nesting System. Princeton University Press. Carere C, Montanino S, Moreschini F, Zoratto F, Chiarotti F, Santucci D, Alleva E (2009) Aerial flocking patterns of wintering starlings, Sturnus vulgaris, under different predation risk. Animal Behaviour 77:101-107. Cavagna A, Giardina I, Orlandi A, Parisi G, Procaccini A (2008) The STARFLAG handbook on collective animal behaviour: 2. Three-dimensional analysis. Animal Behaviour 76:237-248. Cavagna A, Giardina I, Orlandi A, Parisi G, Procaccini A, Viale M, Zdravkovic V (2008) The STARFLAG handbook on collective animal behaviour: 1. Empirical methods. Animal Behaviour 76:217-236. Conradt L, Roper TJ (2005) Consensus decision making in animals. Trends in Ecology & Evolution 20:449-456. Couzin ID, Krause J Self-Organization and Collective Behavior in Vertebrates. Advances in the Study of Behavior 32:1-75. Couzin ID, Krause J, Franks NR, Levin SA (2005) Effective leadership and decisionmaking in animal groups on the move. Nature 433:513-516. 26 Cresswell W (1994) Flocking is an effective anti-predation strategy in redshanks, Tringa totanus. Animal Behaviour 47:433-442. Feare CJ (1999) The Starling. Shire Publications. Fernández-Juricic E, Kacelnik A (2004a) Information transfer and gain in flocks: the effects of quality and quantity of social information at different neighbour distances. Behavioral Ecology and Sociobiology 55:502-511. Fernández-Juricic E, Kacelnik A (2004b) Information transfer and gain in flocks: the effects of quality and quantity of social information at different neighbour distances. Behavioral Ecology and Sociobiology 55:502-511. Gerlotto F, Bertrand S, Bez N, Gutierrez M (2006) Waves of agitation inside anchovy schools observed with multibeam sonar: a way to transmit information in response to predation. ICES J. Mar. Sci. 63:1405-1417. Hamilton WD (1971) Geometry for the selfish herd. J. Theor. Biol 31:295-311. Kastberger G, Schmelzer E, Kranner I (2008) Social Waves in Giant Honeybees Repel Hornets. PLoS ONE 3:e3141. Kenward RE (1978) Hawks and Doves: Factors Affecting Success and Selection in Goshawk Attacks on Woodpigeons. Journal of Animal Ecology 47:449-460. Krause J, Godin JJ (1995) Predator preferences for attacking particular prey group sizes: consequences for predator hunting success and prey predation risk. Animal Behaviour 50:465-473. Leif Nøttestad and Bjørn Erik Axelsen, Leif Nøttestad et Bjørn Erik Axelsen, Government of Canada NRCC, Gouvernement du Canada CNDRC Herring schooling manoeuvres in response to killer whale attacks. Lima SL (1995) Collective detection of predatory attack by social foragers: fraught with ambiguity? Animal Behaviour 50:1097-1108. Long WJ (2005) How Animals Talk: And Other Pleasant Studies of Birds and Beasts. Bear & Company. Makris NC, Ratilal P, Jagannathan S, Gong Z, Andrews M, Bertsatos I, Godo OR, Nero RW, Jech JM (2009) Critical Population Density Triggers Rapid Formation of Vast Oceanic Fish Shoals. Science 323:1734-1737. Michaelsen T.C. & Byrkjedal I. (2002) 'Magic carpet' flight in shorebirds attacked by raptors on a migrational stopover site. ARDEA 90 (1): 167-171 27 Parrish DJK, Hamner WM Animal groups in three dimensions. Parrish JK[ (1991) Do Predators 'Shape' Fish Schools: Interactions Between Predators and Their Schooling Prey. Netherlands Journal of Zoology 42:358-370. Pepperberg I (2006) Grey parrot numerical competence: a review. Animal Cognition 9:377-391. Pulliam HR (1973) On the advantages of flocking. J. Theor. Biol 38:419-422. Radakov D (1973) Schooling in the ecology of fish : Translated from the Russian by H. Mills. Wiley [and] Israel Program for Scientific Translations Jerusalem, N.Y. Ryer CH, Olla BL (1991) Information transfer and the facilitation and inhibition of feeding in a schooling fish. Environmental Biology of Fishes 30:317-323. Sirot E (2006) Social information, antipredatory vigilance and flight in bird flocks. Animal Behaviour 72:373-382. Sumpter D, Buhl J, Biro D, Couzin I (2008) Information transfer in moving animal groups. Theory in Biosciences 127:177-186. Sumpter DJ, Krause J, James R, Couzin ID, Ward AJ (2008) Consensus Decision Making by Fish. Current Biology 18:1773-1777. Tinbergen N (1969) The Study of Instinct. Oxford Univ Pr. Vicsek T, Czir�k A, Ben-Jacob E, Cohen I, Shochet O (1995) Novel Type of Phase Transition in a System of Self-Driven Particles. Phys. Rev. Lett. 75:1226. Ward AJW, Sumpter DJT, Couzin ID, Hart PJB, Krause J (2008) Quorum decisionmaking facilitates information transfer in fish shoals. Proceedings of the National Academy of Sciences 105:6948-6953. Williams GC (1996) Adaptation and natural selection. Zheng M, Kashimori Y, Hoshino O, Fujita K, Kambara T (2005) Behavior pattern (innate action) of individuals in fish schools generating efficient collective evasion from predation. Journal of Theoretical Biology 235:153-167. Zimmer C (2007) From Ants to People, an Instinct to Swarm. The New York Times. Zoratto F, Santucci D, and Alleva E (2008) Theories commonly adopted to explain the antipredatory benefits of the group life: the case of starling (Sturnus vulgaris). Masters Internship Thesis. 28