j.1365-2656.2010.01775.x

  • Upload
    eulier

  • View
    215

  • Download
    0

Embed Size (px)

Citation preview

  • 8/13/2019 j.1365-2656.2010.01775.x

    1/10

    Predicting the optimal preygroupsize from predator

    huntingbehaviour

    WillCresswell

    1

    * and JohnL. Quinn

    2

    1School of Biology, University of St. Andrews, Bute Building, St Andrews, Fife KY16 9TS, UK; and2Edward Grey Institute,

    Department of Zoology, University of Oxford, Oxford OX1 3PS, UK

    Summary

    1. How group size affects predator attack and success rate, and so prey vulnerability, is important

    in determining the nonlethal consequences of predation risk on animal populations and communi-

    ties. Theory predicts that both predator attack success rate and the dilution effect decline exponen-

    tially with group size and that selection generates optimal group sizes at a risk threshold above

    which antipredation benefits are outweighed by costs, such as those owing to higher attack rates.2. We examined whether flock size risk thresholds for attack rate, success rate or dilution differed,

    and therefore whether the strength of selection for group size differed for these three factors, using

    a system of redshank Tringa totanus flocks being hunted by Eurasian sparrowhawks Accipiter

    nisus. We also asked which of the three thresholds, on their own or in combination, predicted the

    most commonly observed group size.

    3. Mean redshank flock size increased with a very gradual quadratic function (i.e. approximately

    linearly) with population size, although at a rate half that possible; when population size was not

    limiting, individuals almost always avoided flocks of less than 30 and birds were frequently in

    flocks up to at least 80. Sparrowhawk attack rate showed a quadratic relationship with flock size

    and peaked at 55 redshanks. Sparrowhawk attack success rate, however, declined exponentially,

    becoming less steep at flock sizes of about 40 and remaining uniformly low (a 95% decrease) by 70.

    Combined with dilution, individual risk of death per attack decreased by 95% when group size

    reached 30 (20 for the dilution effect alone).

    4. Redshanks most commonly formed group sizes that gained the maximum individual predation

    risk reduction. They also commonly formed group sizes far above any further substantial advanta-

    ges from the dilution effect or from reducing attack rate, but that continued to reduce predation

    risk by lowering attack success rate. Individuals did not always form the largest groups possible

    which we suggest is because individual variation in risk-taking subdivides the population. This

    places a constraint on the ability of individuals to compensate for predation risk and will have a

    variety of important effects on animal populations.

    Key-words: group living, hunting success, nonlethal effects, predation risk, trait-mediated inter-

    actions

    Introduction

    How animals respond behaviourally to increased predation

    risk is a key part of understanding how trophic interactions

    structure ecosystems (Lima 1998; Agrawal 2001; Werner &

    Peacor 2003; Abrams 2010). Prey behaviour in general, and

    behavioural responses to predators specifically, can deter-

    mine the immediate outcome of an encounter between preda-

    tor and prey (e.g. Cresswell & Quinn 2004; Quinn &

    Cresswell 2004), but they can also determine population

    dynamics and community structure indirectly through nonle-

    thal, trait-mediated effects as prey respond to the fear of

    predation (e.g. Cowlishaw 1997; Schmitz, Krivan & Ovadia

    2004; Minderman, Lind & Cresswell 2006; Owen-Smith &

    Mills 2006; Ripple & Beschta 2006; Creel et al. 2007). Group

    size is one of the main behavioural mechanisms used by

    animals to manage their vulnerability to predation risk (Kra-

    use & Ruxton 2002; Caro 2005) and has a major influence on

    the outcome of predatorprey interactions (Abrams 1993;

    Lima 1998; Brown, Laundre & Gurung 1999; Cresswell

    2008). However, despite the marked prevalence of this strat-

    egy in animal populations, explaining the role of predation in*Correspondence author. E-mail: [email protected]

    Journal of Animal Ecology 2011, 80, 310319 doi: 10.1111/j.1365-2656.2010.01775.x

    2011 The Authors. Journal compilation 2011British Ecological Society

  • 8/13/2019 j.1365-2656.2010.01775.x

    2/10

    predicting the group sizes observed in nature remains a major

    gap in our knowledge of how populations are structured.

    Theory and empirical data suggest that optimal group size

    reflects a dynamic interplay between a diverse range of costs

    and benefits associated with joining a group. Important costs

    of grouping include competition, kleptoparasitism andincreased conspicuousness to predators, while common bene-

    fits include gain of foraging information, group defence, vigi-

    lance and dilution of risk (Krause & Ruxton 2002). Despite

    intensive study for several decades, empirical data linking

    observed group sizes to costs and benefits, in terms of influ-

    encing probability of predation, are lacking from natural sys-

    tems (Lima 2002; Lind & Cresswell 2005; Roth, Lima &

    Vetter 2006), so limiting our ability to predict optimal group

    size, the ability of animals to compensate for increased preda-

    tion risk via grouping or avoidance (e.g. Anholt & Werner

    1995, 1998), and so the lethal and nonlethal effects of preda-

    tion risk on animal populations (see Werner & Peacor 2003).

    Animals commonly form groups because an individual

    usually has a lower probability of predation in a group

    (Krause & Ruxton 2002; Caro 2005). Decreased predation

    risk in groups arises because of the dilution effect (Foster &

    Treherne 1981; Krebs & Davies 1981), enhanced predator

    detection (Pulliam 1973; Elgar 1989) and the confusion effect

    (Neill & Cullen 1974; Parrish 1993; Schradin 2000; Ioannou

    et al.2008). All of these factors predict exponential declines

    in predation risk as group size increases, and therefore, above

    a particular group size threshold, only very small decreases in

    predation risk are expected (Pulliam 1973; Elgar & Catterall

    1981; Krakauer 1995; Roberts 1996). This means that the

    predation risk for individuals in all groups above a moderatesize will be similar, and so the selective advantage in terms of

    predation risk for individuals to form larger groups, when

    already in a large group, will be low. Measuring both the

    strength of the relationship between group size and predation

    risk, and the point at which further increases in group size no

    longer give antipredation benefits (or incur net costs), deter-

    mines the degree to which behavioural compensation can

    mediate predation risk and therefore the strength of lethal

    and nonlethal effects.

    Grouping is also generally associated with costs (Milinski

    et al. 1991; Krause & Ruxton 2002), and when these costs

    are significant, selection will act on individuals to choosethose groups that give the lowest predation risk relative to

    the number of individuals in the group (Pulliam & Caraco

    1984). The most important costs of grouping in the context

    of predation include direct competition for food, interfer-

    ence competition and enhanced attack rates from predators

    (Krause & Ruxton 2002). For example, larger groups are

    more conspicuous (Vine 1973) and generally attract higher

    attack rates from predators (Lindstrom 1989; Cresswell

    1994b; Botham et al. 2005; Carere et al. 2009; but see Fitz-

    gibbon 1990; Fernandez-Juricicet al.2004 for the opposite),

    so selecting for smaller group size. Consequently, animals

    should only join a larger group than the one they are in if it

    increases antipredation benefits. Beyond a certain group size

    (hereafter called the risk threshold), no further benefits will

    arise because predation risk will remain relatively uniform.

    Assuming no other benefits for increasing group size and

    that animals are risk minimizing, the optimal flock size will

    occur at the risk threshold where maximal antipredation

    benefits accrue while incurring the lowest possible costs. In

    this paper, we explore this prediction by examining whethera bird species that gains no foraging or other benefits from

    grouping (Sansom et al. 2008) usually forms flocks that are

    no larger than the risk threshold and therefore how optimal

    group size arises from predation risk.

    The determinants of optimal group size in the context of

    predation risk can be summarized for an individual as fol-

    lows:

    Predation risk 1 Group attack rate 2

    Attack success rate 3 Individual attack rate

    (1) Groups of different sizes are likely to be attacked at dif-

    ferent rates because they vary in conspicuousness (Vine

    1973). Generally, larger groups attract higher attackrates because of predator functional or numerical

    responses, although this may reach an asymptote if pre-

    dators suffer interference competition (Sutherland

    1996). Selection therefore acts initially to reduce group

    size (Fig. 1a).

    (2) As individuals join larger groups, however, there is usu-

    ally a reduction in the likelihood of any attack on the

    group being successful (e.g. Kenward 1978; Cresswell

    1994b, 1996; Krause, Ruxton & Rubenstein 1998; Fun-

    ston, Mills & Biggs 2001; Roth & Lima 2003; Cresswell

    & Quinn 2004). This arises through a number of mecha-

    nisms, for example because of shared vigilance, where

    the probability of any animal scanning and so detecting

    a predator at any one time increases (Pulliam 1973; El-

    gar & Catterall 1981; Roberts 1996). The confusion

    effect is also important, where predators find it hard to

    track multiple moving targets (Krakauer 1995). Both

    vigilance and confusion predict exponential increases in

    effect with group size (Fig. 1b) and therefore an expo-

    nential decline in attack success with group size

    (Fig. 1c).

    (3) Individual risk correlates negatively with group size

    because of the dilution effect (Krebs & Davies 1981). If

    all individuals have an equal chance of being targeted,

    then the probability that an individual will be killed onattack will be 1group size, and so an exponential

    decline in predation risk with group size will occur

    (Fig. 1d).

    Therefore, negative selection on group size is expected to

    minimize attack rate, but positive selection is expected

    because of attack success reductions and the dilution effect.

    Because both attack success rate and dilution effect decline

    exponentially with flock size, there should be a clear risk

    threshold above which predation risk becomes reasonably

    constant. If there is negative selection owing to the attack rate

    relationship (e.g. foraging costs with increasing group size),

    this risk threshold should predict the group size chosen by

    most individuals (i.e. optimal group size; Fig. 1).

    Optimal group size and predation risk 311

    2011 The Authors. Journal compilation 2011British EcologicalSociety, Journal of Animal Ecology, 80, 310319

  • 8/13/2019 j.1365-2656.2010.01775.x

    3/10

    Optimal group size should, however, depend on the rela-

    tive strength of the differing selection relationships. For

    example, see Fig. 1 and compare the effects of arbitrary

    differences in risk threshold between the relationships that

    comprise predation risk on the likely observed group size.

    If attack rate was the main factor determining predationrisk, optimal group size will be as small as possible, assum-

    ing only antipredation benefits to grouping (Fig. 1e). For

    example, for a large animal such as a baleen whale ( Mysti-

    ceti sp.) hunting krill (Euphausiacea sp.), where the entire

    group is easily engulfed so removing any variation in cap-

    ture success with group size, or the dilution effect, then

    selection would lead to smaller groups that would be able

    to escape detection, or that would be uneconomic for a

    whale to pursue.

    If attack success rate was the main factor determining pre-

    dation risk, and again assuming only antipredation benefits

    to grouping, optimal group size will always be above the risk

    threshold of the attack success rate relationship (Fig. 1f,

    dashed line). For example, consider a colonially breeding

    bird attacked by a mammalian predator. In this system,

    attack rate might not depend on group size because a breed-

    ing colony of any size is always conspicuous, nor the dilution

    effect because the predator always targets the weakest indi-

    viduals identified during the attack. Selection would then act

    to promote groups that minimized capture success per attack

    (i.e. above the risk threshold). If there was a cost to grouping,

    such as competition for food or nest sites, resulting in selec-

    tion for smaller groups, then we would predict that group

    sizes would converge around the risk threshold, the exact

    location dependent on the relative strength of the costs andbenefits to grouping (Fig. 1f, solid line).

    If the dilution effect was the main factor determining pre-

    dation risk, and again assuming only antipredation benefits

    to grouping, optimal group size will always be above the risk

    threshold of the dilution effect relationship (Fig. 1g, dashed

    line). For example, a migratory ungulate crossing a river,

    where any individual or group is highly conspicuous to a

    crocodile Crocodylus niloticus attacking from below and

    where the prey have no effective escape behaviour. Again if

    there are costs to grouping, such as large groups interfering

    with each other as they crossthe river, reducing their speed of

    travel, then we would again predict that group sizes wouldconverge around the risk threshold, the exact location depen-

    dent on the relative strength of the costs and benefits to

    grouping (Fig. 1g, solid line).

    This paper explores the importance of group sizedepen-

    dent attack rate, attack success rate and individual attack

    rate in predicting group size in an animal population under

    natural conditions. We recorded daily mean group size in

    wintering redshanks Tringa totanus in a system where the for-

    aging benefits from shared vigilance are cancelled by interfer-

    ence competition, leading to no net effect of group size on

    foraging success among redshanks (Sansom et al. 2008).

    Therefore, optimal group size is probably entirely determined

    by variation in predation risk, and so the group size selected

    by most individuals should reflect this. We then measured the

    attack rate and attack success rate relationship with group

    size for Eurasian sparrowhawks Accipiter nisus attacking and

    killing groups of redshanks (see Cresswell & Whitfield 2008

    for an overview of the system), and along with the dilution

    effect relationship, we determined the likely location of the

    risk threshold for all three factors. Finally, we examinedwhether the risk threshold matched the most favoured group

    size. First, we tested the general prediction that risk thresh-

    olds should predict optimum group size by calculating the

    overall predation risk relationship (using the equation men-

    tioned earlier) and its risk threshold. Second, we assessed

    whether any one component attack rate, success or dilution

    might have the strongest selective effect on group size by

    comparing which individual risk threshold best predicted

    group size.

    Materialsandmethods

    The study area consisted of salt marsh habitat backed by

    woodland or dunes at the Tyninghame Estuary, East Lothi-

    an, Scotland (see Whitfield 1985 for further study site

    details). Data were collected from September to early March,

    198992, and 200106. The salt marsh (c. 15 ha) provides a

    feeding habitat for wintering redshanks, in particular for first

    winter (juvenile) birds (Cresswell 1994a).

    O C C U R R E N C E O F R E D S H A N K S I N D I F F E R E N T F L O C K

    S I Z E S

    Data on the occurrence of redshanks with respect to the flock sizes

    they chose were collected by scan samples on the salt marsh during

    the winter of 20012 (see Quinn & Cresswell 2004) during 1- to 6-h

    observation periods from fixed locations overlooking the whole salt

    marsh, except during high tide periods when the salt marsh was cov-

    ered (to any degree) by water. During observation periods, the total

    number of redshanks and the number of flocks were recorded on the

    salt marsh every 30 min, along with a potentially confounding vari-

    able affecting predation risk, the mean distance to predator-conceal-

    ing cover for each flock. A flock was defined as a cluster of birds in

    which the maximum nearest neighbour distance was 25 m and typically varied in size from 1 to 100

    birds.

    Observations were made on 57 days with 89 07 scan samples a

    day. Mean flock size SE was 183 15 birds (n = 57 observa-

    tion days), with a mean SE salt marsh population of 387 27

    birds available to form flocks (n = 57 observation days). Each set of

    focal samples from a sample time had an overall population size (i.e.

    the sum of all flock sizes), and this was classified into a population

    size class (110,in intervalsof 10 until 91100and then 101150, with

    class sizes chosen to equalize the variance in each class). When the

    same population size class occurred in more than one focal sample in

    a day, a daily mean value for flock size was calculated for that popu-

    lation size class. The sampling unit for analysis was therefore the

    number of separate days in which that population size class was sam-

    pled. We then tested whether there was an optimum flock size by

    determining the flock size at which further increases in daily salt-

    marsh population size resulted in no further major increases in meandaily flock size.

    312 W. Cresswell & J. L. Quinn

    2011 The Authors. Journal compilation 2011British Ecological Society, Journal of Animal Ecology, 80, 310319

  • 8/13/2019 j.1365-2656.2010.01775.x

    4/10

    Population size was frequently limited duringa scan andprecluded

    theformation of large flocks. We,therefore, analysed a subsetof scan

    samples where the total population of redshanks on the salt marsh

    exceeded 100. This gave 21 observation days, with mean SE

    27 05 scans and 89 18 flocks recorded per day. We then cal-

    culated the total number of birds recorded per day in 12 flock size

    classes (1, 35, 610, 1115, 1620, and then increments of 10 until

    80, 81100: increments defined in this way to be consistent with those

    used to calculate capture success). We divided these totals for each

    day by the number of scan samples for each day to obtain compara-

    ble relative frequencies across days. We then summed the number of

    birds recorded across days within each flock size class. We deter-

    mined the existence of a flock size class most favoured by redshanks

    by determining whether there was a quadratic relationship between

    thefrequency of redshanks withinthe flock size classes and flock size.

    Relative flock availability with respect to sparrowhawk attack rate

    was calculated by pooling flock size records into the same 12 flock

    size classes as used previously and summing the total number of

    flocks recorded in eachsizeclass in a day.The proportion of the over-

    all total that each flock-size-class total represented was then calcu-lated for each day (i.e. standardizing daily relative flock size class

    availability because of variable daily sample effort). The daily pro-

    portions were then summed for each flock size class across the

    57 days sampled. The overall mean relative abundance was then

    scaled to the abundance of group size 1 (e.g. if the abundance of

    group sizes 1 and 10 was 15 and 5, respectively, relative abundances

    were calculated as 1515 = 1and 515 = 033).

    S P A R R O W H A W K A T T A C K D A T A

    Sparrowhawk attacks were seen on 288 separate days over 8 winters,

    198992 and 200106 (mean SE = 2 5 03 attacks per day on

    days when attacks were recorded). During the initial more intensivestudy in the first three winters (2557 h spent at the study site), it was

    estimatedthat at least six sparrowhawksattackedduringeach winter.

    Some individuals were probably recorded across winters (see

    Cresswell & Whitfield 1994). The 16-year period over which observa-

    tions took place, however, makes it likely that many different birds

    were involved.

    An attack was defined as a rapid flight directed towards a flock or

    a single bird. A kill was recorded when the raptor captured a

    redshank. For each attack, flock size and distance from predator-

    concealing cover were estimatedwhen possible. Flocks almost invari-

    ably contained only redshank; the presence of additional birds was

    ignored in flock size estimates, and the attack was not considered if

    other species were targeted, which happened very rarely. Markers

    were placed at regular intervals around the edge of the salt marsh to

    facilitate estimating distances. Flock size and distance to cover were

    usually recorded before attacks. For some attacks, details were

    obtained from videos (see Quinn & Cresswell 2006). In total, we saw

    641 surprise (where sparrowhawks attacked directly from concealing

    cover see Cresswell 1996) attacks where flock size was accurately

    recorded, resulting in 101 captures. All attacks were included where

    we had a measure of flock size; usually, this was recorded before any

    escape response had occurred because flocks were under continuous

    observation. After theattack, we confirmed theflock size by counting

    the flock again. Sample sizes of attacks and captures are larger than

    those in Cresswell, Lind & Quinn (2010) because flock size informa-

    tion could almost always be recorded accurately during an attack,

    whereas distance to cover could not be for some attacks, where fore-shortening, or an unclear view of the entire flock before flight

    prevented an accurate and unbiased estimate of initial distance from

    cover.

    Attack success rate was calculated by pooling into the same 12

    flock size classes used for flock availability above, with class sizes

    determined to spread captures as equally as possible between classes.

    The relationship between sparrowhawk attack success rate and flock

    size may be confounded by distance to cover, which also strongly

    affects capture success rate in our system (Cresswell 1994b; Cresswell

    & Quinn 2004). However, any variation in attack rate arising from

    distance to cover was uniform with respect to group size and should

    not bias our estimates of the relationship between hunting behaviour

    and flock size. On average, there was no significant variation in dis-

    tance to cover across group size for attacking sparrowhawks (e.g.

    KruskalWallisv211 = 126, P = 032 data pooled withinthe 12 clas-

    ses of increasing flock size).

    G E N E R A L S T A T I S T I C A L M E T H O D S

    Analyses were carried out in R (R Development Core Team 2009),

    predominantly using linear models (lm command in R); all finalmodels met the assumption of normally distributed data and

    homogeneity of variance as demonstrated by the (plot) command

    in R and according to criteria in Crawley (2007). Relationships

    between variables were modelled with the three most likely biologi-

    cal models: a linear relationship (continuous increase), a quadratic

    relationship (a peak or optimum) and a logarithmic relationship (a

    rapid increase becoming gradual). Where biologically appropriate

    (i.e. flock size must be zero when population is zero), models were

    constrained to pass through the origin. Models were compared

    using AICC values because of small sample sizes (Burnham &

    Anderson 2002). For the relationship between flock size and popu-

    lation size, a logarithmic relationship was not expected and there-

    fore tested because flock size must always be less than or equal to

    population size. For the relationship between attack rate and flock

    size, attack rate was adjusted for variation in overall availability of

    the different flock size classes by dividing attack rate by the relative

    abundance of the different flock size classes. Variation in attack

    success across flock size classes was further tested using Chi-square

    tests (summary command in R) and using logistic regression (glm,

    family = binomial, command in R) to determine how attack suc-

    cess (capture or escape) was dependent on flock size and distance

    to cover in the unpooled data.

    Results

    G R O U P S I Z E F R E Q U E N C Y

    Mean daily flock size on the salt marsh varied typically from

    1 to 60 (mean SE = 183 15). Flock size was best pre-

    dicted by a shallow quadratic relationship with population

    size [(056 004 SE) population size]) [(00015

    000042 SE) population size2]; population size, F1,9 =

    12673,P < 0001; population size2,F1,9= 125,P= 0006;

    adjustedR2 = 099: Fig. 2. A linear function gave a poorer

    fit (DAICC= 79).

    When population size was not constraining on the salt

    marsh (i.e. more than 100 birds available), the observed flock

    size in which bird frequency peaked occurred at about 75

    (Fig. 3). The total number of birds found in a flock size classwas best predicted by a logarithmic relationship which

    Optimal group size and predation risk 313

    2011 The Authors. Journal compilation 2011British EcologicalSociety, Journal of Animal Ecology, 80, 310319

  • 8/13/2019 j.1365-2656.2010.01775.x

    5/10

    Atta

    cksuccessrate

    Strong selection

    for larger

    groups

    Weak selection forlarger groups

    Attack success rate

    RT

    Frequency

    Group sizeRT

    Strong selection for small groups

    Frequency

    Predation risk

    group attack rate(a)

    (c)

    (b)

    (d)

    Likely observed number

    of individuals(e)

    Frequency

    RT

    0 20 40 60 80 100

    0 20 40 60 80 100

    0 20 40 60 80 100 0 20 40 60 80 100

    0 20 40 60 80 100

    0 20 40 60 80 1000

    1

    Weak selection for

    larger groups

    Strong

    selection

    for

    larger

    groups

    Relativeprobabilityofflock

    beingattacked

    Probability

    ofindividualbeingattacked

    wh

    enaflockisattacked

    Group size

    Dilution

    RT

    0 20 40 60 80 10000

    05

    10

    Probabilityofdetection

    ormaximal"confusion"

    (g)

    (f)

    Fig. 1.Graphs showing predicted values of

    howgroup attackrate (a), attacksuccess rate

    (c) and the dilution effect (d) (where individ-

    ual risk of being targeted is 1group size)

    should vary with group size, and how corre-

    sponding selection pressure should lead to

    preference of certain group sizes by individu-

    als minimizing predation risk, and so greater

    observed frequencies of individuals in less

    risky flocks (eg).Graph (a) shows two pote-

    ntial functions linking attack rate and group

    size; others are possible, but all predict that

    attackrate increases initially with group size.Attack success and dilution both have a

    threshold (risk threshold labelled RT)

    where further change in the function makes

    little further difference to predation risk.

    Correspondingly, we predict that there will

    be strong selection for individuals to avoid

    risky flock sizes below these thresholds (lead-

    ing to the dashed line distributions in f and

    g). If there is also strong selection from

    attack rate to minimizeflocksize (a), we pre-

    dict a peak of preference for individuals in

    flock sizes close to the risk threshold (leading

    to the solid line distributions in f and g).

    Inset graph (b) shows how the probability of

    detection and confusion changes with groupsize leading to the predicted exponential

    change in attack success rate function with

    group size.

    0 20 40 60 80 100 1200

    10

    20

    30

    40

    50

    60

    28

    27

    3326

    3130 24

    18

    1111

    17

    Meanflock

    size(+1SE)

    Mean population size

    Flock size observed

    Flock size maximised

    Fig. 2.Mean flock size as a function of the total population of red-

    shanks available (pooled into population size classes). Overall mean

    flock size values (acrossall days that a population size class was sam-

    pled) were then plotted, withn being the number of separate days in

    which that population size class was sampled (shown by the numbers

    above each point). The line plots the quadratic function of best fit

    (F2,9= 6400, P < 0001, Adj. R2 = 099).

    0 20 40 60 80 1000

    50

    100

    150

    200

    250

    300

    350

    400

    Frequenc

    y

    Flock size

    Fig. 3.The frequency of individual birds in each flock size class for

    scan samples when the population of the salt marsh exceeded 100

    birds. Scans were carried out on 30 separate days, and each days

    samples were weighted so each day contributes equally, irrespective

    of sampling effort. The line plots the logarithmic function of best fit

    (F1,11= 1233, P < 0001, Adj. R2 = 091).

    314 W. Cresswell & J. L. Quinn

    2011 The Authors. Journal compilation 2011British Ecological Society, Journal of Animal Ecology, 80, 310319

  • 8/13/2019 j.1365-2656.2010.01775.x

    6/10

    showed that very few birds were in flock sizes of less than 30,

    with a gradually increasing frequency of birds in flocks above

    this flock size [total number of birds = (556 50

    SE) ln (flock size); F1,11 = 1233, P < 0001, adjusted

    R2 = 091]: Fig. 3. A quadratic relationship that peaked at

    about flock size 80 was a poorer (DAICC= 15), although

    also reasonable, predictor of the relationship (flock size,

    B = 69 14 SE, F1,10 = 1081, P < 0001; flock size2,

    B = )0044 0020,F1,10= 52,P = 0045;F2,10 = 567,

    P< 0001, adjusted R2 = 090). A linear function gave a

    much poorer fit (DAICC = 49).

    A T T A C K R A T E

    Sparrowhawk attack rate, relative to the availability of each

    flock size class, increased with flock size, with peak observed

    values at flock sizes of about 65 (Fig. 4). Attack rate was best

    predicted by a quadratic relationship with flock size, with

    predicted values peaking at flock sizes of about 55 (attack

    rateflock availability = [(00076 0024 SE) flock size]

    ) [(000068 000027 SE) flock size2] + ( 04 7 040

    SE); flock size, F1,9= 78, P = 0021; flock size2, F1,9= 62,

    P= 0030; adjusted R2 = 052). A logarithmic function

    gave a poorer fit (DAICC = 26) as did a l inear fit

    (DAICC= 43).

    A T T A C K S U C C E S S

    Attack success declined steeply with group size until about

    40, after which the decline became relatively shallow (logistic

    regression flock size B =)

    0029 0

    007, P < 0

    001,

    n = 540 unsuccessful attacks, 101 captures). This relation-

    ship was very similar, B = )0026 0007, see Cresswell,

    Lind & Quinn (2010), when also including distance to cover,

    as would be expected from the equal distribution of all dis-

    tances to cover across the flock size classes see Materials

    and methods). When data were pooled into flock size classes

    to equalize samples of kills attack success was best predicted

    by a logarithmic function [attack success rate = ()0-

    0063 00086 SE) ln (flock size) + (037 0028 SE);

    flock size, F1,10 = 530, P < 0001; adjusted R2 = 083]:

    Fig. 5. A quadratic function gave a poorer fit (DAICC= 27)

    as did a linear fit (DAICC = 66). Attack success was signifi-

    cantly affected by flock size up to 40 birds (v26 = 243,

    P < 0001), but not for flock sizes above 40 (v24 = 29,

    P = 057).

    C O M P A R I N G G R O U P S I Z E T O A T T A C K R A T E , A T T A C K

    S U C C E S S A N D D I L U T I O N R I S K T H R E S H O L D S

    Almost all (95%) of the decline in risk occurred by flock size

    of about 20 for the dilution effect (Fig. 6), by about30 for the

    overall individual predation risk relationship (Fig. 6) and by

    about 70 for the attack success relationship (Fig. 5). The

    0 20 40 60 80 10000

    05

    10

    15

    20

    25

    30

    35

    Attackrate/flockavailab

    ility

    Flock size

    Fig. 4.The relative attack rate per availability of each flock size class

    (scaled to the attack rate for group size 1); sample sizes of attacks as

    in Fig. 5. The line plots the quadratic function of best fit (F2,9= 70,

    P = 0014, Adj. R2 = 052).

    0 20 40 60 80 100

    000

    005

    010

    015

    020

    025

    030

    035

    65

    69 7157

    54

    9782

    53

    45

    24

    9

    16Attacksuccess

    rate

    Flock size

    Fig. 5.Capture success rate with flock size with attacks pooled into

    classes so rates of capture can be calculated; numbers of attacks in

    each class are given beside each point. The line plots the logarithmic

    functionof bestfit (F1,10 = 530, P < 0

    001, Adj. R

    2

    = 083).

    0 20 40 60 80 100

    00

    05

    10

    During the whole

    study average

    flock size was 183

    Relativerisk

    Flock size

    Dilution

    Overall individual predation risk function

    Attack success

    Attack rate

    Most birds were in

    flocks of 75 95

    when population

    size allowed

    During the whole

    study most birds

    were in flocks of

    40 60

    Fig. 6.A summary of the relationships forattack rate, attacksuccess,

    dilution and overall predation risk (=attack success attack rate

    function 1group size) with group size.

    Optimal group size and predation risk 315

    2011 The Authors. Journal compilation 2011British EcologicalSociety, Journal of Animal Ecology, 80, 310319

  • 8/13/2019 j.1365-2656.2010.01775.x

    7/10

    attack success relationship suggested selection for joining

    flocks as large as possible (Fig. 4). Optimum flock size, as

    measured by the one adopted by most birds when uncon-

    strained by population size, was at least above 30, with most

    birds being in larger flocks than this with a possible peak at

    around 80 (Fig. 3), above all of the risk thresholds, but clos-est to that of attack success.

    Conclusions

    W H I C H P R E D A T I O N R I S K F A C T O R D E T E R M I N E S G R O U P

    S I Z E ?

    Redshanks continued to increase group size as population

    size increased, both independently of attack rate and

    beyond the threshold predicted by both the dilution effect

    and the individual risk of being killed per attack (see

    Fig. 2). Redshanks were found most commonly in large

    flocks, although there was little increase in the frequency

    with which redshanks favoured larger flocks above a flock

    size of about 30 (see Fig. 3). This relationship between the

    preferred flock size and population size has two possible

    interpretations. First, the quadratic fit suggests an optimum

    flock size of about 80, but the logarithmic relationship sug-

    gests redshanks avoid small flocks (less than 30) but still

    attempt to form large flocks at least up until 100 (the maxi-

    mum flock size we included). Either way, redshanks clearly

    preferred large flock sizes up to around 80100, and the

    group sizes most frequently used by birds were therefore

    closest to those that might be predicted from the overall

    individual predation risk relationship and the attack successrate relationship.

    The strength of selection for the dilution effect above flock

    sizes of about 30 was minimal predicting few larger flocks if

    this were the main driver of group size (assuming no other

    advantages of flocking), and this is shown by the reduction in

    the strength of increase in the relationship shown in Fig. 3.

    Advantages of increasing group size from the attack success

    rate function, however, were still substantial, even up to

    group sizes of 70, and we found a preference for flocks up

    until at least 80, as shown by the continuing increase (or

    peak) in number of birds using flocks up to about 80 in

    Fig. 3. Some of the threshold estimates were uncertain. Forexample, the attack success threshold may have been as low

    as 40, andoptimum group size may nothave been reached by

    a group size of 80. This does not, however, greatly change

    our conclusions: the group size most commonly used by indi-

    viduals was above the risk thresholds of all the antipredation

    relationships with group size we examined, suggesting that

    even the very small decreases in predation risk accrued by an

    individual moving from a large group to a slightly larger

    group give fitness benefits sufficient to have led to selection

    for this behaviour. Our results also clearly showed that red-

    shanks avoided flocks of less than about30 as predicted from

    the very large fitness benefits of joining a flock above the

    overall individual predation risk threshold of about 30 (see

    Fig. 6).

    Attack success rate may be an important selective factor

    driving the flock size increases above the overall individual

    predation risk threshold in redshanks because behaviourally

    driven variation in flock vulnerability is the main determi-

    nant of attack rate and mortality (Quinn & Cresswell 2004)

    and because sparrowhawks are opportunistic, generalist pre-dators (Cresswell 1995). Any reduction in success rate should

    encourage a shift to other more vulnerable prey. If sparrow-

    hawks attack prey that they can most efficiently catch (i.e. the

    smallest number of attacks to gain the largest prey possible),

    then their prey choice will be determined by the relative vul-

    nerability and size of different prey species. Redshanks are

    relatively large prey for sparrowhawks (Newton 1986; Cres-

    swell 1995) and so might be preferentially attacked even when

    there is a low chance of success. But any reduction in attack

    success, even when this is already low, may lead to a sparrow-

    hawk shifting prey species, as another prey becomes the most

    profitable (Newton 1986), so leading to strong selection on

    redshanks to reduce attack success rates as much as possible.

    Reducing attack success rate as low as possible should greatly

    reduce attack rate, particularly if the prey-switching thresh-

    old is passed, although this threshold will be dependent on

    the availability and vulnerability of alternative profitable

    prey (Jackson et al. 2006).

    Although redshanks on the study site were primarily

    attacked and killed by sparrowhawks, peregrines Falco pere-

    grinus also attacked and killed redshanks frequently (Cres-

    swell & Whitfield 1994) and therefore observed flock sizes

    may also be influenced by the effects of variation in peregrine

    attack rate and success with flock size. The decline in success

    rate with group size for surprise attacks was the same forboth predator species; however, peregrines attacked smaller

    flocks more frequently than sparrowhawks (Cresswell &

    Quinn 2010) and therefore further increase the predation risk

    associated with very small flocks, in effect making the

    increase in attack rate with flock size illustrated in Fig. 4 shal-

    lower. The lack of a strong effect of attack rate on observed

    favoured flock size may then partly be a consequence of a

    more uniform overall attack rate when all predators are con-

    sidered, but this effect is likely to be minor because peregrines

    attack redshanks much less frequently than sparrowhawks

    (Cresswell & Quinn 2010).

    A T T A C K S U C C E S S R A T E

    Most previous studies have related attack success rate to

    broad flock size categories, rather than to continuous mea-

    sures, but have found broadly similar results: attack success

    rate drops off very rapidlyfor group size (e.g. Kenward 1978;

    Lindstrom 1989; Krause & Godin 1995; Roth & Lima 2003).

    Our results show, however, that there are still reasonable

    advantages in terms of reduction in attack success rate for

    redshanks in flocks up to about 70. This is perhaps surprising

    considering that there is unlikely to be any enhanced benefit

    in terms of predator detection in groups above 20 birds, both

    theoretically (Pulliam 1973; Roberts 1996) and empirically

    for redshank (Cresswell 1994b), and also in terms of the

    316 W. Cresswell & J. L. Quinn

    2011 The Authors. Journal compilation 2011British Ecological Society, Journal of Animal Ecology, 80, 310319

  • 8/13/2019 j.1365-2656.2010.01775.x

    8/10

    confusion effect, both theoretically for groups of more than a

    few individuals (Krakauer 1995) and empirically for Daphnia

    above group sizes of about 30 (Jeschke & Tollrian 2007). Yet

    our results show that reductions in attack success do occur

    over larger group sizes well above where further vigilance

    and confusion benefits should accrue. Why this should be sois not clear, but potential explanations include enhanced col-

    lective detection, because accurate and rapid escape

    responses rely not only on detection but on simultaneous

    detection and response of several flock members (Cresswell,

    Hilton & Ruxton 2000; Quinn & Cresswell 2005a), and

    increase in the probability of larger flocks containing unusual

    or poorly behaving individuals that are identified and prefer-

    entially targeted during an attack, thus countering the confu-

    sion effect(Ohguchi 1978; Quinn & Cresswell 2006).

    A T T A C K R A T E

    Large flocks (55) were attacked approximately two and a half

    times as often as small flocks. Although our data are sparse

    for flocks above 55, very large flocks seem to also have been

    preferentially avoided by sparrowhawks. Flock conspicuous-

    ness is likely to have influenced attack rate: single birds are

    often relatively difficult to detect on the salt marsh from the

    relatively low vantage points normally used by sparrow-

    hawks, particularly when redshanks are feeding in creeks or

    depressions (W. Cresswell, pers. obs.). On many occasions,

    we witnessed a single redshank flying away after a sparrow-

    hawk had flown over the redshank on its way to attack a

    group of redshanks further away from it, suggesting that the

    closer and more vulnerable redshank (see Cresswell & Quinn2004) had not been detected. But above five or so redshanks,

    any group on the salt marsh is conspicuous: redshanks are

    very mobile when feeding because of interference competi-

    tion (Minderman, Lind & Cresswell 2006) and so are detect-

    able to a human observer at a distance of hundreds of metres

    (pers. obs.). It therefore seems unlikely that the increase in

    attack rate above small flock sizes is driven by conspicuous-

    ness. One reason for attacks on large flocks, even when suc-

    cess rates are on average low, may be that because

    sparrowhawk hunts are very brief (Cresswell 1996), and the

    redshanks remain in the general area after attack (Cresswell

    & Whitfield 1994), sparrowhawks can probe for weakenedindividuals or even create more vulnerable prey with repeated

    attacks (see Charnov, Orians & Hyatt 1976; Lima 1990,

    2002; Brown, Laundre & Gurung 1999).

    D I L U T I O N E F F E C T A N D O V E R A L L R I S K F U N C T I O N

    The dilution effect ensures that any effects of variation in

    attack rate and attack success rate are swamped with increas-

    ing group size. For example, elsewhere it has been shown in

    this system (Cresswell 1994b although with much lower sam-

    ple sizes) that individual risk still decreases rapidly even with

    elevated attack rates on larger flocks. The dilution effect,

    however, relies on an unrealistic assumption that all prey

    have an equal chance of being targeted, when in reality

    differences in the vulnerability of individuals within flocks

    lead to preferential targeting (Quinn & Cresswell 2006).

    Therefore, it is perhaps not surprising that flocks above the

    threshold for the dilution effect are preferred because any

    vulnerable individual is likely to have more vulnerable com-

    panions in larger flocks.

    W H Y D O N O T R E D S H A N K S A L W A Y S M A X I M I Z E F L O C K

    S I Z E ?

    Redshanks increased flock size where possible, increasing

    average flock size as the pool of available redshanks

    increased (Fig. 3). They did not, however, frequently form

    one flock when populations were below 80. The most likely

    hypothesis for this discrepancy may be because, although we

    counted the whole population of redshanks available to form

    flocks on the salt marsh, these birds were likely to have varied

    in their position on the starvationpredation risk trade-off

    continuum (Cresswell & Whitfield 2008; Sansom, Lind &

    Cresswell 2009). Some birds may have been prepared to feed

    closer to cover and at higher risk than others because their

    risk of starvation exceeded their risk of predation. Other

    birds that were prioritizing risk avoidance, rather than forag-

    ing would notbe prepared to join birds that were taking more

    risks (Cresswell, Lind & Quinn 2010). Individual variation in

    body condition and starvation risk may inevitably always

    lead to a subdivided population which constrains the ability

    of all individuals to maximize group size. Similarly, the use of

    alternative antipredation strategies linked to personality

    (Quinn & Cresswell 2005b) could also lead to the nonrandom

    distribution of behavioural phenotypes across groups.

    G E N E R A L P O P U L A T I O N A N D C O M M U N I T Y S T R U C T U R E

    C O N C L U S I O N S

    Our results have several implications for how predatorprey

    interactions might influence trophic interactions and commu-

    nity structure in ecosystems generally. First, species that are

    able to form groups may be able to respond to increased

    predation risk without having to avoid areas frequented by

    predators, and generalist predators may then switch to more

    profitable prey in other areas. This may lead to predation on

    populations being inversely density dependent and to Alleeeffects because predators may primarily impact small popu-

    lations that are unable to form large groups (e.g. Mooring

    et al. 2004; Angulo et al. 2007; Watson, Aebischer &

    Cresswell 2007). Second, if a predator has a high attack suc-

    cess rate that declines steeply with larger group sizes of prey,

    then selection will favour prey that form groups. Formation

    of groups will then change the local density of a species so

    affecting its competitive interactions with conspecifics and

    the prey species with which it shares a food supply, as well as

    affecting the spatial availability and distribution of this food

    supply (Minderman, Lind & Cresswell 2006). Third, if the

    threshold for losing further antipredation benefits from

    grouping occurs at low group sizes, then nonlethal effects will

    be relatively large because optimum group size is unlikely to

    Optimal group size and predation risk 317

    2011 The Authors. Journal compilation 2011British EcologicalSociety, Journal of Animal Ecology, 80, 310319

  • 8/13/2019 j.1365-2656.2010.01775.x

    9/10

    be constrained by the availabilityof individuals. Fourth, even

    when increasing group size may confer substantial advanta-

    ges, unless all individuals are prepared to feed in the same

    areas (i.e. take the same risks), maximum group sizes will not

    occur unless the population is sufficiently large to allow opti-

    mal subpopulations to form. Increased variation in condi-tion, dominance or the competitive ability of individuals will

    therefore reduce the likelihood of optimal group size forming

    and so decrease the strength of nonlethal effects. The precise

    manner in which any of these effects arise is likely to be sys-

    tem dependent.

    Acknowledgements

    We thank NERC, the Royal Society and the Leverhulme Trust for funding.

    We also thank Philip Whitfield, East Lothian District Council, the Tyningh-

    ame Estate and Sue Holt. We thank the Editors, Andrews Jackson and two

    anonymous referees for helpful comments.

    References

    Abrams, P.A. (1993) Why predation rate should not be proportional to preda-

    tor density. Ecology, 74, 726733.

    Abrams, P.A. (2010) Implications of flexible foraging for interspecific interac-

    tions: lessons fromsimple models. Functional Ecology, 24, 717.

    Agrawal, A.A. (2001)Phenotypic plasticityin the interactions and evolution of

    species. Science, 294, 321326.

    Angulo, E., Roemer, G.W., Berec, L., Gascoigne, J. & Courchamp, F. (2007)

    Double allee effects and extinction in the island fox.Conservation Biology,

    21, 10821091.

    Anholt, B.R. & Werner, E.E. (1995) Interaction between food availability and

    predation mortality mediated by adaptive behaviour. Ecology, 7 6, 2230

    2234.

    Anholt, B.R.& Werner, E.E. (1998) Predictable changes in predation mortality

    as a consequence of changes in food availability and predation risk. Evolu-

    tionaryE cology, 12, 729738.

    Botham, M.S., Kerfoot, C.J., Louca, V. & Krause, J. (2005)Predator choicein

    the field; grouping guppies, Poecilia reticulata, receive more attacks. Behav-

    ioralEcology andSociobiology, 59, 181184.

    Brown, J.S., Laundre, J.W. & Gurung, M. (1999) The ecology of fear: optimal

    foraging, game theory, and trophic interactions.Journal of Mammalogy,80,

    385399.

    Burnham, K.P. & Anderson, D.R. (2002) Model Selection and Multimodel

    Inference:A Practical Information-Theoretic Approach. Springer,New York.

    Carere, C., Montanino, S., Moreschini, F., Zoratto, F., Chiarotti, F., Santucci,

    D. & Alleva, E. (2009) Aerial flocking patterns of winteringstarlings, Sturnus

    vulgaris, underdifferent predationrisk. Animal Behaviour, 77, 101107.

    Caro, T.M. (2005)Antipredator Defenses in Birds and Mammals. University of

    Chicago Press, Chicago.

    Charnov, E.L., Orians, G.H. & Hyatt, K. (1976) Ecological implications of

    resource depression. American Naturalist, 110, 247259.

    Cowlishaw,G. (1997) Trade-offs betweenforagingand predationrisk determinehabitat usein a desert baboonpopulation. AnimalBehaviour, 53, 667686.

    Crawley, M.J.(2007) TheR Book. John Wiley & Sons Ltd, Chichester.

    Creel, S., Christianson, D., Liley, S. & Winnie Jr, J.A. (2007) Predation risk

    affects reproductive physiology and demographyof Elk. Science, 315, 960.

    Cresswell,W. (1994a) Age-dependent choice of redshank (Tringa totanus) feed-

    ing location:profitabilityor risk?Journalof Animal Ecology, 63, 589600.

    Cresswell, W. (1994b) Flocking is an effective anti-predation strategy in Red-

    shanks, Tringa totanus. Animal Behaviour, 47, 433442.

    Cresswell,W. (1995)Selection of avianprey by wintering sparrowhawks Accip-

    iternisus in southern Scotland. Ardea, 83, 381389.

    Cresswell, W. (1996) Surprise as a winter hunting strategy in Sparrowhawks

    Accipiter nisus, Peregrines Falco peregrinusand Merlins F. columbarius. Ibis,

    138, 684692.

    Cresswell, W. (2008) Non-lethal effects of predation risk in birds. Ibis,150, 3

    17.

    Cresswell, W.,Hilton,G.M. & Ruxton, G.D. (2000)Evidence fora rule govern-

    ing the avoidance of superfluous escape flights. Proceedings of the Royal

    Societyof London Series B, 267, 733737.

    Cresswell, W., Lind, J. & Quinn, J.L. (2010) Predator hunting success and prey

    vulnerability: quantifying the spatial scale over which lethal and non-lethal

    effects of predationoccur. Journalof Animal Ecology, 79, 556562.

    Cresswell, W. & Quinn, J.L. (2004) Faced with a choice, predators select the

    most vulnerable group: implications for both predators and prey for moni-

    toring relative vulnerability. Oikos, 104, 7176.

    Cresswell, W. & Quinn, J.L. (2010) Attack frequency, attack success and choice

    of prey groupsize for two predatorswith contrasting hunting strategies. Ani-mal Behaviour, 80, 643648.

    Cresswell, W. & Whitfield, D.P. (1994) The effects of raptor predation on win-

    tering wader populations at the Tyninghame estuary, southeast Scotland.

    Ibis, 136, 223232.

    Cresswell, W. & Whitfield, D.P. (2008) How starvation risk in Redshanks

    results in predationmortalityfrom Sparrowhawks. Ibis, 150, 209218.

    Elgar, M.A. (1989) Predator vigilance and group size in mammals and birds: a

    critical reviewof the evidence.Biological Review, 64, 1333.

    Elgar, M.A. & Catterall, C.P. (1981) Flocking and predator surveillance in

    housesparrows:test of an hypothesis. Animal Behaviour, 29, 868872.

    Fernandez-Juricic, E., Jokimaki, J., McDonald, J.C., Melado, F., Toledano,

    A.,Mayo, C.,Martin, B., Fresneda,I. & Martin, V. (2004)Effectsof oppor-

    tunistic predation on anti-predator behavioural responses in a guild of

    groundforagers. Oecologia, 140, 183190.

    Fitzgibbon, C.D.(1990) Why do hunting Cheetahs prefer male gazelles? Animal

    Behaviour, 40, 837845.

    Foster, W.A. & Treherne, J.E. (1981) Evidence for the dilution effect in the

    selfish herd from fish predation on a marine insect. Nature, 293, 466

    467.

    Funston, P.J., Mills, M.G.L. & Biggs, H.C. (2001) Factors affecting the hunt-

    ing success of male and female lions in the Kruger National Park.Journal of

    Zoology, 253, 419431.

    Ioannou,C.C.,Tosh,C.R.,Neville, L. & Krause, J. (2008)The confusion effect

    from n eural networks to r educed predation risk.Behavioral Ecology, 19,

    126130.

    Jackson, A.L., Beauchamp, G., Broom, M. & Ruxton, G.D. (2006) Evolution

    of anti-predator traits in response to flexible targeting strategy by predators.

    Proceedings of the Royal Societyof London Series B, 273, 10551062.

    Jeschke, J.M. & Tollrian, R. (2007) Prey swarming: which predators become

    confused and why? Animal Behaviour, 74, 387393.

    Kenward, R.E. (1978) Hawksand doves: factors affectingsuccess and selection

    in goshawk attacks on woodpigeons. Journal of Animal Ecology, 47, 449

    460.Krakauer, D.C. (1995) Groups confuse predators by exploiting perceptualbot-

    tlenecks a connectionist model of the confusion effect. Behavioral Ecology

    and Sociobiology, 36, 421429.

    Krause, J. & Godin,J.G.J. (1995)Predator preferences for attacking particular

    prey group sizes consequences for predator hunting success and prey pre-

    dation risk.Animal Behaviour, 50, 465473.

    Krause, J. & Ruxton, G.D. (2002) Living in Groups. Oxford University Press,

    Oxford.

    Krause, J., Ruxton, G.D. & Rubenstein, D. (1998) Is there always an influence

    of shoal size on predator hunting success? Journal of Fish Biology,52, 494

    501.

    Krebs, J.R. & Davies, N.B. (1981) An Introduction to Behavioural Ecology.

    Blackwell, Oxford.

    Lima, S.L. (1990) Evolutionarily stable antipredator behavior among isolated

    foragers on the consequence of successful escape. Journal of Theoretical

    Biology, 143, 7789.

    Lima, S.L. (1998) Nonlethal effects in the ecology of predator-prey interactions

    what a re the ecological effects of a nti-predator decision-making?BioSci-

    ence, 48, 2534.

    Lima, S.L. (2002) Putting predators back into behavioral predator-prey inter-

    actions. Trends in Ecology andEvolution, 17, 7075.

    Lind, J. & Cresswell, W. (2005) Determining the fitness consequences of anti-

    predation behaviour. BehavioralEcology, 16, 945956.

    Lindstrom, A.(1989) Finchflock size andriskof hawk predationat a migratory

    stopover site.Auk, 106, 225232.

    Milinski, M. & Parker, G.A. (1991) Competition for resources. Behavioural

    Ecology: An Evolutionary Approach (eds J.R. Krebs & N.B. Davies),

    pp. 137168. BlackwellScientific Publications, Oxford.

    Minderman, J., Lind, J. & Cresswell, W. (2006) Behaviourally mediated indi-

    rect effects: interference competition increases predation mortality in forag-

    ing redshanks. Journal of Animal Ecology, 75, 713723.

    Mooring, M.S., Fitzpatrick, T.A., Nishihira, T.T. & Reisig, D.D. (2004) Vigi-

    lance, predation risk, and the allee effect in desert bighorn sheep.Journal ofWildlife Management, 68, 519532.

    318 W. Cresswell & J. L. Quinn

    2011 The Authors. Journal compilation 2011British Ecological Society, Journal of Animal Ecology, 80, 310319

  • 8/13/2019 j.1365-2656.2010.01775.x

    10/10

    Neill, S.R.S.J. & Cullen, J.M. (1974) Experiments on whether schooling by

    their prey affects the hunting behaviour of cephalopods and fish predators.

    Journalof Zoology, 172, 549569.

    Newton, I. (1986) The Sparrowhawk. Poyser, Calton.

    Ohguchi,O. (1978)Experimentson theselection against colouroddity of water

    fleas by three-spined sticklebacks.Zeitschrift fur Tierpsychologie-Journal of

    Comparative Ethology, 47, 254267.

    Owen-Smith, N. & Mills,M.G.L. (2006)Manifoldinteractiveinfluences on thepopulation dynamics of a multispecies ungulate assemblage. Ecological

    Monographs, 76, 7392.

    Parrish, J.K. (1993)Comparison of thehunting behavior of 4 piscine predators

    attackingschoolingprey. Ethology, 95, 233246.

    Pulliam, H.R.(1973) On the advantages of flocking. Journalof Theoretical Biol-

    ogy, 38, 419422.

    Pulliam, H.R. & Caraco, T. (1984) Living in groups: is there an optimal group

    size?Behavioural Ecology an Evolutionary Approach, 2nd Edition (eds J.R.

    Krebs & N.B. Davies), pp. 122147. Blackwell Scientific Publications,

    Oxford.

    Quinn, J.L. & Cresswell, W. (2004) Predator hunting behaviour and prey vul-

    nerability. Journalof Animal Ecology, 73, 143154.

    Quinn, J.L. & Cresswell, W. (2005a) Escape response delays in wintering red-

    shank, Tringa totanus, flocks: perceptual limits and economic decisions. Ani-

    mal Behaviour, 69, 12851292.

    Quinn, J.L. & Cresswell, W. (2005b) Personality, anti-predation behaviour and

    behavioural plasticity in the chaffinch Fringilla coelebs. Behaviour, 142,

    13771402.

    Quinn, J.L. & Cresswell, W. (2006)Testing fordomains of dangerin theselfish

    herd: sparrowhawks target widely spacedredshanks in flocks. Proceedings of

    theRoyal Societyof London Series B, 273, 25212526.

    Ripple, W.J.& Beschta, R.L.(2006) Linking wolves to willows via risk-sensitive

    foraging by ungulates in the northern Yellowstone ecosystem. Forest Ecol-

    ogyAnd Management, 230, 96106.

    Roberts, G. (1996) Why individual vigilance declines as group size increases.

    Animal Behaviour, 51, 10771086.

    Roth, T.C. & Lima, S.L. (2003)Hunting behavior and diet of Coopers hawks:

    an urban view of the small-bird-in-winter paradigm. The Condor,105, 474

    483.

    Roth, T.C., Lima, S.L. & Vetter, W.E. (2006)Determinants ofpredation risk in

    small wintering birds: the hawks perspective.Behavioral Ecology and Socio-

    biology, 60, 195204.

    Sansom, A., Lind, J. & Cresswell, W. (2009) Individual behavior and survival:

    the roles of predator avoidance, foraging success, and vigilance. BehavioralEcology, 20, 11681174.

    Sansom, A., Cresswell, W., Minderman, J. & Lind, J. (2008) Vigilance benefits

    andcompetitioncosts in groups: do individual redshanksgain an overall for-

    agingbenefit? Animal Behaviour, 75, 18691875.

    Schmitz, O.J., Krivan, V. & Ovadia, O. (2004) Trophic cascades: the primacy

    of trait-mediatedindirectinteractions. EcologyLetters, 7, 153163.

    Schradin, C. (2000) Confusion effect in a reptilian and a primate predator.

    Ethology, 106, 691700.

    Sutherland, W.J. (1996) From individual behaviour to population ecology.

    Oxford UniversityPress, Oxford.

    Team, R.D.C. (2009) R: A Language and Environment for Statistical Comput-

    ing. R Foundationfor Statistical Computing, Vienna, Austria.

    Vine, I. (1973) Detection of prey flocks by predators. Journal of Theoretical

    Biology, 40, 207210.

    Watson, M., Aebischer, N.J. & Cresswell, W. (2007) Vigilance and fitness in

    grey partridges Perdix perdix : the effects of group size and foraging-vigi-

    lance trade-offs on predation mortality.Journal of Animal Ecology,76, 211

    221.

    Werner, E.E. & Peacor, S.D. (2003)A reviewof trait-mediated indirect interac-

    tionsin ecologicalcommunities. Ecology, 84, 10831100.

    Whitfield,D.P. (1985)Raptorpredation on wintering waders in southeastScot-

    land. Ibis, 127, 544548.

    Received 1 April2010;accepted18 October 2010

    HandlingEditor:Tom Webb

    Optimal group size and predation risk 319

    2011 The Authors. Journal compilation 2011British EcologicalSociety, Journal of Animal Ecology, 80, 310319