You are on page 1of 10

Anim. Behav.

, 1996, 51, 1077–1086

Why individual vigilance declines as group size increases
Department of Zoology, University of Melbourne, Australia

(Received 27 March 1995; initial acceptance 23 June 1995;
final acceptance 8 September 1995; MS. number: 4887)

Abstract. A reduction in individual vigilance with an increase in group size is one of the most frequently
reported relationships in the study of animal behaviour. It has been argued that this phenomenon may
not be a direct consequence of an increase in group size but may be due to other factors relating to
increased group size, such as increased foraging competition. However, there is evidence for a direct
relationship between group size and vigilance where other variables have been controlled. The aim of
this review is to highlight the fact that the functional explanation of the group size effect remains
unclear. Some authors have considered just one hypothesis, the group vigilance or ‘many eyes’
hypothesis. This states that, by taking advantage of the vigilance of other group members, individuals
can reduce their own vigilance. However, there is an alternative, or additional, possibility that if
individual vigilance declines with a reduction in individual predation risk, the group size effect could be
accounted for by a reduction in individual risk at higher group sizes, as is widely thought to occur
through encounter, dilution and confusion effects. In this review, it is shown that evidence previously
interpreted in terms of one hypothesis may also be interpreted in terms of the other. Future research
should be directed towards explicit consideration of the two effects and empirical tests to distinguish
their relative importance. It is proposed that the individual risk hypothesis, with group vigilance as one
element, provides a more general framework for understanding variation in vigilance behaviour with
group size and with other factors. ? 1996 The Association for the Study of Animal Behaviour

A reduction in individual vigilance with increasing vigilance hypothesis’. It has also been referred to
group size is one of the most frequently reported as a ‘many eyes effect’ (e.g. Powell 1974), as a
relationships in the field of animal behaviour (see ‘collective detection effect’ (Lima 1995) and as
reviews in Barnard & Thompson 1985; Elgar a ‘detection effect’ (e.g. Dehn 1990, but not as in
1989; Lima 1990; Lima & Dill 1990; Quenette Lima 1995). It states that individuals in larger
1990). However, in spite of a large number of groups can enjoy the same or improved predator
published studies on the subject, the functional detection rate while scanning less frequently and
interpretation of this decline remains poorly having more time to feed (e.g. Pulliam 1973).
understood. I also examine another hypothesis. If vigilance
There are two main hypotheses to explain the depends on predation risk and if that risk declines
widespread existence of an inverse relationship with increasing group size, vigilance should also
between group size and vigilance (‘the group size decline with increasing group size. The possibility
effect’ e.g. Lima 1995). In a highly influential of such an effect has been considered by Bertram
paper, Pulliam (1973) advanced the hypothesis (1978), Pulliam et al. (1982), Packer & Abrams
that animals benefit by flocking because the (1990), Lima (1990), Dehn (1990) and McNamara
vigilance of flock-mates leads to an increase in the & Houston (1992).
probability of detecting a predator within the time There is a third possibility in the form of the
it takes to attack. I refer to this as the ‘group familiar caution that correlation does not imply
causation. Elgar (1989) concluded that ‘most
if not all studies fail to adequately demonstrate
Correspondence and present address: G. Roberts,
Department of Psychology, University of Newcastle,
an unambiguous relationship between vigilance
Newcastle upon Tyne NE1 7RU, U.K. (email: behaviour and group size’. A reduction in
GILBERT.ROBERTS@DURHAM.AC.UK) vigilance with increasing group size might arise

0003–3472/96/051077+10 $18.00/0 ? 1996 The Association for the Study of Animal Behaviour

studies are therefore consistent with the earlier It may be that this lack of progress is due to the work of. My aims in this paper are to highlight the fact ceived level of risk of attack may decline with the that there are alternative explanations of the passing of time at a site without the appearance of group size effect. less time on other activities such as vigilance. although Pöysä (1994) found that this was due to there remains a lack of understanding of the an effect of neighbour distance. to encourage explicit consideration of the by recording the vigilance of individuals in both two hypotheses in vigilance research. The object of this except Catterall et al. As initially alternative or additional explanation. while groups may build up through to show where it is open to alternative interpret- that time. spite of the large literature on vigilance behaviour. Thus. in individual vigilance with increasing group size. Sterna bergii. Lazarus apparent lack of awareness among authors (1979). Most recent selective factors involved in the group size effect. Parker & dilution of risk may also be an important factor in Hammerstein 1985. although but the model may be rearranged to predict widely cited as a general review of relationships scanning frequency in terms of group size (e. It will remain a possibility that increased group Catterall et al.1078 Animal Behaviour. the hypothesis that increased group concerns by measuring or controlling some or size provides better detection will remain untested. 1982. stood as developments of this model (Pulliam tions of the group vigilance hypothesis because a et al. all of these factors (Burger & Gochfeld 1992. (1992) found a reduction in review is to focus attention on the fact that. All the pressure to be vigilant. For example. One study present some ideas as to how group vigilance and which does this is Roberts (1995) where individual predation risk effects may be distinguished. described the probability of at least one member Given this lack of progress. Pöysä 1994. and to increasing and decreasing groups. In a recent investigation of the group group size remains influential and most later vigilance advantages of flocking. responded to flock arrivals by increasing their inter-scan inter- vals and to flock departures by decreasing their TESTS OF THE PULLIAM (1973) inter-scan intervals. the model gave expected probabilities It is particularly pertinent to ask this question of group detection given individual vigilance levels given that the review by Elgar (1989). group size and predator hypothesis is an adequate explanation for a reduc. for example. to evaluate critically the evidence a predator. models of vigilance behaviour can be under- cluded that his results may not have fit the predic. Packer & Abrams 1990. 51. Caraco (1979). 5 if group size relates to some other factor which in between group size and vigilance. 1989. such as Elgar (1989) that there are alternative An additional potentially confounding factor functional explanations. No consideration was given to other animals feeding on better food supplies may spend hypotheses such as the individual risk hypothesis. Roberts 1995). even if all workers direct their research so as Other potentially confounding effects include dis. This echoed Lazarus’ (1979) 1989. age. interpreted the turn affects vigilance. McNamara & comment that part of the group size effect may be Houston 1992). it is timely and of a group detecting a predator in terms of pertinent to ask whether the group vigilance individual vigilance rates. Recent studies have addressed these flock density. Lima (1995) con. This possibility may be eliminated ations. sex and observer proximity Elgar (1989) and others such as time at a site and (Elgar 1989). larger groups evidence solely in terms of the group vigilance may tend to feed on better food supplies and hypothesis. MODEL If it is accepted that there is evidence for a direct relationship between group size and vigilance. 1992. As originally formulated it due to a reduced risk of capture in larger groups. Lazarus & Symonds 1992. presented. . Pulliam’s (1973) model of how group detection of then there remains the question of its functional a predator may be expected to vary in relation to explanation. to control for the confounding effects listed by tance from cover. Lima the group size effect. Hart & Lendrem 1984. approach time. size does not provide better detection but reduces Cresswell 1994. tion in vigilance with increasing group size and Several studies have tested the Pulliam (1973) whether the individual risk hypothesis provides an model (see Elgar 1989 for a review).g. preening crested terns. Kaitala et al. Barnard (1980) and Bertram (1980). not considered by Elgar (1989) is that the per.

The probability of an found by Catterall et al. of group (or ‘corporate’) vigilance. however.g. as increase in group size. there are numerous observations of individual vigilance of silvereyes. Predator attack rate depends on an in part from an increase in group detection and in ‘encounter effect’ (e. Each of these may extreme they might maintain their levels of vigi. Powell 1974. ever. (1992) concluded that their results may have predator to make its final approach and the provided support for Pulliam’s (1973) model for probability that at least one member of the group another reason: the lack of predators on the will detect a predator are assumed to remain island study site. even if the Pulliam (1973) model were mate fit to a curve representing the reciprocal of to be rejected it would not necessarily entail group size. it is clear that difficult to falsify. predation risk (e. (1992) which increase with the risk of predation. such that they benefit Davies 1993). range of ways in which individuals might trade off and the probability of escaping given the attack detection and investment in vigilance. given that a group is attacked . individual awareness of the vigilance of others. with advantages arising through accepted that a reduction in predation risk is an increased foraging time. Gluck 1987. including the lack of a relationship. However. Returning to the 1990). Inman & Krebs 1987). scan rates were lower than rejection of the concept of group vigilance (Lima predicted for small group sizes. Krebs & vigilance levels rather less. In fact. At one and detection probabilities. How- size. Other problems with the Pulliam (1973) model ing an inverse relationship between group size and have been discussed elsewhere (e. Or. It is generally tion is a constant. vigilance. Lima 1990). If both the time for a et al. and competition for food.g. benefiting from the between group size and predator detection above. then the scan rate between group size and vigilance is the null should equal the reciprocal of group size multi. vary with group size. At But vigilance may depend not only on group the other extreme they might reduce their levels of detection probability but also on other aspects of vigilance such that the probability of group detec. a fit could be found to the model PREDATION RISK by relaxing the initial assumption that group detection remained constant. Caraco et al. In accordance reported no relationship between group size and with this. predation. That is to say. Thus.g.g. I discussed the relationship lance regardless of group size. 1980. original form of the model. the probability of detect- vigilance should decline with group size. In fact. Turner & Pitcher 1986. sons why the Pulliam model might fail to hold in Lendrem 1983. the model has been widely interpreted as predict. be interpreted as falsifying the Pulliam (1973) An individual’s risk of predation depends on model because the model does not state that the predator attack rate. But allowing both Vigilance may have a number of functions in sides of the equation (detection and individual obtaining information about the environment vigilance) to vary with group size makes the model (e. Zosterops increases in vigilance on or after exposure to a lateralis. any of a range of relationships between be more likely to be detected by a predator but group size and vigilance may be consistent with this increase is unlikely to keep pace with an the model. Dehn 1990). hypothesis. predator (e. as no relationship constant across group size. Nevertheless.g. Thus. it cannot be concluded that the plied by a factor depending on group detection lack of a relationship is due to the absence of and approach time (Elgar & Catterall 1981). one would expect vigilance to by the study of Catterall et al. part from a decrease in time invested in vigilance. Elgar & Catterall (1981) used observed scan rates to calculate that the probability of group detection increased with VIGILANCE IN RELATION TO group size. Catterall individual escaping. Lima 1990). that is. given a constant level These include the perception of group size. (1992). The authors considered a series of rea. Sullivan 1984. increase in the probability of group detection. Roberts: Vigilance and group size 1079 Elgar & Catterall 1981). they may reduce their important benefit of grouping (e. There is a ing a predator given that the group is attacked. predator detection is a major function in many The problem of falsification is also highlighted species. As such. Elgar & Catterall (1981) found an approxi.g. the object vigilance may be decreased with increasing group of vigilance. their result cannot 1987). larger groups may Thus. Pöysä the case of their study.

Evidence in these Lima (1990). in a larger group an individual has a lower making them difficult to test. However. these studies consider detection not test such models. there are multiple probability of detection within a certain time . Lima 1987b). That is. ocean skaters. Lazarus 1979. Inman & Krebs 1987). this presup- ing both detection and dilution effects explains poses how an animal scales the value of increased more of the variance in the group size–vigilance feeding time against the value of increased pred- relationship of foraging elk. Tringa totanus (Cresswell 1994). However. The combined effects have been referred to as an encounter-dilution effect GROUP SIZE AND PREDATOR (Mooring & Hart 1992). the number of studies providing in predation risk in larger groups. where N represents group size. predator. than the number demonstrating a reduction in Reduced predation risk as a factor affecting individual vigilance with increasing group size. late detection of aerial predators. van individual vigilance may be lower in larger groups Schaik et al. strating improved detection. Pulliam et al. (2) when another animal detects a then the probability of detecting each of 10 pred- predator and (3) when the animal itself detects a ators decreases considerably. depends on the ‘dilution effect’. A dilution effect predicts that vigilance should Other studies have used the Pulliam (1973) be dependent upon the reciprocal of group size. Although Pulliam (1973) was largely concerned Halobates robustus (Foster & Treherne 1981) and with showing how increased group size could lead redshanks. vigilance has been considered by Bertram (1978). the effect declines as group size fore provides some evidence of the importance of increases. Even in those studies where detec- does a solely detection based model and there. on the basis of tion. or ‘early warning’ encounter and dilution effects lead to a reduction (Lazarus 1979). from 0. A model incorporat. Danaus plexippus (Calvert et al. the two hypotheses are difficult to distin. dependencies on group size in such models.g. observed individual vigilance levels. Perhaps the most vigilance to predation risk is that vigilance might compelling evidence comes from Cresswell (1994) actually be reduced in order to lessen the feeding who found that redshanks in small groups time when the animals are most exposed to suffered high mortality. further by incorporating separate terms for the However.1080 Animal Behaviour.60 to 0. The differ. Further empirical work is required to Furthermore.g. tion does increase. apparently as a result of predators (Lima 1987a). would gain no advantage in detection. Packer & Abrams (1990). if the probability of detecting a pred- probability of death (1) when no animal detects ator decreases only a little. (1982). Kenward 1978. model to calculate the probability of a group As this prediction is similar to that of the member detecting a predator while it makes Pulliam (1973) model with constant group detec. This may reflect the relative difficulty of demon- Pulliam et al. Evidence for such effects DETECTION has been found in monarch butterflies. then the group evidence of an increase in predator detection with size effect could be explained by postulating that increased group size (Powell 1974. Turner & Pitcher 1986. Dehn (1990) suggested that of studies the probability of detection increases group vigilance (or ‘detection’) should have a with group size. Cervus elaphus.31. 1982. Underhill 1975.90. If to improved predator detection. However.95 to 0. 51. say from 0. Dehn (1990) and McNamara & studies comes from earlier detection or an increase Houston (1992). effect. Cresswell 1994) is much smaller because individual predation risk is lower. even where individual vigilance 1/N effect while dilution should have a ln(N)/N declines (e. Elgar & Catterall (1981) found the calculated ence arises from the assumption that dilution is increased probability of detecting a predator to be relevant only if detection fails. Siegfried & vigilance increases with predation risk. chance of being taken (e. That is. than ator detection. An extreme example is Pöysä’s (1987) both effects (Dehn 1990). A complicating factor in relating in the probability of detection. finding that groups of more than two teal Anas McNamara & Houston (1992) took the model crecca. its final uncovered approach. as McNamara & as how soon the predator is detected but as the Houston (1992) pointed out. In a number guish. a predator. 5 by a predator. 1983. an assumption small and concluded that the major benefit from whose validity will depend on the particular joining a group came through a reduction in time predator–prey situation. 1979). spent in predator vigilance.

it is possible other group members cannot be valued as highly that at high group sizes vigilance is responding not as an individual’s own vigilance. shanks increased their pecking rates as vigilance Therefore. That is. Most CONSEQUENCES OF REDUCED studies of vigilance have found an increase in INDIVIDUAL VIGILANCE feeding time as vigilance declines. The value of an individual’s own sizes independent of the detection rate calcu. Such mission has received support from a number of results suggest that an increase in time for other studies (e. few studies transmission’ of the detection (Lazarus 1979). more time becomes available for other activities. Gazella thomsoni. there may be other advantages in reduced . demonstrated by FitzGibbon (1989): cheetahs. yet the main advantage of grouping required in order to elicit a flight response from may in fact be increased predator detection or non-detectors (Lima 1994a). ing such a reduction in wild boar. However. if vigilance is tages associated with individual vigilance in that costly in terms other than time lost from feed- sparrows which were vigilant at the time of an ing. reduced predation risk. Acinonyx jubatus. behaviour of little or no value. However. Pulliam 1973). there is evidence of advan. declined. Lima If it is advantageous not just to detect a predator (1994b) has shown that vigilant birds which could within a certain time. individual vigilance. An advantage of being vigilant has also been Quenette & Gerard (1992) were unusual in find. The lower vigilance of Related to the idea of collective detection is the animals in large groups may simply reflect a lower assumption that an individual values the vigilance need to be vigilant because of reduced predation of other group members as equivalent to its own risk and thus the animals do not perform a vigilance. that were less vigilant. Or. with their heads down (Elgar et al. there is less need to look out for such findings are inconsistent with the model. A have measured whether there is an increase in common observation is that if one member of a intake resulting directly from a reduction in group detects a predator. Roberts: Vigilance and group size 1081 period (corresponding to the time during which alarm stimulus responded more quickly than birds the predator makes its final uncovered approach). only one individual need detect a pred. their intake did not increase: they simply ator for all to benefit. chased those Thomson’s Cresswell (1994) also found a reduction at very gazelles. Cresswell (1994) found that.g. Although these authors did not Such studies demonstrate that the vigilance of interpret these results in such terms. 1988). it a predator. although risk: if risk is low. as is assumed by to the probability of detection but to predation the Pulliam (1973) model. However. while red- and the whole group is alerted (e. Lima (1995) found vigilant. even without explain a reduction in individual vigilance. Godin et al. its activities may not be an important advantage of generality has recently been questioned. Thus. vigilance relative to the vigilance of another group lated from observed individual vigilance. an alarm call is given vigilance. it detecting the source of danger. was advantageous. The idea of social trans.g. increased the number of unsuccessful pecks. However. at least in species ance or predation risk effects. but also to detect the not have detected an approaching ball never- predator as soon as possible. As individual vigilance declines. This is because in most studies feeding and vigilance are mutually The Pulliam (1973) model depends on what Lima exclusive activities and account for a large pro- (1995) refers to as ‘collective detection’. What member will be an important factor determin- appears to be lower detection may be due to lower ing the shape of any group size and vigilance perceived risk and hence delayed response. then it is more likely theless flushed more quickly than non-vigilant that larger groups will continue to be beneficial. multiple detections may be advantages. it is difficult to provide a remains possible that there is a more limited group demonstration of lower detection at higher group vigilance effect. whether through group vigil- detection. many studies concluded that limited evidence for social transmission of the reduced vigilance. relationship (McNamara & Houston 1992). 1986). birds in response to the alarm flight of a detecting While the concept of group vigilance can bird. In a series reduced vigilance and that this increased time is of experiments involving ‘ball attacks’ which only simply a by-product of a reduced need to be one individual could detect. high group sizes. Sus scrofa. However. brings foraging without alarm calls. cannot explain a reduction in corporate vigilance. It appears that. the ‘social portion of the time budget.

it is fundamental to the models of availability of vigilance mates. 51. He concluded that the basis of the number of birds within a certain redshanks benefit from grouping primarily proximity. He concluded on vigilance is also consistent with a ‘selfish herd’ that both group vigilance and dilution effects were (Hamilton 1971) explanation whereby animals are important. they may attack birds in small groups the effects of both group size and density were preferentially. suggests that the reduction in group density on vigilance (Lazarus 1978. Further- 1984a. In fact. the two hypotheses are not mutually exclusive GROUP DENSITY AND POSITIONAL alternatives. That is.g. If animals do respond to group through the dilution effect (or. Dehn (1990) found a good fit to a model in tected a predator is easier to obtain from nearer which non-vigilant individuals were considered to individuals (Pöysä 1994). From an observer’s point of pretation. Lazarus 1978. Roberts 1988. group vigilance being particularly less at risk when density around them is greater. sizes. The way in which benefits associated Jennings & Evans 1980. whereby animals on the edge of effect continuing to be important at higher group a group are more vigilant (e. This leads to difficulties of inter- and difficult to define. individual risk of attack bird rather than absolute group size (e. Cresswell (1994) found view this can lead to measurement of the number that although larger groups of redshanks were of birds within a certain radius of the focal preferentially attacked. which are based on time spent feeding as a currency in which costs of vigilance are measured. Thus. from the animal’s own per. Pöysä (1994) found that group risk associated with grouping may not come size did not relate to vigilance when neighbour through dilution per se but through the vigilance distance was controlled for.g. 1984. more. Metcalfe decreased with increasing group size. But a density effect dilution but not to group vigilance. the probability of being captured declined spective. nearer individuals contribute to lowering individual risk through make better ‘vigilance mates’. 5 vigilance. Such a density effect is understandable in terms In perhaps the only empirical study to attempt of the group vigilance hypothesis: information to distinguish group vigilance and individual risk about whether other group members have de. Holmes 1984. explanation and an explanation based on the Furthermore.1082 Animal Behaviour. Pöysä group vigilance effect. density. size. then this might be either because they are through a combination of encounter and dilution using group density as a convenient indicator of effects). the finding that predators were group size. In a recent study in which groups. more precisely. tinguish from the predation risk hypothesis because the two make similar predictions. However. b) Similarly. late or no detection was the cause of many of Few studies have investigated the effects of these captures. effects. It appears logical that DISTINGUISHING GROUP there should be a trade-off between activities VIGILANCE AND PREDATION RISK but in practice it is not clear that this trade-off EFFECTS provides a useful basis for predicting behaviour. if predators are 1994). Packer & Abrams (1990) and Lima (1990) . the reduction in predation considered. behavioural decisions may be made on with increasing group size. Elgar capture probability may have been due to a et al. Inglis & Lazarus 1981) with reduced vigilance are manifested is important are also consistent with both a selfish herd in understanding the selection pressures involved. Detecting a predator is just one EFFECTS element of avoiding predation and so the group vigilance effect can be seen as one aspect of Groups of animals are often irregularly spaced predation risk. whereas neighbour benefits of large groups putting predators off distance did have an effect after controlling for attacking them at a rate proportional to group group size. or because group density is actually more successful in attacking small groups and that of more functional significance than group size. For example. more successful in attacking birds in smaller bour distance increased. important at small group sizes and a dilution Edge effects. This may be because animals are not always The group vigilance hypothesis is difficult to dis- time-limited. McNamara & Houston (1992). Thus. All found that vigilance increased as neigh.

It could be hypothesized that should respond in a Tat-for-Tat manner to the vigilance in the presence of a predator should not behaviour of others (Pulliam et al. If so. 1982). Alvarez considered problems in testing vigilance models. the vigilance effect should reduce individual vigilance reduction in vigilance with increasing flock size with increasing group size.g. If the two types risk. that animals referred to above. it would provide an interesting benefit from the vigilance of others. According to the individual risk hypothesis. Roberts: Vigilance and group size 1083 also showed theoretically that dilution of risk with angled towards a predator. several species fits may provide only a weak test. vigilance hypothesis that advantages arise from ation is diluted. roosting. an alternative test of the the vigilance of other group members as opposed advantages of group vigilance would be to carry to their mere presence. this is. Thus. while qualitative detected a predator. tests relying on quan. As the risk that a member of the group according to the vigilance of other group will be taken increases with the appearance of a members. Birds standing by the group vigilance effect: once all members of vigilant may be expected to contribute more to the group have become aware of the danger and any group vigilance effect than roosting birds so. a distinction vigilance hypothesis. that is. the of vigilance can be distinguished observationally. preen- former may be expected to respond to group size ing. One problem is the possibly confounding One method for distinguishing the two hypoth- effect that at larger group sizes individuals may eses depends on the inference from the group respond more slowly because their risk of pred. standing vigilant). Thus. Only the performing different activities (feeding. If birds ation risk. For example. achieve if vigilance serves functions additional Some species show signs of apparently having to that of detecting predators. less vigilant members may contribute more to a for example by distinguishing head movements reduction in risk of the other birds because they . However. This study is therefore consistent with predator. A recent study involving birds that were individual vigilance with increasing group size. the reduction in ant. Although he was actually exposure to. As discussed above. then speed or between how vigilant an animal is and how great probability of predator detection should increase is the risk it perceives. food-deprived and therefore less vigilant than Higher vigilance in the presence of. of a predator should decline with group size all flock members contribute to a reduction in according to a dilution effect. then the relationships increasing group size can account for much of the of the two types of vigilance with group size could group size effect. Thus. whether they are vigilant or not. for modifying vigilance group size. In fact. as embodied model system for investigating the relationship in the group vigilance hypothesis. on the basis of which flocks of different sizes detect a stimulus the group vigilance hypothesis. However. Such evidence dicted that a focal individual should not reduce is important because it indicates a group vigilance its vigilance in line with the number of other advantage more unambiguously than does the group members if those individuals are not vigil- opposite side of the coin. or after others is relevant here. this has been risk and vigilance is fundamental to the hypoth- tested before. it may be pre- indicating the availability of food. However. a predator is a common observation testing a contrasting prediction. Tail There is a clear need for studies that distinguish flicking may be an indicator of perceived pred- group vigilance and predation risk effects. Thomson’s gazelles stot (Caro 1986). Lima (1990) of rails (Rallidae) flick their tails (e. as is the case in the out an experiment to measure the speed with individual risk hypothesis. it is less clear that a group according to the group vigilance hypothesis. be investigated in the presence and absence of a titative fits to models may be impossible to predator. it would be should be less where more of the flock are roost- predicted that individual vigilance in the presence ing. detection may become less important the individual risk hypothesis and not the group relative to dilution of risk. This relationship between with flock size. more studies are needed esis that a decline in vigilance may result from a in order to determine how widespread an effect decline in risk with increasing group size. are more vigilant. might be made between vigilance at some back. Lima only be greater than in the absence of a predator (1995) found no evidence for ‘behavioural but may not show the same relationship with monitoring’. A related test would be to study birds that had ground level and vigilance that is a response to different levels of vigilance because they were something that has been observed. 1993).

J. McGowan & Woolfenden 1989).1084 Animal Behaviour. S. The signifi. 1992. 60. M. 1980. T. 1980.. T. A. Behav. Flocking is an effective anti- total group size. P. they should be of equal value Bertram. of others is where some individuals act as sentinels Caraco. vigilance behaviour through an individual’s risk ual vigilance than the presence of the less vigilant of predation. R. number of birds between it and the predator than Calvert. individuals between the focal individual pp. R. C. Vigilance and group size in diluting a focal individual’s risk. Behav. Anim. A. no value in group vigilance because they are in positions from which they would be unable to REFERENCES detect a predator. 64–96. If the Barnard. 1978. 1986.g. detect the predator. and the predator should be of greater value in Bertram. 44. The functions of stotting in Thomson’s gazelles: some tests of the predictions.. Time budgeting and group size: a test (e. 46. Mortality of the monarch butterfuly (Danaus An extreme case of benefiting from the vigilance plexippus L. If non-sentinels retain some vig. 97. Krebs & N. Science. cance of sentinel behaviour in this discussion is Caraco. 1979. to the total group size. 204. their vigilance rather than simply through diluting Anim. 1979. it is possible to imagine situations Barnard. M. Anim. B. H. 3. Effect of group-size on vigilance while drinking in the coati.. testable predictions. J. detection hypothesis. Behav. were known to have consistently higher vigilance Future work should develop a general framework than others: according to the group vigilance in which group vigilance is considered as one hypothesis. Thus. & Thompson. important are advantages obtained through A similar test could be carried out in situations shared vigilance as a factor influencing individual where certain classes of individuals (e. 5 may be slower to respond and so be more likely Despite the considerable research effort into to be taken by the predator. could be tested by determining whether a focal Burger. 663–684.. 47. 34. associating with sentinels specifically because of Caro. 847–851. R. Hendrick. in individual’s vigilance was more responsive to the Costa Rica. individual risk. diluting other individuals’ risk of attack. M. 51.. However. 207–210. as would be predicted on the predation strategy in redshanks. Avian that here it is clear that advantages arise from time budgets and distance to cover. W. B. ilance. & Brower. 28. 278–286. Oxford: Blackwell Scientific Publications. B. & Pulliam. 1980. 1053–1057. to the total number of animals present.. C. C. Behav. 1992. Behav. R. Tringa totanus. T. J. & Gochfeld. Anim. an increase in the number of sentinels rather than Cresswell. Flock feeding and time budgets in where individuals vary in their contribution to the house sparrow (Passer domesticus L. Plovers: The Ecology of Mixed Species Feeding then provided all individuals are in a position to Groups. B. 618–627. one would predict by the group vigilance Vigilance does not covary with group size in an island hypothesis that this vigilance should decline with population of silvereyes (Zosterops lateralis). Ecol. 1229–1231.. 1985. vigilance should be reduced according Alvarez. W. for support and for providing departmental and not according to the number that have facilities. Behav. F. Martindale. Living in groups: predators and prey. Another way in which individuals may differ in their value as vigilance mates is if some animals ACKNOWLEDGMENTS cannot contribute to group vigilance because their vision is obscured. a prediction that ostriches. Ecology.. C.): avian predation at five overwintering sites in Mexico.). Conversely. Catterall. L. of a theory. P. Christopher Barnard and Neil Metcalfe. H. Davies).. Anim. J. Elgar. basis of a simple dilution of risk. Auk. class. In: Behavioural Ecology: An Evolutionary in reducing the vigilance of some focal individual. London: Croom Helm. Anim. L. 28. animals in an exposed I acknowledge the support of a Royal Society position should reduce their vigilance according to Fellowship and thank Professor Marilyn Renfree the number of other animals in exposed positions. Nasua narica. Behav. Approach (Ed. for example by having a I greatly appreciate the comments of John partially roofed feeding station. 433–442. C. 1993.g. Gulls and predominant risk to a group comes from one side. D. it remains to be seen how hypotheses make opposing.. . the presence of the more vigilant class of a number of factors affecting individual should have a greater impact in reducing individ. 295–309. Alertness signalling in two rail species. & Kikkawa. 872–875. 1994. E. males) vigilance and as a factor affecting gregariousness. According to the individual risk hypothesis. the two vigilance behaviour. According to the Lazarus. by J.

329–345.. Behav. Behav.. & Houston. 48. 921–928. E. H. M. Lond. in brent geese: the edge effect. Krebs. L. dilution effect and the selfish herd from fish predation J. Behav. risk. domesticus L. Theor. Burren. L. vigilance: the group size effect. 38. Behavioral decisions Group vigilance and shoal size in a small characin made under the risk of predation: a review and fish. B. vigilance and predator avoidance in a single bird. C. M. B. H. 48... J. 1990. J. interpretation of vigilance. R. Boulder. Behaviour. Slatkin). 173–191. E. Inglis. J. Behav. M. 37. Hawks and doves: attack success and animal behaviour. Biol. 1987. Sturnus vulgaris. J. 25. 142.. 1993. Harvey & M. Experimental analysis of the Scientific Publications. Anim. 1994a. N. & Posen. Game theory Kenward. 1989. Behav. 124. 31–32. 828–838. J. & Woolfenden. 64. hypothesis. K. 31. Iterated prisoner’s dilemma: an and perception of flock size in foraging house approach to evolutionarily stable cooperation. & Lendrem. W. Vigilance for predators: detection Lendrem. Mooring. 1974.. 102. The influence of models on the Elgar. J. 104.. stable levels of vigilance as a function of group size.. Lima. M. & Abrams. Group foraging. A. J. Anim. theor.. A. J. 22. Predation and group trust? Anim. 295–311. I. E. 868–872. D. Vigilance and flock size Behav. 1995. Collective detection of predatory Foster. behaviour of the hoary marmot in Alaska. 1989. J. system in the florida scrub jay. 501–505. theor. Hart. 981–985. 1987b. 73–94. Jamieson). G. M. Vigilance. 43. M. 337– the blue tit (Parus caeruleus). Behav. Metcalfe. 342. distance to cover and vigil- Lazarus. B.. Behav. The early warning function of flocking (Anas crecca): effects of feeding method and predation in birds: an experimental study with captive quelea. 1987a. Anim. Ecol. Contrasting effects of ance in the teal. Behav. 108–122. Lima. evidence. 29–40. J. Avian Biol. J. 1973. 134. flocking on the vigilance of shorebirds: who do they Inman. prospectus. Z. W. Behav. flock size and domain in relation to predation and foraging. Anim. Anim. H. A. S. Godin. Oxford: Blackwell Powell.. 1892–1894. Nature. Behaviour. W. The effects of mixed-species 193–200. sparrows (Passer domesticus L. L. McGowan. Biol. by M. Anim. G. Behaviour. and vigilance in house sparrows. On the advantages of flocking. Feeding–vigilance trade-off in the teal Lazarus. Behav.. Anim. protective and obstructive cover on avian vigilance. 1981. Behaviour. Pöysä. 1984. Vigilance Lima. 1989. J. 1978. 1990. A sentinel J. 1989. Geometry for the selfish herd. & Krebs. scanning patterns in birds. J. 37. R. 519–521.. McKay. Rev. 90. C.. D. V. G. The effects of visibility on the Ecol. E. Evolution and Adaptation (Ed. S.. Parker. Elgar. L. & Evans. L. D. J. Influence of position ing for protection from parasites: selfish herd and in the flock and flock size on vigilance in the starling. Behav. W. theor. N. 32. Behav. Anim. L. 26. 3rd edn.. 34. Should co-operative Kaitala. Elgar. A. study of ESS. groups be more vigilant than selfish groups? J. In: Evolution: Essays in and selection in goshawk flights at wood-pigeons. 1984. 1983. W. 11–20. P. Hamilton. On the personal benefits of anti- Gluck. 215–223. vigilance of shorebirds: is visibility important? Anim. 1987.. Honour of John Maynard Smith (Ed. 341–357. M. & Catterall. on a marine insect. L. 1992. Behaviour. In: Interpretation and pecking and alarm flights in house sparrows (Passer Explanation in the Study of Animal Behaviour: Vol. Evol. 1987. 27. & Lazarus. Am. 135–145. & Dill. Pulliam. A. 123.). 1984. 466–467. Tierpsychol. & Hart. 49. 1216– McNamara. A. Lima. 1990. Trends Ecol. Behav. Sociobiol. vigilance and risk predation in birds: a dynamic game Biol. visual obstruc- predator surveillance in house sparrows: test of an tions. Distance to cover. 47. 508–510.. Ecol. 1980. social value of flocking by starlings (Sturnus vulgaris) Lazarus. G. Vigilance and 1000–1006.. Animal group- Jennings.. L. Evidence for the attack by birds in the absence of alarm signals. Biol. 43. P.. S. V. 1979. Anim. Behav. Lima. Classon. S. 1992. Anim. N. 449–460. 138. & Woon. predatory vigilance. Lima.). Pöysä.. 14.. Nat. Lindstrom.. 32. R. 246–267. Holmes. 103. Predation risk and foraging Anim. Anim. 57. 1984a. theor. Wildfowl. Anas crecca. 634–635. & Abrahams. E. C. 1990. 419–422. S. . 1986. Ecol. R. Sociobiol. 1988. Anim. A. S. FitzGibbon. S.. Greenwood. hunted by cheetahs. Sociobiol. An Introduction New York: Cambridge University Press. M. living. Back to the basics of anti-predatory Oecologia (Berl. H. D. 29. L. Roberts: Vigilance and group size 1085 Dehn. P. An experimental study of feeding. 9–13. Biol. I. 1985. 268–272. M. 293–301. A... Foraging. of danger in the white-fronted goose. & Hammerstein.. 15. M. A cost to individuals with Bekoff & D.. 1971. by P. Behav. 303–316. H. R. S. N. Flocking and Lima. 1981. pp. 2. pp. M. 1984b. V. L. M. Behav. 986–993. J. P. Scanning.. & Treherne. Packer. B. 1994b. M. 32. Elgar.. Anim. Zool.). T. 1981.-G. A. J. encounter-dilution effects. 29.. Predator vigilance and group size in Lima. Evolutionarily 1224. L. 855–865. Can. 1989. L. P. 1994. 13–33. 231–238. & Symonds. 1992. & Ranta. reduced vigilance in groups of Thomson’s gazelles Colorado: Westview Press. to Behavioural Ecology. J. & Davies. Behav. 619–640. Explanation. Metcalfe. 28. P. 2.. 319–326. 1978. S.. 641–658. Biol. 71. Vigilance while feeding and its mammals and birds: a critical review of the empirical relation to the risk of predation. J. 293. Predation risk and vigilance in and dilution effects. S. 734–736. J. 68.

G. Biol. 1990.1086 Animal Behaviour. C. mammals: a review. The advantages of social for- Roberts.. B. G. . Party size and early detection of approach. Flocking as an Quenette. 11. 128. Warsono. 905–913.. 1995. Zool. Am. Quenette. 1992. Can. scanning behaviour of juncos: a game-theoretical & Sutriono. 1986. Pyke. 1975. 51. Acta oecol... R. C. 32. Behav. The van Schaik. 24. P. Attack abatement: a Roberts..-Y. E. & Caraco. T. J.-F. 1632–1635. T. P. & Pitcher. Siegfried.. Social influences on vigilance in model for group protection by combined avoidance rabbits. 70. 504–508.-Y. predators in Sumatran forest primates. 1983. 1371–1374. H. K. Turner. J. 89–103. & Gerard. J. F. 5 Pulliam. and dilution. 1984. J. 23. Nat. From individual anti-predator strategy in doves. Anim. Functions of vigilance in 211–221. A. W. 228–240. 36. A real-time response of vigilance aging in downy woodpeckers. Behav. P. 16–22. R. Anim. M. Anim. & Underhill.. Primates. 95. L. G. Behav.. Anim. Sullivan.. van Noordwijk. 50. S.. 1982. G. 801–818. H. theor. 1988.. behaviour to changes in group size. A. Behav. to collective vigilance in wild boar (Sus scrofa).