Antipredator Behaviour: How Prey Detect, Deter and Escape Their Predators

Topic: Animal Behaviour & Ethology Subject: Predation Risk, Vigilance, Alarm Signals and the Landscape of Fear Last verified: 2026-09-23

Almost every animal is potential prey for another, and predation has long been recognised as one of the strongest selective forces shaping animal bodies and behaviour. Prey are not passive victims: they assess their risk of being eaten and adjust what they do from moment to moment, trading safety against feeding, mating and rest [Lima & Dill 1990]. The resulting repertoire of antipredator behaviour is remarkably diverse. It includes scanning for danger and gathering in groups, giving alarm calls that can distinguish between types of predator, mobbing and eavesdropping on other species, advertising unprofitability to a pursuer, hiding through camouflage, and responding to chemical cues released by injured companions [Elgar 1989; Seyfarth et al. 1980; Magrath et al. 2015; FitzGibbon & Fanshawe 1988; Stevens & Merilaita 2009; Chivers & Smith 1998]. Field experiments have also shown that fear itself, independent of any killing, can reduce the number of offspring wild animals produce and alter entire food webs [Zanette et al. 2011; Suraci et al. 2016].

This article explains how behavioural ecologists think about predation risk, then describes the main forms of antipredator behaviour and the evidence behind them. It closes with the practical relevance of this research for reintroductions, for predator-naive species and for managing human disturbance. The honeybee waggle dance, filial imprinting, songbird vocal learning and dominance hierarchies are covered in companion articles in this series.

Predation Risk as a Behavioural Decision

In a landmark review, Steven Lima and Lawrence Dill argued that animals also assess and behaviourally influence their risk of being preyed upon in ecological time, that is, within their own lifetimes [Lima & Dill 1990]. They broke the predation process down into components of risk: how often a prey animal encounters predators, how likely it is to be killed once an encounter occurs, and how long it spends exposed to danger. They then reviewed evidence that animals' ability to assess and control one or more of these components strongly influences their decisions [Lima & Dill 1990].

The key insight of the review was that such decisions reflect trade-offs. The literature showed antipredator decision-making in feeding animals, in animals deciding when and how to escape, when and how to be social, and even, in fishes, when and how to breathe air at the surface [Lima & Dill 1990]. In each case, avoiding risk carries a cost in lost feeding, mating or other opportunities, and the balance an animal strikes depends on how dangerous its situation is. Lima and Dill identified the influence of predation risk on reproduction as the most important gap in the field at the time [Lima & Dill 1990].

This framing explains why animals can be affected by predators they are never caught by: the costs of avoiding risk are paid whether or not an attack ever comes [Lima & Dill 1990; Zanette et al. 2011].

Safety in Numbers: Vigilance, Dilution and the Selfish Herd

One of the most commonly cited benefits of foraging in groups is earlier detection of predators. An early mathematical model by H. Ronald Pulliam predicted that as group size increases, each individual can afford to spend less time scanning, because the group as a whole is still likely to spot an approaching predator [Elgar 1989]. This negative relationship between group size and individual vigilance, often called the group-size effect, has been reported in many birds and mammals [Elgar 1989; Roberts 1996].

The evidence is less straightforward than the idea. Mark Elgar reviewed more than fifty studies of birds and mammals that reported the relationship and assessed how well each had controlled for confounding variables. These included food density, competition, distance to refuge, visibility and group composition [Elgar 1989]. He concluded that most studies failed to demonstrate an unambiguous relationship between vigilance and group size, while noting that many revealed features of the relationship that deserved further research [Elgar 1989].

Why individual vigilance declines in larger groups is itself debated. One explanation, the "many eyes" or group-vigilance hypothesis, emphasises collective detection. The other emphasises that each individual's own risk falls in a larger group, for example because an attacking predator is likely to take only one victim, so that risk is diluted among group members [Roberts 1996]. Gilbert Roberts reviewed these alternatives and proposed that the individual-risk hypothesis, with group vigilance as one element within it, offers the more general framework for understanding how vigilance varies with group size and with other factors [Roberts 1996].

Groups need not be cooperative to be protective. In a celebrated paper, W.D. Hamilton argued against the view that gregarious behaviour evolves because it benefits the population or species [Hamilton 1971]. He imagined frogs sitting on the rim of a circular pond that shelters a water snake which, when it surfaces, seizes the nearest frog. Each frog's "domain of danger" is the stretch of rim on which the snake could appear and find that frog closest, and a frog can shrink its domain by jumping into a gap between neighbours [Hamilton 1971]. Using such simple models, Hamilton showed that selection is likely to favour individuals that move close to others, so that aggregations can arise purely from each animal's attempt to put others between itself and danger. He called this the selfish herd [Hamilton 1971].

Alarm Calls, Eavesdropping and Mobbing

Many animals call when they detect a predator, and some calls carry specific information about the danger. The classic demonstration comes from vervet monkeys (Chlorocebus pygerythrus), which give acoustically different alarm calls to different predators [Seyfarth et al. 1980]. Robert Seyfarth, Dorothy Cheney and Peter Marler played recordings of these calls to monkeys when no predator was present. Leopard alarms caused the monkeys to run into trees, eagle alarms caused them to look up, and snake alarms caused them to look down [Seyfarth et al. 1980]. Adults called mainly to leopards, martial eagles and pythons, whereas infants gave leopard alarms to various mammals, eagle alarms to many birds and snake alarms to various snake-like objects. Predator classification improved with age and experience [Seyfarth et al. 1980]. Because the recorded calls alone were enough to elicit the escape response appropriate to each predator, the study became a standard example of alarm calls that carry information about the type of danger [Seyfarth et al. 1980].

Alarm calls can also encode how urgent a threat is. In meerkats (Suricata suricatta), a social mongoose of southern Africa, Marta Manser found that acoustically different alarm calls are given to different types of predator, but that call structure also varies with the level of urgency [Manser 2001]. Low-urgency calls tended to be harmonic across all predator types, while high-urgency calls were noisier. Information about urgency followed a general acoustic rule, whereas the information about predator type was not encoded in an obvious way [Manser 2001]. Meerkat alarm calls therefore combine referential and motivational information in a single graded system [Manser 2001].

Some birds combine call elements into sequences whose meaning depends on their order. Toshitaka Suzuki, David Wheatcroft and Michael Griesser studied the Japanese tit (Parus minor), which has more than ten different note types and uses them singly or in combination [Suzuki et al. 2016]. In playback experiments, receivers responded to "ABC" notes by scanning for danger and to "D" notes by approaching the caller. When the two were combined as "ABC–D", birds did both. When the order was artificially reversed to "D–ABC", receivers rarely scanned and approached [Suzuki et al. 2016]. The authors presented this as the first experimental evidence for compositional syntax in a wild animal, and suggested that this mechanism may have evolved independently of human language [Suzuki et al. 2016].

Alarm calls are not heard only by the caller's own species. Robert Magrath and colleagues reviewed eavesdropping on the alarm calls of other species and found that it had been shown experimentally in about 70 vertebrate species, involving both closely and distantly related species [Magrath et al. 2015]. Benefits include prompting immediate antipredator responses, indirect gains in foraging or habitat use, and learning about predators [Magrath et al. 2015]. Some species act as "keystone" information producers, while others mostly seek information, which may shape how mixed-species groups form and function [Magrath et al. 2015].

Prey sometimes turn on the predator instead of fleeing. Mobbing, in which prey approach and harass a predator, has attracted at least ten suggested functional hypotheses across fishes, birds and mammals [Curio et al. 1978]. Eberhard Curio and colleagues tested one of them experimentally with captive European blackbirds (Turdus merula). A blackbird that saw a mobbing companion at the same time as a novel, harmless bird went on to mob that innocuous bird, and continued to do so when the novel bird was later presented alone [Curio et al. 1978]. Enemy recognition could be passed along a chain of at least six individuals, supporting the idea that one function of mobbing is the cultural transmission of knowledge about which animals are dangerous [Curio et al. 1978].

Signalling to Predators: Stotting, Warning Colours and Mimicry

Some antipredator signals are aimed not at companions but at the predator itself. When chased, Thomson's gazelles (Eudorcas thomsonii) sometimes stot, leaping with stiff legs in a conspicuous display [FitzGibbon & Fanshawe 1988]. Clare FitzGibbon and John Fanshawe found that gazelles were far more likely to stot in response to coursing predators such as African wild dogs (Lycaon pictus) than to stalking predators such as cheetahs (Acinonyx jubatus) [FitzGibbon & Fanshawe 1988]. During hunts, gazelles that wild dogs selected had stotted at lower rates than those they did not select. Gazelles that were chased but outran the dogs were more likely to have stotted, and stotted for longer, than those that were caught and killed [FitzGibbon & Fanshawe 1988]. In the dry season, when gazelles were probably in poorer condition, they were less likely to stot and stotted at lower rates. The authors suggested that stotting is an honest signal of a gazelle's ability to outrun predators, which coursing predators take into account when choosing whom to chase [FitzGibbon & Fanshawe 1988].

Defended prey often advertise their defences. Aposematism is the association of conspicuous warning signals, such as bright colours, with unprofitability to predators through toxins, distasteful chemicals or other defences [Mappes et al. 2005; Ruxton et al. 2018]. Johanna Mappes, Nicola Marples and John Endler noted that the theory of warning signals dates back to Alfred Russel Wallace yet remains confusing, controversial and complex [Mappes et al. 2005]. Because predators learn to avoid warningly coloured prey through experience, it is hard to explain how a conspicuous form could spread when it is initially rare and predators have not yet learned what it means. Mappes and colleagues argued that resolving this paradox requires attention to variation among predators, prey and the predation process [Mappes et al. 2005].

Warning signals open the door to mimicry. In Batesian mimicry, named after the naturalist Henry Walter Bates, a palatable species gains protection by resembling a defended one. In Müllerian mimicry, named after Fritz Müller, two or more defended species converge on a shared signal, so that predators learn to avoid all of them more quickly [Ruxton et al. 2018].

Hiding in Plain Sight: Camouflage

Where aposematic prey try to be noticed and remembered, cryptic prey try not to be noticed at all. Martin Stevens and Sami Merilaita define camouflage broadly as all strategies involved in concealment, including the prevention of both detection and recognition [Stevens & Merilaita 2009]. Within this, crypsis covers traits that reduce the risk of an animal being detected when it is potentially visible to an observer [Stevens & Merilaita 2009].

Several distinct mechanisms of crypsis are recognised. In background matching, an animal's colour, lightness and pattern generally match its surroundings. Disruptive colouration uses markings that create false edges and boundaries, hindering recognition of the animal's true outline and shape. Countershading, a darker upper surface and a lighter underside, may work by cancelling shadows or by obliterating the appearance of three-dimensional form [Stevens & Merilaita 2009]. Masquerade works differently: the animal may be seen, but it is not recognised as prey because it resembles an uninteresting object such as a leaf or a twig [Stevens & Merilaita 2009].

Camouflage has illustrated natural selection since Darwin and Wallace, yet Stevens and Merilaita observed that research on it had progressed more slowly than many other areas of biology, partly because earlier work relied on subjective human judgements of what looked concealed [Stevens & Merilaita 2009]. A central theme of modern camouflage research is that concealment must be assessed through the eyes of the relevant predator, whose visual system may differ considerably from our own [Stevens & Merilaita 2009; Ruxton et al. 2018].

Chemical Alarm Cues in Water

In aquatic environments, where vision is often limited, chemical information is a powerful source of warning. In 1938 the Austrian zoologist Karl von Frisch reported on the behaviour of shoals of European minnows (Phoxinus phoxinus), including a fright reaction shown by fish exposed to cues from an injured shoal-mate [von Frisch 1938]. The substance involved, later traced to the skin and detected by smell, became known by its German name, Schreckstoff ("fright substance") [Chivers & Smith 1998].

Douglas Chivers and Jan Smith reviewed chemical alarm signalling in aquatic predator–prey systems. They noted that such cues, released by prey animals when they detect or are attacked by a predator, are particularly common in aquatic systems [Chivers & Smith 1998]. The behavioural responses of prey to alarm cues resemble their responses to known predators and are therefore likely to reduce their vulnerability [Chivers & Smith 1998]. The benefits extend beyond the immediate reaction of nearby fish over a few minutes: alarm cues help animals learn to recognise unfamiliar predators and dangerous habitats, and have been implicated in induced morphological defences and shifts in life history [Chivers & Smith 1998]. In some systems, predators may have ways of counteracting the alarm signals of their prey [Chivers & Smith 1998].

The Landscape of Fear

If avoiding predators is costly, then predators should affect prey even when they kill few of them. The reintroduction of grey wolves (Canis lupus) to Yellowstone National Park in the winter of 1994–1995 offered a natural test [Laundré et al. 2001]. John Laundré, Lucina Hernández and Kelly Altendorf compared the vigilance of elk (Cervus canadensis, referred to in the paper as Cervus elaphus) and bison (Bison bison) in areas with and without wolves. Male elk and bison showed no response, remaining the least vigilant throughout the study, at about 12% and 7% of their time respectively [Laundré et al. 2001]. Females of both species were significantly more vigilant where wolves were present. Female elk with calves in wolf areas reached the highest levels, 47.5% of their time, by the second year and maintained them for the following three years [Laundré et al. 2001]. As wolves spread into previously wolf-free areas, female elk there gradually became more vigilant too [Laundré et al. 2001]. The authors suggested that these responses might have more far-reaching consequences for elk and bison ecology than the killing itself, and described the result as the re-establishment of a "landscape of fear" [Laundré et al. 2001].

Whether such fear effects cascade down to plants in Yellowstone remains debated. The idea that wolves are aiding the recovery of trembling aspen (Populus tremuloides) by causing elk to avoid risky places has been widely accepted and popularised, as discussed in this series' article on trophic cascades [Kauffman et al. 2010]. Matthew Kauffman, Jedediah Brodie and Erik Jules tested it by reconstructing aspen demography from tree rings and monitoring browsing in experimental elk exclosures placed across a gradient of wolf predation risk for three years [Kauffman et al. 2010]. They found that aspen were not recovering even in the presence of a large wolf population, and that the impact of elk browsing on aspen was not reduced where elk faced higher predation risk [Kauffman et al. 2010]. They noted that the strength of fear-driven cascades has been proposed to depend on predator hunting mode, and concluded that how such cascades are mediated by predator–prey life history and ecological context needs further evaluation [Kauffman et al. 2010]. Direct experimental evidence that fear alone can drive such effects in free-living wildlife was, for a long time, lacking [Suraci et al. 2016].

Experiments have since addressed that difficulty directly. Liana Zanette and colleagues protected nests of free-living song sparrows (Melospiza melodia) on the Gulf Islands of British Columbia from real predators, then used loudspeakers to broadcast either predator calls and sounds or non-predator sounds [Zanette et al. 2011]. The perception of predation risk alone reduced the number of offspring produced per year by 40%. Parents exposed to predator sounds laid fewer eggs and fed their young less [Zanette et al. 2011]. The authors concluded that perceived predation risk is itself powerful enough to affect wildlife population dynamics and deserves greater consideration in vertebrate conservation and management [Zanette et al. 2011].

A later experiment by Justin Suraci and colleagues, on several small Gulf Islands where wolves and cougars had been extirpated, tested whether fear of large carnivores could cascade through a food web [Suraci et al. 2016]. For a month at a time, they broadcast either the barks of domestic dogs, the only remaining large carnivore there, or the calls of harbour seals and Steller sea lions as a non-predator control, along stretches of shoreline used by raccoons (Procyon lotor) [Suraci et al. 2016]. During dog playbacks raccoons spent 66% less time foraging. After the month-long predator playbacks there were 97% more intertidal crabs, 81% more intertidal fish, 59% more polychaete worms and 61% more subtidal red rock crabs than after non-predator playbacks [Suraci et al. 2016]. The effects reached species that raccoons do not eat. The abundance of staghorn sculpins, a competitor of red rock crabs, declined, and the survival of periwinkle snails, which red rock crabs eat, fell [Suraci et al. 2016]. By restoring fear of large carnivores experimentally, the researchers reversed much of the raccoons' impact on the shore [Suraci et al. 2016].

Conservation and Welfare Relevance

Antipredator behaviour is not fixed. It develops through experience and can be lost when animals live without predators, either during their own lives or over many generations [Griffin et al. 2000]. Andrea Griffin, Daniel Blumstein and Christopher Evans noted that most animal reintroductions and translocations are unsuccessful and that predation is a principal cause of failure [Griffin et al. 2000]. Animals raised in captivity or drawn from predator-free populations may no longer express appropriate antipredator behaviour. The authors predicted that animals isolated from predators only within their own lifetimes should be easier to train than those isolated over evolutionary time, and that training should be more effective where a species already has effective responses to similar predators [Griffin et al. 2000]. They concluded that pre-release training, using classical conditioning in which model predators are paired with aversive events, can enhance the expression of existing antipredator behaviour.

Island species offer a natural illustration of evolutionary loss of wariness. Charles Darwin conjectured that animals on remote islands without mammalian predators are unusually tame. William Cooper, Alexander Pyron and Theodore Garland tested this with a phylogenetic comparative analysis of flight initiation distance, the distance at which prey begin to flee from an approaching threat, in 66 lizard species [Cooper et al. 2014]. Flight initiation distance was shorter in island than mainland populations and decreased with distance from the mainland, showing that island tameness is a real phenomenon in lizards [Cooper et al. 2014]. Such prey naivety can leave species highly vulnerable when predators they have not evolved with are introduced.

Many threatened populations suffer excessive predation because individuals cannot recognise, avoid or escape introduced predators with which they share no evolutionary history [Moseby et al. 2016]. Katherine Moseby, Daniel Blumstein and Mike Letnic argued that reintroduction programmes which rely solely on excluding introduced predators ignore, and may even worsen, the problem of prey naivety [Moseby et al. 2016]. They proposed exposing reintroduced prey to carefully controlled levels of introduced predators, so that in situ predation encourages both learned antipredator responses and natural selection for appropriate antipredator traits. They emphasised that this should be viewed as a long-term process [Moseby et al. 2016].

Antipredator theory also applies to people. Alejandro Frid and Lawrence Dill argued that nonlethal human disturbance, from a low-flying helicopter to a quiet wildlife photographer, creates trade-offs similar to those created by predation risk, because prey have evolved responses to generalised threatening stimuli such as loud noises and rapidly approaching objects [Frid & Dill 2002]. On this view, disturbance can affect vigilance, fleeing, habitat selection, mating displays and parental investment, and may indirectly affect populations through the energetic and lost-opportunity costs of avoiding perceived risk [Frid & Dill 2002]. The song sparrow experiment adds weight to this view, since perceived risk alone substantially reduced reproduction [Zanette et al. 2011]. For people who watch or photograph wildlife, an animal that stops feeding, becomes vigilant or moves away is paying a real cost, even when it is not harmed directly [Frid & Dill 2002].

References

[Chivers & Smith 1998]     Chivers, D.P. & Smith, R.J.F. (1998). Chemical alarm signalling in aquatic predator–prey     systems: a review and prospectus. Écoscience, 5(3), 338–352.     https://doi.org/10.1080/11956860.1998.11682471

[Cooper et al. 2014]     Cooper, W.E., Jr., Pyron, R.A. & Garland, T., Jr. (2014). Island tameness: living on islands     reduces flight initiation distance. Proceedings of the Royal Society B: Biological Sciences,     281(1777), 20133019.     https://doi.org/10.1098/rspb.2013.3019

[Curio et al. 1978]     Curio, E., Ernst, U. & Vieth, W. (1978). Cultural transmission of enemy recognition: one     function of mobbing. Science, 202(4370), 899–901.     https://doi.org/10.1126/science.202.4370.899

[Elgar 1989]     Elgar, M.A. (1989). Predator vigilance and group size in mammals and birds: a critical review     of the empirical evidence. Biological Reviews, 64(1), 13–33.     https://doi.org/10.1111/j.1469-185X.1989.tb00636.x

[FitzGibbon & Fanshawe 1988]     FitzGibbon, C.D. & Fanshawe, J.H. (1988). Stotting in Thomson's gazelles: an honest signal of     condition. Behavioral Ecology and Sociobiology, 23(2), 69–74.     https://doi.org/10.1007/BF00299889

[Frid & Dill 2002]     Frid, A. & Dill, L.M. (2002). Human-caused disturbance stimuli as a form of predation risk.     Conservation Ecology, 6(1), 11.     https://doi.org/10.5751/ES-00404-060111

[Griffin et al. 2000]     Griffin, A.S., Blumstein, D.T. & Evans, C.S. (2000). Training captive-bred or translocated     animals to avoid predators. Conservation Biology, 14(5), 1317–1326.     https://doi.org/10.1046/j.1523-1739.2000.99326.x

[Hamilton 1971]     Hamilton, W.D. (1971). Geometry for the selfish herd. Journal of Theoretical Biology, 31(2),     295–311.     https://doi.org/10.1016/0022-5193(71)90189-5

[Kauffman et al. 2010]     Kauffman, M.J., Brodie, J.F. & Jules, E.S. (2010). Are wolves saving Yellowstone's aspen? A     landscape-level test of a behaviorally mediated trophic cascade. Ecology, 91(9), 2742–2755.     https://doi.org/10.1890/09-1949.1

[Laundré et al. 2001]     Laundré, J.W., Hernández, L. & Altendorf, K.B. (2001). Wolves, elk, and bison: reestablishing     the "landscape of fear" in Yellowstone National Park, U.S.A. Canadian Journal of Zoology,     79(8), 1401–1409.     https://doi.org/10.1139/z01-094

[Lima & Dill 1990]     Lima, S.L. & Dill, L.M. (1990). Behavioral decisions made under the risk of predation: a review     and prospectus. Canadian Journal of Zoology, 68(4), 619–640.     https://doi.org/10.1139/z90-092

[Magrath et al. 2015]     Magrath, R.D., Haff, T.M., Fallow, P.M. & Radford, A.N. (2015). Eavesdropping on heterospecific     alarm calls: from mechanisms to consequences. Biological Reviews, 90(2), 560–586.     https://doi.org/10.1111/brv.12122

[Manser 2001]     Manser, M.B. (2001). The acoustic structure of suricates' alarm calls varies with predator type     and the level of response urgency. Proceedings of the Royal Society B: Biological Sciences,     268(1483), 2315–2324.     https://doi.org/10.1098/rspb.2001.1773

[Mappes et al. 2005]     Mappes, J., Marples, N. & Endler, J.A. (2005). The complex business of survival by aposematism.     Trends in Ecology & Evolution, 20(11), 598–603.     https://doi.org/10.1016/j.tree.2005.07.011

[Moseby et al. 2016]     Moseby, K.E., Blumstein, D.T. & Letnic, M. (2016). Harnessing natural selection to tackle the     problem of prey naïveté. Evolutionary Applications, 9(2), 334–343.     https://doi.org/10.1111/eva.12332

[Roberts 1996]     Roberts, G. (1996). Why individual vigilance declines as group size increases. Animal     Behaviour, 51(5), 1077–1086.     https://doi.org/10.1006/anbe.1996.0109

[Ruxton et al. 2018]     Ruxton, G.D., Allen, W.L., Sherratt, T.N. & Speed, M.P. (2018). Avoiding Attack: The     Evolutionary Ecology of Crypsis, Aposematism, and Mimicry (2nd ed.). Oxford: Oxford     University Press.     https://doi.org/10.1093/oso/9780199688678.001.0001

[Seyfarth et al. 1980]     Seyfarth, R.M., Cheney, D.L. & Marler, P. (1980). Monkey responses to three different alarm     calls: evidence of predator classification and semantic communication. Science, 210(4471),     801–803.     https://doi.org/10.1126/science.7433999

[Stevens & Merilaita 2009]     Stevens, M. & Merilaita, S. (2009). Animal camouflage: current issues and new perspectives.     Philosophical Transactions of the Royal Society B: Biological Sciences, 364(1516), 423–427.     https://doi.org/10.1098/rstb.2008.0217

[Suraci et al. 2016]     Suraci, J.P., Clinchy, M., Dill, L.M., Roberts, D. & Zanette, L.Y. (2016). Fear of large     carnivores causes a trophic cascade. Nature Communications, 7, 10698.     https://doi.org/10.1038/ncomms10698

[Suzuki et al. 2016]     Suzuki, T.N., Wheatcroft, D. & Griesser, M. (2016). Experimental evidence for compositional     syntax in bird calls. Nature Communications, 7, 10986.     https://doi.org/10.1038/ncomms10986

[von Frisch 1938]     von Frisch, K. (1938). Zur Psychologie des Fisch-Schwarmes. Die Naturwissenschaften, 26(37),     601–606.     https://doi.org/10.1007/BF01590598

[Zanette et al. 2011]     Zanette, L.Y., White, A.F., Allen, M.C. & Clinchy, M. (2011). Perceived predation risk reduces     the number of offspring songbirds produce per year. Science, 334(6061), 1398–1401.     https://doi.org/10.1126/science.1210908