THE INFLUENCE OF GROUP SIZE AND NEIGHBORS ONVIGILANCE IN TWO SPECIES OF ARBOREAL MONKEYS

by

ADRIAN TREVES'''>

(Dept. of Amhropoiogy, Harvard University, Cambridge, MA 02138, L)SA)

(Acc. 5-VlIl-l997)

Summary

ln theory, onc of the main benefits of group-living is lhe sharing of vigilance among group-males. Howevcr, data on scanning in redtail and red eolobus monkeys indicate that only oneclass of individuals in each species derives clear benefits from shared vigilance. Moreover,the expected negative relationship between individual scanning and social group size wasnot mei in thesc monkeys. Nor was time spent scanning influcnced by the sex or speciescomposition of groups. Shared vigilance was observed only among red colobus adult malesand redlail adult females and only when lhey had ncighbors within 2 m. Red colobus adullmales savcd 10% of their scanning time when lhey had onc ncighbor within 2 m, whileredtail adult fcmaIes saved 16% of their time under the same conditions. No other age-sexclass demonstrated a significant decrease. <

The role of ncar ncighbors has been underemphasized in previous work on grouping andvigilance, an oversighl made more serious bccause of the often confounded relaiionshipbctwccn spatiaI cohesion and group sifce. In redtails, but not in red colobus, the number ofncighbors within 2 m was signih'eantly positivcly correlated with group si/e. This promplslhc hypothesis that Ihe inconsislency and poor explanatory power of group si/e in sludics

" Current address: Department of P.sychology, University of Wisconsin, 1202 West John-son Strect, Madison, WI 53706-1696, USA; e-tnail: atreves@facstaff.wisc.edu'' The field re.search was conducted with the permission of the Makercre University Bio-logical Field Station and ths Nalional Rcsearch Council of Uganda (EC 281). The authorwas supported by a National Sdencc Foundation Graduate Pellowship, Sigma Xi GlAR,Boise Fund Grant, Nalional Geographic Research Grant and Mellon Foundation Grantsfur lieldwork and write-up. 1 wouId like to thank C. Chapman, M. Hauser, M. Leighton,L, Naughton and R. Wrangham for helpful advicc throughout the project. This work couldnoi have been completed withoul thc help and goodwill of Pastal Baguma Araali, LisaNaughton and Isaac Treves. lnvaluablc logistical support was provided by lhe KibalcChimpanzee Project.

<0 Koninklijke Brill NV, Leiden, 1998 Rehaviour 135. 453-4S!

454 ADRlAN TREVES VIGD_ANCE IN ARBOREAL MONKEYS 455

of vigiLancc may be due to the underlying and undetccted rolc ol' neai neighbors, Theparamount role oF overall group size in optimality models is therefore questioncd,

Keywords: vigilance, arboreal primates, group-living, neighbors.

Introduction

Compared to solitary animals, individuaIs living in groups are believed tosuffer lower rates of mortality from extra-group attack (Hamilton, 1971;Goss-Custard et al, 1972; Edmunds, 1974; Curio, 1976; Morgan & Godin,l985; Fitzgibbon, 1990), Extra-group attack often derives from predatorsbut the benefits of grouping may apply to interspecific and conspecificcompetition as well (Wrangham, 1980; Young, 1982; Watts, 1989; Treves& Chapman, 1996). Group-living provides protection via several possi-ble mechanisms, including improved detection of outside threats (Galton,1883; Powell, 1974; Siegfried & Underhill, 1975; Hart & Lendrern, 1984),reduction in the per capita probability of being selected as a target (Hamil-ton, 1971; Morgan & Godin, 1985; Landeau & Terborgh, 1986), and/orimproved deterrence of attackers (DeVore & Washbum, 1963; DeVore &Hall, 1965; Klump & Shalter, 1984; Maisels et at., 1994).

Existing models of per capita benefits of group-living argue for increas-ing benefits with increasing group size up to a point beyond which incre-mental benefits do not counteract the costs of added members (Terborgh,1983; Terborgh & Janson, 1986; Quenette, 1990). In this view, group sizeis the key to benefits of sociality. Thus, the literature on the evolution ofgrouping usua!ly treats overall group size as the unit upon which theory isbuilt (Pulliam, 1973; Caraco, 1979; Elgar & Calterall, 1981; Sibly, 1983;van Schaik, 1983; van Schaik & van Hooff, 1983; van Schaik et aL, 1983;Pulliam & Caraco, 1984). Yet, the influence of other aspects of socialstructure have been neglected as a result.

Vigilance research provides a means to examine the importance of groupsize. Twenty-five years ago, Pulliam {1973) argued that individuaIs couIdreduce their own investment in scanning through reliance on the vigilanceof other members of foraging groups. Shared vigilance could thus providecontinuousbenefitsintheabsenceofathreat(Elgar, 1989; Quenette, 1990).The central role of foraging group size was tacitly accepted in ensuingyears, as study after study demonstrated that individual vigilance declined

with increasing group size — hereafter Uie group-size effect — (Berger,1978; Caraco, 1979; Bertram, 1980; Lipetz & Bekoff, 1982; Lendrem,I983a; Alados, 1985; de Ruiter, 1986; LaGory, 1986; Lendrem ei al.,1986; Jarman, 1987; Lima, 1994). However, in 1989, Elgar drew attentionto several careful studies which failed to detect a group-size effect (e.g.impala, Aepyceros melampus: Underwood, 1982; bald eagles, Haliaeetus[eucocephalus: Knight & Knight, 1986; teal, Anas crecca; Pb'ysa, 1987a,b), and criticized the lack of statistical and experimental control in moststudies of vigilance (Elgar, 1989).

Recently, group size and other social stracture variables have been com-bined in multivariate tests searching for determinants of vigilance. Suchanalyses of vigilance in birds and primates revealed no group size effect,especiaUy when confounding variables were controlled (Cords, 1990; Cat-temttetal., 199^;Bekoff, 1995;Rose&Fedigan, 1995). Asimilar,recentstudy of mammalian vigilance uncovered a slight effect of group size incertain activities and not others {Blumstein, 1996). These studies uncov-ered several attributes of groups which couId confound group size, suchas group geometry (Bekoff, 1995), group composition (van Schaik & vanNoordwijk, 1989;Rose&Fedigan, 1995),anddensityofneighbors(P6ysa,1987a).

An example of this confound is an early study of vigilance in two groupsof wedge-capped capuchins (Cebus olivaceus]. Differences in group sizewere identified as the source of differences in vigilance without consid-ering densities of neighbors or differences in numbers of adult males (deRuiter, 1986). Yet, later studies of white-faced capuchin monkeys (Cebuscapucinus), reported that anirnals in large groups had more neighbors thanthose in small groups (Phillips, 1995), and the number of adult males weremore effective than group size in predicting vigilance (Rose & Fedigan,1995).

Li fact, density of neighbors or spatial cohesion within groups may ac-count for the poor explanatory power and inconsistency of the group sizeeffect in cross-taxonomic comparison (Elgar, 1989). In species which in-crease spatial cohesion with increasing group size, one would expect to findthe predicted group size effect. In contrast, species that maintain consistentinter-individual spacing would not display a group-size effect because thedensity of neighbors would be invariant across a large range of group sizes.

456 ADRIAN TREVES VIGTLANCE TN ARBOREAL MONKEYS 457

While researchers have documented that neighbor density influences vig-ilance(Lazams&Inglis, 1978;Robinson, 1981;Metcatfe, I984a),onlyonestudy has shown that group size and density of neighbors covary (Poysa,1994). Even with clear acknowledgment of the importance of spatial co-hesion, group size and the number or density ofneighbors are rarely con-sidered together (Roberts, 1995, p. 1371).

Accordingly, I designed my research to incorporate spatiai cohesionwithin groups, social group size and the composition, numericaI size andnumber of participating species in polyspecific associations. This paperexamines individual vigilance behavior of redtail, Cercapithecus ascaniusschmidtii, and red colobus monkeys, Procolobus badius tephrosceles, inrelation to several measures of aggregation. The associations betweengroup size, neighbor density and vigilance behavior are illustrated withcomparisons of several groups and through an investigation of polyspecificassociations. Specifically, this research sought to identify the proximatemeasure of aggregation (if any) upon which individuals based their de-cisions to invest in vigilance. The u1timate objective was to understandvariation in the individual beneflts of group living with a particular focuson shared vigilance.

Methods

Study site and animals

Redtail and red colobus monkeys of ihe Kibale National Park, Uganda, werc studied fromFebruary 1994 lhrough May 1995 at lhe Kanyawara Rescarch Station, siie of scveral previ-ous studies of the same species (Struhsaker, 1978, 1980, 1988; Struhsaker & Leland, 1979,1988; lsbell, 1983; Chapman & Chapman, 1996), fci the remainder of this paper I refer to'social group size' for the numbcr of individuals living in a singlc-species group and 1 referto 'numerical sizc ol' associations' for thc number of individuals in polyspecific associaiionsor in contcxts where spccies composition is not specified. Social group size among cercop-ithccid monkeys is a fairly stable characieristic showing minor fluctuations on a monihly orannual scaie due to emigration, immigration, birth and death (Struhsaker, 1975, 1977, I980;Struhsaker &Popc, I991). Yet, thcnumerical sizeofassociatk>ns fluctuates greatly overihccourse of a day as encountcrs wiih other arboreal specics lead to polyspecific associationslasting a few minules to 11 hours (Struhsaker, I981; Treves, 1997). Similar]y, clusters ofnear neighbors are highly labilc in the short term, because indepcndent movcments andactivities lake individuals into and out of proximity to cach other.

Kibale redtails are small (aduit fcmales 2.9 kg, adult males 4.2 kg) frugivore-insectivoresliving in groups averaging 25 individuals including one resident adult male typically (Struh-saker, 1988). Females are thought to remain in lheir natal groups all of tfteir lives (Struh-

saker & Pope, 1991}- Soliiary non-group malcs may obtain mating access (Jones & Bush,1988; see Cords, 1987, 1988 for review) or take over groups (Struhsaker, 1977; Jones &Bush, 1988). In this study, four bisexual groups and one solitary male rcdtail wcre sampled(Table 1).

Kibale red colobus (adult femalcs 5.8 kg, adult males 10.5 kg) are primarily folivorousand lbund in groups averaging 50 animals containing multiple adult males. Fcmales dispcrseat higher rates than malcs (Struhsaker & Pope, 1991). Solitary rcd colobus males are rare(Struhsaker & Leland, 1979j and none were sampled in !his study. Five groups of redcolobus were studied for this analysis; one group C split into SC and BC aftcr a fewmonths fTab!e 1).

Red co!obus and redtail monkcys are l'requenlly found in polyspecific associalions (PSAhereafter). In this sludy, PSAs were obscrved lo consist of up to 5 specics and numberedup to 109 individuals. Redtails and rcd colobus were each other's most frequent associationpartners (Treves, 1997). This conforms lo prior work at Kibalc (Struhsaker, 1981; Waser,1987).

The major predators ol" ihese primates arc chimpan/ees (Pan troglodytes schweinfurthii:Wrangham et al., J991), felids (Struhsaker, 1975), and raptors (Skorupa, 19R9; Struhsaker& Lcakcy, I990; Treves, 1997). Feral cats and dogs may pose a thrcat as well (pers. obs.;Lambert, pers. comm.). Human hunting has been eliminated from ihis siie for two decadesat least (Butynski, 1990).

Observational melhods

The aim was to quantify vigilance with respect to social conlexls of visual scanmng, Visualscanning functions in iniormation-gathering and therel'ore serves raany needs (includingfood search, dangcr detection, social monitoring, travel path planning). Although visualscanning can serve many needs, its intended usc in any one situation docs nol preclude itsutility for othcr functions (e.g. planning a travel path may reveal a predalor). Therefore,all scanning may serve in threat detection except that directed to lhe immediate vicinily ofthe scanner. Visual search devotcd lo food wilhin arm's reach or lo grooming are majorcategories of scanning which are unlike!y to detect threats.

To concenirate on sclf-protection, I defined vigilance as 'scanning directcd beyond arm'sreach'. Scanning directed within arm's reach differcd irorn lhai directed beyond arm's reachbecause it co-occurred with motion of the hands, raising and inspecting body parts, rapidadjustments of head position, and/or medial position of ihe pupils. Also, focal animalswcre sampled only ifthey traveled > 10 m in 60 s, and iftheir faces were '> 50% visiblethroughout lhe 60 s. Averaging across species, 9.3% ofrccords (SD 4.1%, range 1.6-21.6%)were aborted and discarded every day, as a result of the focal animal moving out of view,hiding its face or traveling > 10 m within 60 s.

Because scanning ofgroup-matcs can re!lectself-protection as wcll (Alberts, 1994; Yaber& Herrera, I994), no efforl was made to separate scanning directed at group-members fromthat dirccted elsewhere. Visual monitoring of group-maies may in fact, be part of ananti-predafor siraiegy (Lima, 1995a).

The exclusion of anirnals moving ^10 m per minute eliminated traveling as an aclivityand vigilance during travel. There is no means of estimating the effect of this on (hedata bccause sampling vigilance in animals lraveling through a dense canopy is lypically

458 ADRL^N TREVES VlGH,ANCE lN ARBOREAL MONKEYS 459

k.gouun

JJ

-&

U

to

"S

&,

^>

C.2--

lQ"*"i

yl-*^

" 1C

Ql

(j ^^^

t-i j|T,

_'O '- '

V. f:^J> C

C Q3 "*-*

^&1-SCi,

E-2'?olo

" 3

a

S

l— -*S

3-:U3

EQ

H

cO2oQ.

Sou

uuB,

t/^

u

-

-On3J

E9C

<

a.

S.tsjj

3oH

6E

c c.2 o

oo.

cU

l<

Q*- "P,J3

2fib13'Go

o"ou^"

1J TJ "O^tj cj o u_ T3 -q j^ J^" xj' ' r- j: m' ^ rf a c*--

o% ^^ O ^ v ^t O - ' C C O %E^ - ' CO f^ - 1[- 1T|M

c j a > * 3 * ^ ' f A E ^ K I [ ^ T ^ ^ -Oi r ^<M [N r-^>n-i" t f

^ r ^ C ^ ^ ' - ^ c O O ^ ' OT- - r ^ rN r - .u - j f -Jcn^D

o o

^ r v * i T t ^ - o t * i ^ t ^ c * orj oi rJ

'O

i- r-li- 5 fN^ - - o

m "/"_ _ ^ , c - i r - - - (^ r- *t vi w

^P^. -^ ^ -* R" ^Q - C L ^ _ t - i - - Q , i ^ C _ 3

= ^ ^ = ^ S ^ 2 2 " 5^ o ^ P P o u b o ^ t

u & , m ^ S u c a ^ a - ^ o

>>

l |' —5 <J

ri uc^ c^

lfl

T3^J

fls-ooTJTJ.s<u

- -J

3"o >i_g "Sg Sfli "Osjj ajra ti.

^Z- 2*" 5v^E "<^ ^E ^J ^

ll <J

G 'S

llc

^ "S

3 S-o --<^ ^n ^p

u- -^u? "2qj cg

- -^s z,P r^

-3 Oa rj

, Wll "

rr^ "<C *

^*-2tto -2

! ! .i iP P " ^Si ^ -S S>

1 1 3 3SJ C -Q -^J^ y -S -3^ ^ "S "5u O O O

-C.

uT ^J

^ l' ^ -lE 'H "> "i *

=5 il^l| M S "g ZS -3 3 -O nc ^ ^ <u ^

< ll U ll llf- ^ ^O O "C

S^j

l3

-C)

l!

y-SDO

u^isu>.inu

impossiblc. Yet, the theoretical expectaiions about vigilance and aggregation apply tol'nraging and resting, so few siudies examine vigilance in travcIing animals.

Continuous focal saraples lasting 60 s were collected. Whcn multip!e animals werevisiblc, thc foilowing decision niles were taken: (l) anima]s within 2 m of hcterospedficswcre chosen bel"ore alI othcrs (lhis occurrcd in < l% of samples); (2) animals sampledpreviously were bypassed in favor of those noi sarapled in the last hour; and (3) in order oi"decreasing preference — females with mfanis, aduli l"emales without infants, adult malcsand immaturcs wcrc sclcctcd for sampling. ln practice. the poor observation tonditionsrarcly providcd muLtiple focals, so the same animal was someiime& samplcd lwo or moretimes in quick succcssion. All such seqtiences of records were averaged het"ore cnteringinio anaIyses. The $econd rulc rcquircd that thc obscrvcr move frequently, regardless oi'the group's movement, lo sample as many group-mcmbcrs as possiblc per hour. The lastrule compensated for any iendency to sample the more habituaied or lcss shy aniinalsdisproportionatcly. After each 60 s sample, the focal animal's age-sex class (adult male,adtilt female, immature) and activity were recorded (resting. foraging, .wciali/ing, oiher)with its spatial position as described below.

Scanning bchavior was studicd iising continiious observation during the 60 s focal sample(Metcalfe, 1984a, b). A timer was started as soon as a suitable focal animal was chosen,rcgardlcss of whcthcr thc animal was scanning. A bout of vigilance began when the focalanimal intcrruptcd visual cngagcmcnl in onc activity and started to scan. A bout cndcdwhcn scanning was interruplcd, in lum, by visual cngagcmcnl in any othcr activity. Verybrief glances cuuld be recorded wilh this method unlike other studies using mslantaneousscan samples (de Ruiter, 1986; van Schaik & van Noordwijk, 1989| or duration criteria todefine scanmng fRose & Fedigan, 1995].

While the timer counted out 60 s, a separate stopwatch was used to measure time spentscanning. Observation was continuons so individual boutlengths could not be measurcd.Frequcncy of scanning was not uscd in this analysis bccausc it docs not takc into accounichanges in length ol" bouls (Melcall"e, 1984a|, nor is il a suffidenl measure ol" temporalinvestmeni when bouts vary in duration (E!gar & Caiterall. 1981).

Using a short sampling pcriod (60 s) reduces thc number of abortcd samplcs. Howcvcr,it increascs lhc risk of truncating bouts. As lhc sclcction of focals was madc indcpcndentof aclivity. indtvidual houLs which spanned !"ewer than 6U s wcrc samplcd wilhoul bias (i.e.samples were equally likely to interrupt a bout as ln siart a bout laie, on average coincidingexactly). Singlc bouts which lasted more than 60 s were not sampled in an unbiased maiinerby this methnd. Late in ihe study, I assessed the frequency of bouts exceeding 60 s bynieasiiring the lull duration of cach such bout. Fewer lhan l% of the samples used in thisanalysis containcd single bouts exceeding 60 s in duration; thc avcragc 'e*cess' durationtruncatcd by my methods was ! i .s. This trans!ates to an error of 0.22 s for the estimationof average time speni scanning.

Con:e.xiual variablcs

After a local samplc was completed, the conlext of the focal animal at the end of lhe .samplewas recorded. Contextual information included agc, sex and prcsence of a clinging infant,as wcll as spatial vanables iisted below. Consiitcntly diiTerentiating mothcrs from femaleswiihout clinging infants was not possiblc for redtails bccause of infant lransfcr. Rcd colobusdid not show infant transfcr (Stnihs;Aer & Oules. 1975; Struhsaker, 1980; Treves, I997).

460 ADRIAN TREVES VIGU-ANCE !N ARBOREAL MONKEY5 461

In an effort to describe ihe focal animal's position relaiive to the position of other groupmembers (associates), [ defined ccntral posiiion as any individual with an associate in eachof lhe following five positions: above, frotit, rear, and on both sidcs. To be above the focalanimal, an associate had to be posiiioned higher in thc same tree, lf ihe focal animal hadan associate matching ihis criterion, thcn the four other positions were scarched, Thesepositions had to be separaied from each othcr by more than 60" azimuthally, bui could beseveral meters higher or lower than thc focal animal (i.e. so they werc in approximately thesanie horizonial plane). Early in lhe project, the criterion ofhaving an associate below thcfocal animal was eliminaied because practically no focal animal could be coded as ccntral,

Thc numbers, specics and classes of neighbors within 2 m of the focal's torso wererecorded. In addition, an attempt was madc to rccotd the number of ncighbors within 5 mand within 10 m. Bccause ofdense foliagc, 85% of all samples failed to obtain the numberofneighbors bctween 5 and 10 m. Ovcrall, 65% of.samples failed to obtain the numberofneighbors bctween 2 and 5 m. Fewcr than 10% of samples failed to count neighbors within2m.

In terobserver reliab ility

ln fewer ihan 10% of samples were both observcrs sampling from the samc group; in nocase was the same focal animal the subject of simullaneous or sequential sampling by thetwo observers. Interobscrver reliability tests wcre run four times over the course of thestudy. Each test consisted of a series of 20 independent. continuous 60 s samples takcnsimultaneously by both observers on the same animal without consultation until the endof the test. For this analysis, measurcs of bouts of scanning, time spent scanning, numberof ncighbors ai 2, 5 and 10 m, peripheral/central and group slaius (polyspecinc or not)wcre assessed for reliability. Assessment of reliability was made by direct comparison notcotrclation, Whenever a speciflc mcasure deviated by I0% or more between observers, allthe data collected on that measure in the preceding months were discardcd l"rom the datasct.In fact, after the first tesl, no data were discarded except for ncighbors between 5 and 10 mwhich continued to be unreliable.

Groupitt$ and association

Social group sizcs were determined sevcral times over the course of lhe study. Rcsults ofthe previous compleie count were uscd to assign a social group size to a given focal sample.Social group si/.es werevery stable amongtheredtails. The adultmembershipofrcdcolobusgroups was very stable but old juvenile and subadult counts nuctuaied to a marked degreeas has bccn rcportcd in a previous long-term study (Struh.saker, 3975, p. 14). Average socialgroup size valucs are given in Table 1. Polyspccific associations were rccorded whenevcrhcterospecific monkcys wcrc obscrved wilhin 20 m of each other (Slruhsaker, 1981) orgroups were intermingled. The nuraerical sizc of polyspecific associations was estimatedwhencvcr unknown groups participated hy including the average value for that species atKibalc O^aser, 1982, 1984).

Te.sts of thc cffect of group sizc were pcrformed in two ways. The l"irst examincd cachgroup (or solitary animal) individually and dctcrmincd whethcr avcragc timc spcnt scanningchanged as numericaI size changed (increasing from single-species to various sizes nfPSA).The second examined time spent scanning as a function of social group si^e among lhe

known groups, These methods reduce the risk of inflaiing sample sizes or agglomeraiingheterogcneous data.

To as.sess if PSA membership influenccd vigilancc, three questions were askcd. Didlimc spent scanning in single-species or monospecific groups (MSG hereaficr) differ fromihat spcnt in PSAs regardless of size or compo.sition? Did thc change in number of speciesor identjty of specics influence time spent scanning? Did the change m numerical size ofassociations infiuence time spent scanning? These questions were answered by averagingtime spent scanning prior to a change and comparing lhis average to the average nmespcnt after a change in association state (Roberts, 1995). Each association ihus provides amatched pair of before and after samplcs for each age-sex class. The resulting before-and-afler pairs control for variation between groups, associations and observers. The entire seiof before-and-after pairs was tested against species membership, association size, numberof species per association and a dichotomous PSA or MSG variable.

Samples used in analysis

The data used for*his study consist of baseline vigilancc samples collected from thc mainstudy groups (Tablc 1). Ba.seline samples cxclude all rccords coiIected up to two hoursafter poientially threaiening events. Thus, the data do not include aclual evasive actionor scarching bchavior which may follow danger. Shared vigi]ance shouId operate in theabsence of imminent danger (e.g. Lendrcm, 1983a), or it cannot provide thc theorizedbene!its (Pulliam, 1973; Lendrcm, 1983a, b).

The most severe thrcats (raptor attack and chimpanzee sightings) iriggcred obvious alarm,escape or concealmcnt for no longer ihan 45 minules once the threat disappeared, Thus,the cnterion of two hours treats data con.servatively by discarding samples taken longafter a potcntially threatcning event. The entire set of poientially lhreatcning events wasdcfined a priori as any of the following: encounters with prcdators (whcther auditory orvisual), intergroup encountcrs, interspecific agonism, loud sounds made by mammals orbirds; or rare evcnts (earthquake tremors, treefalls, hailstorms, loud breaking branches).I( is reasonablc to expect that the datasct contains a number of samples coJlccted after asignilicant event wcnt unnoticed by the obscrver(s). However, lhe cvents above lendcd toco-occur with voeal rcsponses of the monkeys, providing some assurance that the mostintciiEe events were excluded. ln addition, all samples lakcn during thc day in which acopulation was observed wcre discardcd from ihc present analysis, in ordcr to reduce thebiasing cffects of reproduciivc compelition and incrcased numbcrs of cycling fcmales inlarger groups.

individual identification

The red colobus sub-population under study contained approximately 29 aduIt malcs and 84aduIt fcmaIes in 5 groups. Thc rcdtail sub-population contained approximately 7 adult maIesand 53 aduIt females in 4 groups (Table 1). Bccause this study aimed io samplc behaviorfrom many differeni groups of monkcys, individual identification ol the study anJmals wasseeondary, This might present a problem if thc dataset bccame dominated by observaiionsof one individual or a subsel of individuals (e.g. the mosi habituated individuals). Not onlywould lhis iiitroduce problcms of slatistica! dependcnce but it might weaken the generali!yof the findings. Sevcral melhodological steps wcre taken to avoid biased sampling. First,

e.3'O Mtf, n5! oct nn

l *'

3- "

a I3l'^ g"K) s" n> 3

& 3 -^ ro

ll

cr e^

*^ w?rt n>n t"ft pW J

sr o.

ll" SSi ftCD KJ

S &'iL rt

dala. Tw

o-tailed tests arc used lhroughout.critcrion for signilicance is sct at (J.025 to com

pensate for the number of tests run on the

A m

ultivariale test using number of neighbors

at two diffcrenl distances reports t values

for thc Eignificant variables as well as overall significance of the regrcssion

modcl.

The

Non-param

eiric lests werc used to assess dilT

erences duc to group sizc and neighbors.of 53 for Statistical tesls. T

his conservative technique was used for all individual classes.

n>Mps

"S._FT

4-t>Jto

u

-iHT'f

pJpa3o-*.<JIUJ

r?o.E,e?B.E_

P3P&

CQ"la'Po

r,Q.SPO.UHp

-3,CI

Mn

Statistical analysis

To avoid inflating thc significance lcvels, degrees of lreedom

werc reduced to the sam

plesize of individua!s in each class rathcr than to the sam

ple size of data points collctled. For

BT|" fT

!J5 S|-IfS. o.w c"

tjf

E.

o-iH

On:

O- .3-o

<JQ-nOC-a!/i

lCL

noa-fB- 1C6tragS.s.gisrtng,

l

have had difficulty in consisiemly positioning lhem

selves away from

lhe observer, Second,

thc class of individuals (mothers w

ith infanls) least comfortable w

ith observation by humans

was targeted deliberatcly,

Third, obscrver

distance varied from

5-8

0 m

(mean 28 i

7 m)

from the focal anim

al and the subjccts

were

not always aw

are of thc observer. T

hus,shy anim

als could bc sampled w

hile unaware of thc observer or at a suflicient distance io

diminish any tendcncy to hide. Finally, the am

ount of time spent sam

pling the major study

groups (rcdtail C, F and red colobus P, SC

, BC

) was purposefully

equilibratcd lo reducethe possibility that one group's behavior obscured

contrasting trends of other groups.T

o determine if lack of individual recognition affccts conciusions from

this study, onem

ust first determine

if the few know

n individuals wcrc sam

pled proporlionately to ihcirrepresentation in lhe population.

For thc 53 redtail adult fcmales

of my study groups,

proporlional sampling required 20.4 sam

plcs per female.

Thc lhree know

n females (P, C

,B

N) w

ere sampled an average of 18 tim

es in baseline data (range 7-34). Thc sam

e analysisfor redlail m

ales was unnccessary

as thcy were readily distinguishable from

other groupresidents (see T

ahle l; the lw

o males of group B

differed in body size and

appcaranceof tesliclcs).

Few baseline

samples of know

n red colobus were available

as individualidentification w

as difficult for females and confirm

ed only late in the study for males.

The

three well-know

n adult males (B

L, G

O, B

F) were sam

pled an avcrage of 24 timcs (range

9-55) com

pared to the cxpected 30 sam

ples. T

hus, for red colobus

l cannot reject the

claim that know

n individuals werc sam

pled disproportionately. B

ecause disproportionatesam

pling of a subsei of individuals can reduce lhe generality of the findings if the subseibehaves differcnily lhan the population from

which it w

as drawn, l take onc step to im

provegeneraliiy.

Whenever possible, results w

ill be assessed in each group separately. T

hus, a

in order to nole all occurrcnc:es of polyspecific; associations. A

s a result, individuals would

thc observers made efforts to circle and canvass the group(s) under study every half hour,

46

2

AD

RIA

N

TR

EV

ES

TABLE 2. Scanning and neighbors among red colobus and redtail monkeys

Species

Red colobus

Redlails

lndividual class

adult malesfemales without infantsmothersimmalures

adull malesfemales wilhout infantsfemales with clinging infantsimmatures

Time spent scanning (s):mean i l SD

36.4i 17.323.2i I4.324.4i 15.729.1 i8.7

38.7i 13.837.9i 13.938.7i 14.141.3i 12.7

Neighbors within 2 m:mean i l SD

0.9 i 0.61.3i0.81.9i 1.00.9 i 0.6

0.4 i 0.50.5 i 0.50.6 i 0.60.6 i 0.6

Sample size*

557497592

51

587909173245

<c

9

too

>

Q

Number of baseline reeords of individuals from known groups.

4-

ADRIAN TREVES VIGiLANCE OV ARBOREAL MONKEYS 465

neighbors among adult classes. Redtails spent more time scanning than didred colobus.

Group size and time xpenl scanning

Among red coiobus, time spent scanning did not consistently dechne withincreasing numerical size of associations for any of the groups of redcolobus CFig. 1)- Th^ only downward trend (P group) was not signifi-

Time spcntscanning

Tirae spenlscajuring

*#J

30-

2O

IO

0-<

/flBC group

O 70 TO

Assoeiation Size

30

3O

101

O

C group

75 80 9C I00 110

Assotiaucn Size

4f>

SC Group

50 70 90 110

Assodation Si7c

IO 30 50 70

Associadon Sizc

Fig. 1. Time spent scanning (s per minuie] as a function of numerical size of associaiionin 4 groups ol" red colobus. Observations consist of adults cngagcd in all aclivides exccpttravelling. The leflrnost poinl in each panel is the soriaI group size. Group 66 is excludcd

hccause no baseIme samples were collected of ii in PSA. Error bars = 1 SE.

cant (Kendall correlation r = -0.19, p - 0.46). Taking all single-speciesgroups yielded no association between social group size and time spentscanning (N - 5, r - -0.33, p = 0.21). Time spent scanning by the twoadult male red colobus, known throughout the study (BL, BF) revealedno signincant effect of numerical size of association either {K-W df - 5,H = 3.5, p = 0.62; df = 2, H = l.6. p = 0.66). Because the number ofadult males correlated perfectly with social group size (r - 1.0, Table 1),there appears to be no influence of male number on vigilance among redcolobus.

Similarly, among redtails, no effect ofnumerical size of association couldbe detected on time spent scanning [F group: N = lO group size classes(29, 39, 41, 44, 47, 53, 57, 62, 75, 78, 109 individuals), r = +0.16,p = 0.53; C group: N = 9 (18, 22, 27, 32, 50, 54, 60, 65, 77 individuals),r = +0.44, p = 0.10; B group: N = 5 (22, 23, 32, 41, 85 individuals),r = -0.40, p - 0.33). Comparing the four groups to each other in single-species grouping, yielded no relationship between social group size andtime spent scanning (df - 3, r - 0.50, p - 0.09). Although the effect isstrong, the direction of correlation is positive, contrary to predictions. Thefive well-sampled individually identified redtails provided enough data toassess if scanning time changed with numerical size of association (tworedtail females with old radio collars: Pu N - 6 associations, z - 1.3,p = 0.18: Coll 7V = 5, z = 0.01, p = 0.99; the two resident males Co(JV = 9 associations, z = 0.8, p = 0.40) and Fo (JV = 227, z = -0.5,p = o.59); and the solitary male MM (N = 3, H = 2.5, p = 0.28) did notsignificantly alter time spent scanning in PSA.

Foraging/feeding and resting were the rnost common activities. Feedingand scanning are usually seen to conflict more than resting and scanning. Ifgroups of different sizes spent different amounts of time feeding, group sizedifferences could be obscured. There was no evidence that group identityinfiuenced the proportion of samples in which an animal was observed tobe feeding (redtails: G-test = 0.2, p = 0.77; red colobus: G-test = 9.6,p = 0.21). Results of group level analyses were changed little by theseparation of data by activity or age-sex class of the focal animal. Whensplit by age-sex class, the only significant result came from adult male Cowho increased time spent scanning when resting, as the numerica! size ofassociation increased (A^ = 8 resting samples, r = +0.73, p - 0.016).

466 ADRlAN TREVES VTGILANCE IN ARBOREAL MONKEYS 467

If membership in a PSA influenced vigilance, one would expect that timespcnt scanning would change significantly following a change in associa-tion state. The samp!e size used here is thus the number of before-and-aftcrpairs. Among red colobus, the average time spent scanning was not associ-ated with the change in the number of other primate species in association(all classes pooled: K-W, df = 4, H = 4.6, p = 0.33), the change insize of the association (adult males ^V = 35; adult females without infantsN = 42; mothers N = 31; Kenda!l test: 1.4 z ^ -0.8, p > 0.15for all classes), or in response to being in a polyspeciflc association com-pared to a single-species group (aIl classes pooled: A^isg - 40, N^ - 68,t - -0.11, p - 0.90). Redtails showed no effects of polyspecific associ-ation either. Change in the number of species involved in a PSA had noeffect on time spent scanning (df - 4, aduIt males N = 81, H - 4.0;non-mothers N = 99, H = 3.7; immatures N = 36, H = 5.7; p > 0.4 forall classes; females with infants could not be anaIyzed because of smallsample size, JV - 12). With one exception, the same hcld true for changesin association size (Kendall rank test: adult males z = 0.4, p = 0.66; fe-males without infants z - -0.1, p - 0.95; immatures N = 36, z - -1.7,p = 0.09). The exception was females with infants who increased timespent scanning as association size increased (N - 12, r - 0.6, z - 2.8,p - 0.006). A simple comparison of single-species groups with polyspe-cific associations regardless of size or composition revealed a weak trendtoward increased scanning in PSA, when all classes were lumped (MSGN = 15, PSA N= 153: z = -2.0,P = 0.05).

Finally, the identity of participants in PSA was analyzed to determine ifscanning was influenced by the species of association partner. Neither red-tails nor red colobus showed an effect of forming or dissolving associationwith any one of the other species in Table 1 (red colobus: K-W df - 3,forming H = 2.5, p = 0.47; dissolving H = 0.4, p = O.S2; redtails:df = 4, forming H = 3.9, p = 0.28; dissolving H = 2.7, p = 0.44).

In sum, the only significant correlation between scanning and numericaIsize of association or species composition was an increase in time spentscanning by redtails in PSA.

Neighbors and vigilance

Among red colobus males, time spent scanning varied with the number ofneighbors within 2 m (Fig. 2; df = 2, H = 10.3, p = 0.006). Thercfore,over the course of the average 10-hour day, a red colobus adult male whonever had a neighbor would scan for 62 minutes more than a male whoalways had one neighbor. In four of Ihe five red colobus groups, malesspent less time scanning when they had neighbors (Fig. 3). The exception(C group) had only one baselinc sample ofan adult male with no neighbors.This is also the group which divided early in thc study into BC and SCgroups.

Red colobus adult females failed to display a significani response tochanging numbers of neighbors (Table 3; df = 2, H - l.2, p = 0.55:females without-infants: df = 2, H = 3.5, p = 0.17).

The two well-known redtail adult males (Co, Fo) showed a tendency torespond to neighbors within 2 m (t - 2.2, p = 0.029). The solitary maleMM was recorded with a neighbor ooly once in the baseline dataset (evenwhen he was observed with othcr species).

Redtail females' scanning changed signiticantly in relation to neighbors{Table 3; Fig. 4: df = 2, H = I9.5, p = 0.0002). This pattern appearedlhe same for females in all 4 groups (Fig. 5). To make these resuhs moretangible, a redtail female with one neighbor within 2 m, over the course

1 2+# of neighbors

Fig. 2. The influence of neighbors within 2 m on time spenl .scanning ami>ng red colobusadult males (N = 32 individuals). Bars reprcscnt no neighbors (dolted colum), 1 neighbor(black column) and ^2 neighbcirs (hatched column). Samples sizes: no neighbors = I49rccords, one neighbor = 219 rccords, two or more neighbors = 180 rccords. Errorbars = 1

SE.

468 ADRlAN TREVES VIGFLANCE IN ARBOREAL MONKEYS 469

Time spcntscanning (s)

none ore or more

f of neighbors

Time spenlscanning (i)

none one or niore

i* of ncighbore

Timc spenlscanDiDg (s)

none one or more

# of neighbofs

2O

66groiJp

Jikone or more

of neighbors

none onc iH moret orncighbocs

Fig. 3. Timc spcnt scanning split by absence (dottcd columns) or presence (black columns)of ncighbors within 2 m, for red colobus adult males in 5 single-spccics groups. Samplesizes for the 5 groups: SC, N, = 27, N, = 66; BC JV. = 39, N? = I65; P JV* = 79.

7Vp = 153: 66 N, = 3, N^ = 9; C N< = 1, JVp = 6. Error bars = 1 SE.

of a 10 h day, would scan for 96 minutes less than a redtail female withno neighbors, on average.

Because of low sample sizes for measures of ncighbors beyond 2 m,the data from both species and all individuals are grouped together. By

TABLE 3. lnfluetice ofneighborswithin 2m on time spentscanning (me<miSD) among red colobus and redtail monkeys

Nuraber of non-infatit neighbors within 2 m

Red colobus

Redtails

adult malesmoihersfemales wiihoul infanls

adult malesadull females*

0

41.3 i 18.621.2i 17.923.0i I3.0

29.Sil2.640.9i 16-7

1

35.1 i 15.219.9i 15.319.8i22.6

28.1 i 13.33 l . 3 i20.0

2+

24.2 i 20.124.6il9.628.Si 18.9

-19.7 i 20.8

Mothers, females wiihoul infants and subadult females pnoled.

45 n

35 -

time speDtng (s)

1 2+

of neighbors

Fig, 4. Thc influcncc oI" ncighbors within 2 m on limc spent scanning among redtail adultfemales (A' = 53 individuals). Bnrs represem no neighbors (doUecI o>Iumn), 1 neightaor(black tolumn} and ^2 ncighbors (hatched eolumn). Samples si/es: no neighhors = 267records, onc ncighbor = 200 rccords, lwo or more neighbors = 76 records. Error bars -

1 SE.

converting values for neighbors and time spent scanning into deviationsfrotn the appropriate mean fbased on the species and individual class),one can analyze data from both species in one test without introducingexccssive distortions due to interspecific differences. When neighbors at2 m and 2-5 m are both cntered into a multiple regrcssion, the model foradult males explained 22% of the variance in deviations from the mcanfor time spent scanning {JV = 40, r^ = 0.22, F = 5.1, p = 0.011) and

470 ADRIAN TREVES VIGILANCE IN ARBOREAL MONKEYS 471

time speniscanning(s] perminute

50 -

40 -

30 -

20 -

10 -

1

C (19) B (21) F (29) U (?)

Redmi! Gioups

Fig. 5. Time spcnt scanning in relation to absence (dotted columns) and presence (blackcolunins) of ncighbors within 2 m, for redtail females in 4 single-species groups. Groupsizes provided in parenlheses on the x-axis. SampIe sizes for tIic 4 groups: B N* = 13,JVp = 15, C N, = 73, JVp = 53, F W. = 141, N,, = 138, U iV. = 40, Np = 70. Error

bars = 1 SE.

13% for adult females (JV = 89, r* = 0.13, F = 6.5, p = 0.002). Thesemi-partial correlation of the closer measure (2 m) was significant {adullmales: * = -3.2, p = 0.003; adult females: t = -3.4, p = 0.0009),without any signiflcant eftect of neighbors at 2-5 m. bi sum, neighborswithin 2 m appeared more influen(ial than neighbors between 2 and 5 min individual dccisions about scanning.

Confounds of spatial cohesion

Among the red colobus sociaI groups, there was a tendency for number ofneighbors within 2 m of focal anima!s to increase with increasing group size(df = 4, H = 13.3, p = 0.065). Coinparing social groups as they increasedin size with the formation of PSAs revealed no significant change m spatialcohesion, among red co!obus, (group C: MSG mcan = 2.1 i2.6 neighbor.s,PSA mean = 1 .Oi l .9 neighbors: Mann-Whitney z = 1.3; p = 0.21; groupP: MSG mean = I.li 1.2neighbors,PSAmean = 0.75i0.89neighbors:z = 0.2; p = 0.87).

In contrast, focal redtails in the largest group (29 individuals) had signif-icantly more neighbors (mean 0.5 i 0.4) than focal redtails in the smallergroups (!9 and 22 individuals respectively: mean 0.15 i 0.2; df - 2,H - 18.3, p = 0.003). Redtaiis also had significantty more neigh-bors when found with other species (MSG mean - 0.17 i 0.43; PSAmean = 0.53 i 0.72 neighbors; t = 4.8, p < 0.0001). This was true fortwo redtail study groups (B df = 28, t = 3.4, p = 0.002; F df = 90,t - 2.7, p - 0.008) and nearly so for the smallest group (C: df = 73,t - 1.8, p = 0.07). Although group size and neighbors covaried to acerlain extent, the relationship was not strong enough to produce a group-size eftect on vigilance. This highlights the need to examine inter-groupvariation in spatial cohesion carefully.

Peripheral positionNo class of individual in either species altered time spent scanning accord-ing to peripheral or central position. Yet, mothers who were peripheral hadfewer neighbors (N = 18, Central - 2.6 i 3.7 neighbors, compared ioN = 20, Peripheral = 0.95 i 1.6 neighbors, tf-tesl, p = 0.004). Central-ity was therefore confounded with neighbors for this class of individual.Future work must examine mothers' vigilance more closely.

Discussion

The present work joins several studies which challenge the long-standingclaim that timc spent vigilant decreases as group size increases (CatteralIet al., 1992; Rose & Fedigan, 1995). Therefore, it challenges the idea thatgroup size represents the basic unit of analy.sis upon which models of thebenefits of gregariousness must revolve.

For both species of arboreal primates, results ran counter to expecta-tions of existing predation theory. Individuals in groups associated withother species did not decrease time spent scanning when compared to thesame groups in single-species state. This was true regardless of spcciescomposition, size or number of species participating. Comparing single-species groups of different sizes revealed no effect of social group size ontime spent scanning. Nor did the number of resident adult males influenccvigilance for either species.

472 ADRlAN TREVES VIGU,ANCE iN ARBOREAL MONKEYS 473

However, the number of neighbors within 2 m was a significant pre-dictor of individual vigilance for one class of individual in each species.Neighbors beyond 2 m did not appear to influence vigilance when nearerneighbors were statistically controlled. Activity did not influence the re-sponse to neighbors within 2 m or to group size for adults ofeither species.

The sharing of vigilance observed among neighbors was clearest for redcolobus adult males and redtaii adult females. These two classes formthe stable core of their groups. Red colobus adult males form coalitionsthat defend groups from neighboring coalitions and maintain high levels ofproximity and affiHation with each other (Struhsaker, 1975, 1980). Femalesdisperse whereasjuvenile and subadult male.s rare!y do so; the recruitmentofnatal males into the resident coalition is believed to tead lo high related-ness among adult males. Similarly, redtaiI females almost never dispersefrom their natal groups, and thereby form the stable, related core of theirspecies' groups (Struhsaker & Pope, 1991). Males disperse and take-overor attain residency in groups as adults {Struhsaker, 1977; Jones & Bush,1988). Therefore, those individuals who have allies and kin among thegroup-members were observed to reap the benefits of shared vigilance.

Redtail males did not evince a strong response to variation in the numberof neighbors, reflecting perhaps the social!y and spatially peripheral roleoften played by males in this genus (Cords, 1988; Rowetl, 1988). Redcolobus females showed no response to neighbors either. Nevertheless, redcolobus mothers have, on average, 0.8 more neighbors than other adults(Tab!e 2). This suggests some other benefit of close association for mothersand/or their associates. Alternatively, neighbors may differ: wilh sometypes or individuals, mothers may share vigilance, while towards othersthey may be more wary.

Benefits ofgregariousness: shared vigilance

Saving time and effort through shared vigilance is a benefit which accruescontinuously, regardless of the presence of a threat. Tn other words, whenanimals are uncertain of the presence or whereabouts of threats, they mustmaintain a continuous investment in setf-protection; if this investment isshared with neighbors, they save time on a continuous basis. Other anti-predatory benefits of vigilance such as cooperative deterrence, improveddetection and dilution of risk provide benefits only when attack is im-

minent. In this light, the results presented here indicate thaI only oneage-sex class per ,species consistently benefited from group-living, whilethe remaining individuals did not derive the postulated minute-by-minutebenefits ofaggregation (Pulliam, 1973).

The absence of a group size effect on individnal vigilance accords withmost other primate studies. For example, data collected on white-facedcapuchins (Cebus capucinus) by Rose & Fedigan (1995) showed a correla-tion between scanning and the number of resident adult males, rather thangroup size. These data contradict information from a congener (Cebus oU-vaceus: dc Ruiter, 1986). Yet the latter sfudy did not consider the role ofneighbors, differences in numbers of resident adult and subadult males, thepotential problem with the use of frequency data alone, nor the potentialbias caused by excluding bouts of scanning lasting less than 5 s.

The present*analysis joins other studies (Poysa, 1987a, 1994; Cords,1990; Catterall et aL, 1992; Bekoff, 1995) in suggesting a re-evaluation ofthc theory relating group size to vigilance. The poor explanatory power ofgroup size in studies of vigilance in different taxa (Elgar, 1989) may resultfrom interspecific differences in (a) spatial cohesion within groups, (b) thedistance over which neighbors are effective in cooperative vigilance, or (c)the severity ofpredation pressure. Prior research had laid the groundworkfor an effective test of such an hypothesis, yet so many studies focusedon group size to the exclusion of other measures of aggregation that al-ternative hypotheses have only recently begun to emerge (E!gar, 1989;Bekoff, 1995). Overall group size was considered paramount even whenthe researchers were, in fact, observing neighbor clusters (Lendrem, 1984;Alados, 1985, 1986). In Lazarus' (1978J study of white-fronted geese(Anser albifrons), he demonstrated the change in density of neighbors withincreasing flock size, but failed to then determine if neighbor density wasresponsible for the observed decline in vigiiance in large flocks.

Overalf group size may be irrelevant if one is unable to monitor one'sneighbors (Lima, 1995a) or confirm their presence (Elgar ei al., 1984).Metcalfe (1984a) demonstrated that obstructions in visibility influencedshorebirds' vigilance; shorebirds did not lower vigilance with increasingdensity of neighbors when the foca! was surrounded by rocks or othervisual obstructions. Bekoff (1995) argued that the spatial geometry ofgroups detcrmines the efficiency with which group-members can monitor

474 ADRIAN TREVES VlGO.ANCE lN ARBOREAL MONKEYS 475

each other and therefore the effectiveness of shared vigilance, more or lessmdependent of group size. Visibility certainiy plays sonie role but thenature of its effect on shared vigilance is not yet clear.

The vigilance data from red colobus and redtails derive from animals in-habiting an environment within which cooperative detection may be ham-pered by a multitude of conspecific stimuli and limited visibility. It maybe that for the species studied here, the value of shared vigilance declinesrapidIy as distance increases between group-members due to the densityof foliage and complexity of the habitat. If habitat stmcture affect.s thedistance over which neighbors can engage in shared vigilance, one wouldexpect to find inter-specific (and inter-populational) differences iri the rela-tionship between group size and vigilance. Inler-populational differenceshave hardly been explored. For example, Holmes (1984) dcmonstratedthat the number of neighbors within 10 m of hoary marmots (Marmotacaligata) was associated with changes in vigilance, but Blumstein (1996)reports no such association in its congener (Marmota caudata aurea) livingin superficially similar habitats.

Alternative hypotheses

By identifying food density or dispersion as a relevant confound of groupsize, Elgar and his colleagues (Elgar et al., 1986; Elgar, 19S9) presented acase for an alternative hypothesis: clumped resources may lead to greaterconccntrations of foragers which, in turn, leads to rcet vigilance reductionsdue solely to increases in foraging rates as competition increases. Thisis a viable alternative hypothesis to the one presented here, albeit directlyrelated through spatial cohesion of group-members. The real differencebctween the two views rests on the relationships of neighbors: in Elgar'sview, neighbors are competilors for food, so vigilance decreases as eachanimal sets about harvesting rapid!y. fri the view presented in this paper,some neighbors are cooperative and vigilance decreases because of mutualreIiance and shared responsibi!ity. Resolution of this question will haveto rely on experimental evidence or observalion of gregarious individualsin and out of controlled feeding contexts to determine if neighbor densityexerts an effect independent of food dispersion. Studies of different socialstructures in similar contexts can also shed light on this question (Caine &Marra, 19SS).

Prior obstacIes in disentangling an effect due to overall group size fromone due to density ofnearest neighbors derive, in part, from thc difficulty indefining groups among fluid, short-iived aggregations ofbirds, upon whichmost studies of vigilance have been done (Lazarus, 1978; Metcalfe, 1984a,b; Metcalfe & Furness, I984; Beveridge & Deag, 1986). Additionally,group size and neighbors may covary or the distance ovcr which neighborsare itifluential may be as Iarge as the group spread, b!urring the distinctionbetween the two. In this study, numbers of neighbors wilhin 2 m andgroup size were correlated for one species and not the other. Group sizeand cohesion have been shown to covary in whi(e-tailed deer and o!herspecies (LaGory, 1986 for review).

The distance at which neighbors are no Ionger helpful as cooperativedetectors should be measured for a wider distribution of species. If thisdistance is as large as the radius ofthe group, then it may be appropriate tocollapse such a measure and use total group size instead. Until the properdata are collected however, use of numerical size of association as anindicator of vigilance is bound to be misleading especially in inter-specificcomparisons (e.g. van Schaik et al., 1983).

lmplications ofthe importance ofneighbors

If neighbor relations are, as suggestcd, more important than overall group-size for shared vigilance, then future rcsearch and experiments cannot treatall group-members equaIly. The focal animal's nearest associates wouldbe the ones most relevant to its behaviour. For example, from a set ofcareful and revealing experiments, Lima (1995a, b) reported that nearbyindividuals were the most likely to respond to the escape or alert behaviorof focal anima!s. The remainder of the flock rarely followed the focalanimal's escapc when it was the only one to respond to a simuiated threat.Further research showed that flight by two birds was significantly more of-ten associated with flock escape than the flight of one bird (Lima, I995b).The distances at which the second bird (and the ensuing flock) foragedfrom the focal bird were not reported however. This information is crit-ical to understanding whether neighbors or far-flung group-members areinfluencing the response to potential lhreats. In the case of the monkeysstudied here, it would be interesting to determine if low visibility restrictsthe utility of far-flung group-members for individual decisions relating to

476 ADRIAN TREVES VIGlLANCE lN ARBOREAL MONKEYS 477

risk avoidance. This may have implications for diIution of risk and otheranti-predator benefits of sociality.

The evidence pre.sented here for the importance of neighbors implies thefollowing:

(1) Group cohesion and neighbor relationships are raore than just char-acteristics of social networks but are also correlates of ecologicalstrategies influencing time budgets and energetics. Selection may thusfavor the maintenance of alliances, associations or kin-based groupsbecause of the time savings provided by shared self-protection.

(2) The dynamics of group formation and dissolution need to be re-examined in the light of the daiIy benefits which an individual canderive from maintaining proximity to others. The motivation fordispersal, subgroup formation and aggressive interaclions may lie inthe lack of benefits gained by certain classes of individual.s.

(3) Observations of variation in spatial cohesion should suggest that out-side threats impinge differentially on study subjects. This can applyto individuaIs (e.g. pregnant fema!es v.v lactating females), clusters(low-ranking matrilines compared to high-ranking matrilines), groups(groups in dangerous areas vs those in safer areas) or even populationswhich vary in lhe level of threat.

References

Alados, C.L. (1985). An analysis of vigilance in !he Spanish lbcx (Capra pyretiaica). —Z. Ticrpsychol. 68, p. 58-64., (1986). SpatiaI s!ructure in groups of Spanish ibcx (Capra pyrenaica). — Biol.Behav. I1, p. 176-185.

Alheris, S,C. {1994). Vigilance in young baboons: EITects of habilat, age, sex and maternalrank on glance ratc. — Anim. Behav. 47, p. 749-755,

Bckoff, M. (1995). Vigilance, flock size, and flock geometry: lnlormation gathermg bywestern evening grnsbeaks (Aves, Fringillidae). — Ethology 99, p. 150-161.

Berger, J, (1978). Group size, foraging, and antipredator ploys: An analysis of bighornshccp decisions. — Behav. Ecol. Sociobiol. 4, p. 91-99.

Bcrtram, B.C.R. (19SO). Vigilance and group size in ostriches. — Anini. Behav. 28, p. 278-286.

Beveridge, F.M, & Deag, J.M. (1986). The effects of sex, !empcraturc and companionson looking-up and leeding in single and mixed species flocks of house .sparrows(Passer domeslicus), chaffinches (Fringilla coelebs) and starlings (Slurnus vulgaris).— Behaviour 100, p. 303-319.

Blumstein, D.T, (I996). How much docs social group sizc influence golden marmot vigi-lance? -Behaviour 133, p. 1I33-1151.

Buiynski, T.M. (1990). Comparative ecology of blue monkeys (Cercopiihecus mitis) inhigh- and low-dcnsity subpopuIations. —Ecol. Monogr. 60, p. l-26.

Caine, N.C. & Marra, S.L. (1988). Vigilancc and social organization in two species ofprimaies. - Anim. Behav. 36, p. 897-904.

Caraco, T. (1979). Time budgeting and group size: A lesL oftheory. -— Ecology 60, p. 618-627.

Cattcrall, C.P., Elgar, M.A. & Kikkawa, J. (1992). Vigilance does not covary with groupsize in an island population of silvcreyes (Zosterops lateralis). — Behav. Ecol. 3,p. 207-2IO.

Chapman, C.A. & Chapman, L.J. (1996). Mixed-specics primalc groups in the KibaIc Forest:Ecological constraints on associalion, — Ini. J. Primato!, 17, p. 31-50.

Cords, M. (1987). Foresl guenons and patus monkeys: Male-male compeiiiion in one-malegroups, — ln: Primate socielies (B. Smuis, D,L, Cheney, R.M. Seyl'arth, R,W. Wrang-ham & T.T. Struhsaker, eds). University of Chicago Press, Chicago., (1988). Mating systems of foresl gucnons: A preliminary review. — ta: A primatcradiation: Evolutionary biology of thc African gucnons (A. Gauticr-Hion, F. Bourliere,J,-R Gaulier & J. Kingdon, eds). Cambridge Umversity Press, Cambridge., (1990). Vigilance and mixed-species association ofsome Easi African foresl mon-keys. - Behav. Ecol. Sociobiol. 26, p. 297-300.

Curio, E. (1976). The ethology of predaiion. — Springer-Verlag, Berlin.DeVore, l. & Hall, K.R.L. {1965). Baboon ecology. — In: Primate behavior: Field siudies

of monkcys and apes (I. Dcvorc, cd.). Holl, Rhinehart and Winston, New York.& Washbum, S.L. (1963). Baboon ecology and human evolulion. — In: African

ecology and human evolulion (F.C. Howell & F. Bourliere, eds.). Aldine, Chicago.Edmunds, M, (1974). Defencein animals; A surveyofanii-predaiordefences. — Longman,

London.Elgar, M.A. (1989). Predator vigilance and group size in maniinaIs and birds: A critica]

review of lhe empirical evidence. — Biol. Rev. 64, p. 13-33.—, Burren, P.J. & Posen, M. (1984). Vigilance and perccption of flock sizc in foraging

housc sparrows (Passer domeslicus). — Behaviour 90, p. 215-223.— -— & Cattcrall, C.P. (1981). Rocking and prcdalor surveillancc in housc sparrows: Test

of an hypoihesis. - Anim. Behav. 29, p. 868-872,, McKay, H. & Woon, P. (1986). Scanning, pecking, and alarm flights in housesparrows. — Anim. Behav. 34, p. 1892-1894.

Fitegibbon, C.D. (1990). Mixed-species grouping in Thomson's gazelles: The antipredatorbcnefits. - Anim. Behav. 39, p. 1116-1126.

Gallon, F. (1883). Inquirics into human facully and ils dcvclopmcnl. — Deni, London.Goss-Custard, J-D., Dunbar, R.I.M. & Aldrich-Blake, FP.G. (1972). Survival, mating and

rearing straIegies in the evolulion of primate social siruciure. - Folia primatol. 17,p. 1-19.

Hamillon, W.D. (1971). Gcomelry for the selfish hcrd. - J. theor. Biol. 31, p. 295-311.Hart, A. & Lendrem, D.W. (1984). Vigilance and scannmg paiierns in birds. — Animal

Behav. 32,p. 1216-1224.

478 ADRlAN TREVES

Holmes, W.G. (1984). Predalion risk and foraging bchaviorofthe hoary marmot in Alaska.- Behav. Ecol, Sodobiol. 15, p. 293-30l.

IsbeIl, L.A. (1983). Daily rangmg bchavior ofred colobus (Colobus badius tephrosceles)in Kibale Foresi, Uganda, — Folia primatol. 4l, p. 34-48.

Jarman, P.J. fl9S7). Group size and activity in eastern grey kangaroos. — Anim. Behav.35, p- 1044-1050.

Jones, W.T. & Bush, B.B. (1988). Movemcnt and reproduciive bchavior of solitary malcredlail guenons (Cercopithecus ascanius). — Amer. J. Primaiol. 14, p. 203-222.

Klump, G.M. & ShaIlcr, M.D. (1984). Acoustic bchavior of birds and mammals in thepredator contexL, — Z. Tierpsychol. 66, p. 189-226.

Knight, S,K. & Knight, R.L. (1986). Vigilance patterns a! bald eaglcs feeding in groups.- Auk 103, p. 263-272.

LaGory, K.E. (1986). Habitat, group siy.e, and the behaviour of whitc-tailcd deer. — Be-haviour 98, p. 168-179.

Landeau, L. & Terborgh, J. (1986). Oddity and thc 'confusion effeci' in predation. — Anim.Behav. 34, p. 1372-1380.

La/.arus, J. (1978). Vigilance, llod; sizc and domain ol' danger size in ihe white-fronledgoose. - Wild Fowl 29, p. 135-145.& lnglis, I.R. (1978). The breeding bchaviour of the pink-footcd goose: Parental care

and vigilani behaviour during the fiedgling pcriod. — Behaviour 65, p. 62-84.Lendrem, D,W, (1983a). Predation and risk in thc blue tit (Parus caemleus). — Bchav.

Ecol. Sociobiol. 14, p. 9-13.— — (1983b). Sleeping and vigilancc in birds. I. FieId obscrvations of the mallard (Anas

platyrhynchos). - Anim. Bchav. 31, p. 532-53S.{1984), Flocking, feeding and predaiion risk: Absolute and instantaneous feeding

rates. - Anim. Bebav. 32, p. 298-299.— -—, Stretch, D., Metcalfc, N. & Jones, P. (I986), Scanning for predators in thc pur-

plc sandpiper; A iime-depcndcnt or time-independcnt process? — Anim. Behav. 34,p. 1577-1578.

Lima, S.L. (1994). On the personal bcnefits of anti-predatory vigJlance. — Anim. Behav.48, p. 734-736.

(1995a). Back to the basics of anli-predatory vigiIance: The group-size cffect. -Anim. Bchav. 49, p. 11-20.

(1995b). Coliective detection of predaiory attack by social foragers: Fraughl withambiguity? - Anim. Bchav. 50, p. 1097-1108.

Lipeu, V.E. & Bekoff, M, (I982). Group size and vigi!ancc in pronghorns. — Z. Tierpsy-chol. 58, p. 203-216.

Maisels, F.G., Gautier, J.-P., Cniickshank. A. & Bosefe, J.-P. fl994), Atlacks by crownedhawk eagles (Stephanoaetus coronatu$] on monkeys in Zaire, — Foiia prirnatol. 61,p. I57-159.

Metcalfe, N.B. (1984a). The eliecL oF habitat on the vigilance of shorebirds; ls visibilityimponant? - Aniffl. Behav. 32, p. 981-985.

(1984b). The effccts of mixed-spedcs fiocking on the vigilance of shorebirds; Whodo they trust7 — Anini. Behav. 32, p. 986-993.

VIGILANCE lN ARBOREAL MONKEYS 479

& Furness, R,W. (1984). Changing priorities: The effcct of pre-migratory faiieningon the trade-olT betwecn foraging and vigilance. — Behav. Ecol. Sociobiol. 15, p. 203-206.

Morgan, M.J. & Godin, J.J. (1985). Antipredator benefits of schooling behaviour m aCyprinodoniid fish, thc bandcd Killiflsh (Fundulus diaplumus). —- Z. Tierpsychol. 70,p. 236-246.

Phillips, K.A. (1995). Resource patch size and ilexible foraging in white-faced capuchins(Cebus capucinus). - Int. J. Primatoi. 16, p. 509-520.

Powell, G.V.N. (1974). Expcrimental anaIysis of thc sotial value of llocking by starlings(Stumus viitgaris) in relation to prcdation and foraging. — Anim. Behav. 22, p. 501-505.

Poys3, H. (1987a), Costs and benefits of group foraging in the tcal (Anas crecca). -Behaviour 103, p. 123-139.

(1987b). Feeding-vigilance lrade-off in the teal (Anas crecca); Effects of feedingmethod and predation risk. — Bchavjour 103, p. 108-121,(1994). Group foraging, distante lo cover and vigilance in the tea), Anas crecca. —

Anim. Behav:48, p. 921-928.PulIiam, H-R. (1973). On the advantages of flocking. — J. thcor. Biol. 38, p. 419-422,

& Caraco, T. (1984). Living in groups: l.s there an optirnal group sizc7 —In:Behavioral ecology, an evo1ulionary approach (J.R. Krebs & N.B. Davies, eds). SinauerAssociates, Sunderland, MA,

Quenette, P.-Y. (1990). Funciions of vigilance behaviour in mammals: A review. — AttaOecol. 11, p. 801-818.

Robcrts, G. (1995). A real-lime response of vigilance behaviour to changes in group si/.e.- Anim. Behav. 50, p. 1371-1374.

Robinson, J.G, (1981). Spatial structurc in loraginggroups oi'wedge-cappcdcapuchinmon-kcys (Cebus nigrivittatus). — Anim. Behav. 29, p, 1036-i056,

Rose, L.M, & Fedigan, L.M. (1995). Vigilance in white-faced capuchins, Cebus capucinus,in Costa Rica. — Anim. Behav, 49, p. 63-70.

Rowell, T.E. (1988). The social system of guenons, compared with baboons, macaques andmangabeys. — In; A primaie radiation: Evolutionary biology of the African penons(A. Gauticr-Hion, F. Bourliere, J.-P. Gauticr & J. Kingdon, cds). Cambridgc UniversityPrcss, Cambrjdge,

dc Ruiter, J.R. (1986). The influence or group size on predator scanning and faragingbehaviour of wedgc cappcd capuchin monkeys (Cebus olivaceus). — Behaviour 98,p. 240-258.

van Schaik, C.P. (1983). Why are diurnal primatcs living in groups? — Behaviour 87,p. 120-144.

& van Hooff, J.A.R.A.M. (1983). On the ultimate causes of primate social sysiems._Behaviour 85, p. 91-I17.& vanNoordwijk,M.A. (19R9). ThespecialrolcofmaleCffcu.tmonkeysinpredation

avoidance and its effccts on group composition. — Behav. Ecol. Sociobiol. 24, p. 265-272,, — -, Warsono, B. & Sulriono, E. (1983), Party size and early detection ofprtxlalorsin Samatran fores! primates. — Primaces 24, p. 211-221.

Sibly, R.M. (1983). OptimaI group size is uusiable. - Anim. Behav. 31, p. 947-94S.

480 Al)RIAN TREVESVIGlLANCE m ARBOREAL MONKEYS 481

Siegfried, W.R. & Underhill, L.G. (1975). FIocking as an anti-prcdak>r strategy in dovcs.- Anim. Benav. 23, p. 504^508.

Skorupa, J.P, <!989). Cr<wned eagles (Stephanoaeins coronatus} in rainforcst: Observationson brceding chronology and diet nt a nesi in Ugan<fa. — lbis 131, p, 294-298.

Struhsaker, T.T. (1975). The red colobus monkey. - University ofChicago Press, Chicago.(1977). Infanticide and social organization in ihe redlail monkey (Cercopithecus

ascanius schinidiii) in the Kibale Forest, Uganda. — Z. Tierpsychol. 45, p. 75-84.(1978). Food habiis of !ive monkey specics in lhe Kibale Forest, Uganda. — ln;

Reccnt advances in primatology (D.J. Chivcrs & J. Hcrbert, eds). Academic Press,Lond<>n.(1980). Comparison of the hehaviour and ccology ol" red colobus and rcdlail monkcys

in the Kibale Forest, Dganda, - A(r. J. Ecol. 18, p. 33-5l.(1981). Polyspecifk- associations among tnjpical rain-forest primates. — Z. Tierpsy-

chol. 57, p. 268-304.(1988), Male tenure, multi-male influxes, and rcproduclive surccss in redtail monkeys

(Cerrapilhecus ascattius). - In: A prirnate radiation: EvoIulionary biology of LheAfriean guenons (A. Gautier-Hion, F. Bourliere, J.-P, Gautier & J. Kingdon, eds),Cambridge Universiiy Press, Cambridge.& Leakcy, M. (1990). Prcy selcclivity by crowncd hawk-eagles on monkeys in the

KibaIe Forest, Ugancia. — Behav. Ecol. Sodobiol. 26, p. 435-443,- — & Leland, L, (1979). Sodoecology of five sympatric monkey species in the Kibale

Foresi, Uganda. - In: Adv. Study Behav. (J.S. Rosenblatt, R.A. Hinde, C. Beer &M.C. Busnel, eds). Academic Pres.s, New York,& (I9S8). Group fission in redtail monkeys (Cercopithecus ascaniux) in lhe

Kibale Forest, Uganda. — In: A primaie radiation: Evolutiona^ bio!ogy of lheAfncan guenons (A. Gauticr-Hion, F. Bourlicre, J.-P. Gautier & J. Kingdon, erfs).Cambridge University Pres.s, Cambridge,& Oatcs, J.F (1975). Comparison of lhe behavior and ecology of red colobus and

black-and-white colobus monkeys in Uganda: a summary. — In: Socioecology andpsychology o!" primates (R.H. Tuttle, ed.). Mouton, The Hague.

- -— & Pope, T.R, {199I). Maiing system and reproductive success: A tomparison oflwoafritan foresi monkeys (Coli>bus badius and Cercopithecus ascanius]. — Behaviour117,p. 182-205.

Terborgh, J. (1983), Five New World primatcs: A study in comparative ecology, — Prince-ton University Press, Princeton, Ncw Jersey.

& Janson, C.H. (19S6). The socioecology of primaie groups, — Ann. Rcv. Ecol.Syst. 17, p. 111-135.

Treves, A. (1997). Sclf-proiection in primates. — Unpubl. PhD thesis, Harvard Universiiy.& Chapman, C.A. (1996), Conspeci(ic threat, predation avoidance and resource tfc-

fense: Implications for grouping in Iangurs. — Behav. Ecol. Sociobiol. 39, p. 43-53.Underwood, R. (1982). Vigilancc behaviour in grazing African anielopes. — Behaviour 79,

p. 81-107.Waser, P.M. (1982), Frimate polyspecific associations: Do they occur by chance? - Anim.

Behav. 30, p. l-8.— - (1984). 'Chanee' and mixed-spedcs associations. — Bchav. Ecol. Sociobiol. 15,

p. 197-202,

(19S7). Interactions among primatc species. - In: Primate SocieIies (B.B. Smuts,D.L. Chcney, R,M. Seyfarih, R.W, Wrangham & TT. Siruhsakcr, eds). University of

Chicago Press, Chicago,Waits, D,P. (1989), Infantieide in mountain gorillas: New cases and a reconsideration of

ihe evidencc. — E:hology 81, p. 1-18.Wrangham, R.W. {1980). An ecological model of femaIe-bonded primate groups. -

haviour 75, p. 262-300., Conklin, N,L., Chapman, C.A. & Hunt, K.D, (1991). The sigmfitance of fibrousfoods for Kibale Foresi chimpanzees. - Phil. Trans. Roy. Soc. London B 334, p. 171-

178.Yaber, M.C. & Herrera, E.A. (1994). Vigilancc, group size, and social status m capybaras.

- Anim. Behav. 48, p. 1301-1307.Young, O.P. (1982). Aggressive mleraclions betwccn howler monkeys and turkey vuliure.s:

The need to lhermoregulaie behaviourally, - Biotropica 14, p. 228-231.