ANIMAL BEHAVIOUR, 1999, 57, 911–921Article No. anbe.1998.1052, available online at http://www.idealibrary.com on

Infant handling in wild Cebus capucinus: testing bonds betweenfemales?

JOSEPH H. MANSON

Department of Anthropology, University of California, Los Angeles

(Received 31 August 1998; initial acceptance 7 September 1998;final acceptance 29 October 1998; MS. number: A8149R)

The evolved functions, if any, of infant handling (IH) by female primates remain unclear for manyspecies. I tested a new hypothesis, that IH tests social bonds between adult females, using data on a groupof wild white-faced capuchins. I also tested the nonadaptive, learning-to-mother, reciprocity, harassmentand alliance-formation hypotheses. Focal subjects were the mothers of 10 infants that ranged in age from0 to 90 days. The behaviours comprising IH (inspecting, nuzzling, sniffing, and a distinctive vocalization)differed from typical maternal behaviour. Nulliparous females engaged in no more IH than did parousfemales. The median frequency of rough handling was zero, kidnapping was not observed, and maternalrestraint of infants was extremely rare. Infant handling rate was unrelated to the relative dominanceranks of the mother and the handler. Females tended to handle the infants of females with which theygroomed and formed aggressive coalitions more frequently. Dyadic IH rates were not correlated with ratesof allonursing when the infants were older than 90 days of age. There was no evidence for reciprocalexchanges of IH between females. Mother-terminated IH bouts were longer when the mother hadgroomed the handler more frequently before the infant’s birth. Based on these findings, the learning-to-mother, reciprocity and harassment hypotheses are unsupported for this species. Some support existsfor the nonadaptive hypothesis. Contrasting predictions of the bond-testing and alliance-formationhypothesis are presented for future testing.

1999 The Association for the Study of Animal Behaviour

Correspondence: J. H. Manson, Department of Anthropology,UCLA, Los Angeles, CA 90095-1553, U.S.A. (email:jmanson@anthro.ucla.edu).

Care of young by individuals other than their geneticparents occurs in many species (Emlen 1991). Typically,recipients of this care or their parents experienceenhanced survival prospects, and caregivers benefit viaincreased production of nondescendant kin, tolerancefrom the breeders on their territory, or various othermeans (Emlen et al. 1991). In most well-studied group-living primates, females other than the mother appearhighly motivated to touch and carry infants, but theevolved function(s), if any, of this motivation remainsunclear after more than two decades of study and debate(Hrdy 1976; Nicolson 1987; Maestripieri 1994a). Recentauthors, beginning with Wasser & Barash (1981), haveused the neutral term ‘infant handling’ (IH) to denoteinteractions towards infants initiated by individuals otherthan their mothers. Several proposed explanations for IHgenerate testable predictions, although the links betweensome of these hypotheses and their associated predictionsare debatable (see below). The nonadaptive hypothesis(Quiatt 1979; Scollay & DeBold 1980) holds that IH is aby-product of selection on females to bond with and care

0003–3472/99/040911+11 $30.00/0 911

for their own infants. Two predictions generated fromthis hypothesis are that (1) IH should strongly resemblematernal care (see Emlen et al. 1991), and (2) femalescurrently caring for their own small infants shouldengage in less IH than other females (Paul & Kuester1996). An alternative nonadaptive hypothesis, that new-borns attract attention merely because they are novelobjects, predicts that IH rates will peak during the firstfew days of life and decline thereafter. If IH providespractice for motherhood (Lancaster 1971), then nulli-parous females should do a disproportionate amount ofinfant handling, and this experience should increase thesurvival prospects of their first infants. If IH is a form ofkin-selected altruism (Riedman 1982), then it should bedirected preferentially towards the infants of relatedfemales, and should enhance infant survival and/or themother’s future reproductive output. Even unrelatedfemales could benefit from the exchange of infant careservices (Hrdy 1976; Riedman 1982; Stanford 1992) if theconditions for the evolution of reciprocity (Trivers 1971)exist. If IH represents harassment rather than care of theinfant (Silk 1980; Wasser & Barash 1981), handlers andinfants should not be closely related, mothers should belower ranking than the females that handle their infants,

1999 The Association for the Study of Animal Behaviour

912 ANIMAL BEHAVIOUR, 57, 4

and IH should sometimes have harmful consequences forinfants. Finally, if IH functions to establish valuablealliances between the handler and the infant’s mother(Hrdy 1976; Kurland 1977; Maestripieri 1994b), thenmothers should be higher ranking than the females thathandle their infants, and handlers should in fact improvetheir relationships with infants’ mothers as a conse-quence of IH. Maestripieri (1994a) emphasized that IH isa heterogeneous phenomenon, ranging from clearlybenign (e.g. allonursing) to clearly abusive (e.g. engagingin a tug-of-war over an infant). He hypothesized thatinterspecific differences among cercopithecoid monkeysin the relative frequency of benign versus abusive IH stemfrom differences in the overall structure of female–femalerelationships, which in turn are a product of socio-ecological differences as described by van Schaik (1989).Maestripieri (1994a) also proposed that the relative fre-quency of benign versus abusive IH affects species-typicalmaternal style, and argued that considerable insight intothe function of IH can be gained by carefully distinguish-ing benign from abusive IH in data collection protocols.

I have two goals in this paper. First, I propose a newhypothesis for IH: that it functions to test social bonds(Zahavi 1977), as distinct from forging or strengtheningbonds, between adult females. Second, I test this newhypothesis as well as the by-product, learning-to-mother,reciprocity, harassment and alliance-formation hypoth-eses using data collected on wild white-faced capuchins.Relatively little information is available on IH in wildplatyrrhine primates. In callitrichines, it is clearly a formof infant care and part of a system of cooperative breeding(Goldizen & Terborgh 1989; Tardif et al. 1992). SouthAmerican squirrel monkeys, Saimiri sciureus, spend a largefraction of their first several months riding dorsal on andnursing from females other than their mother (Baldwin1969; Williams et al. 1994). In contrast to callitrichines,S. sciureus and Cebus groups contain multiple adults ofboth sexes, including multiple breeding females, andfemales are typically philopatric and form linear domi-nance hierarchies (Robinson & Janson 1987; Mitchellet al. 1991; Fedigan 1993; Perry 1996). These social paral-lels with many cercopithecoid monkeys invite tests ofhypotheses about primate social behaviour which areclaimed to apply throughout the order but which havebeen tested almost exclusively on narrower taxonomicgroups (Di Fiore & Rendall 1994; Strier 1994). Among wildC. olivaceus, females preferentially investigate, associatewith and carry infants whose mothers are kin or lowerranking than themselves, and old juvenile and youngadult females interact more with others’ infants thanexpected by chance (O’Brien & Robinson 1991). Allo-nursing of independently locomoting C. olivaceus andC. capucinus infants and juveniles has also been docu-mented (O’Brien 1988; O’Brien & Robinson 1991; Perry1996), although its adaptive significance remains unclear.However, the frequency and quality of other forms of IHin wild C. capucinus have not even been described indetail, much less explained. Nor is it known whetherrelationships between handlers and particular infants per-sist as infants spend less time with their mothers andsolicit allonursing with increasing frequency.

Zahavi (1977; Zahavi & Zahavi 1997) has argued thatsignals exchanged between ‘bond partners’ (e.g. mates;parents and offspring) often contain stress-producingor aggressive elements because the recipient’s positiveemotional response to these elements provides valuable,reliable evidence of its current willingness to continue toinvest in the social bond. In this view, many affectionatebehaviours, such as children jumping on their parents’laps, derive their communicative value (i.e. their‘honesty’) from being inherently stress producing. (Con-sider an adult’s likely reaction to being jumped on bysomeone else’s child.) The recipient’s willingness toengage in, and prolong, the interaction should be pro-portional to the current strength of its bond with theinitiator. As elaborated recently by Zahavi & Zahavi(1997), bond testing could provide valuable informationeven in well-established social relationships, because eachpartner’s immediate willingness to invest in the otheris likely to vary on short time scales as a function ofeach animal’s health, its relationships with other groupmembers, and other factors.

A problem with this idea is the difficulty of determin-ing, a priori, whether a putative signal of bond strengthactually imposes any stress on the recipient. However,theoretical considerations suggest some behaviours aslikely candidates for stress-inducing tests of the strengthof a bond. One of these, in primates, is the act of handlinga female’s infant. Consider a species in which IH neitherbenefits infants (as required by the kin selection andreciprocity hypotheses) nor harms them sufficiently toincrease the handlers’ relative reproductive success (asrequired by the harassment hypothesis). Infant handlingcould still be mildly stressful to the mother because of thepossibilities of accidental injury to the infant and diseasetransmission. Thus, a mother’s willingness to tolerate andprolong IH interactions could signal her willingness tocontinue investing in her relationship with the handler.This hypothesis generates three predictions. First, IH willbe most common when the mother and handler have anestablished pattern of engaging in relatively high rates ofcooperative behaviour (especially costly cooperativebehaviour such as coalition formation against third par-ties). This prediction differs from the prediction of thealliance-formation hypothesis that IH will increase thefrequency of future cooperative behaviour betweenmother and handler. Second, IH bouts that are termi-nated by the mother will be longer when other measuresof bonding (e.g. grooming) indicate that the mothervalues her bond with the handler. Third, the relativeoccurrence of more stress-inducing forms of IH (e.g.touching an infant rather than merely inspecting it atclose range) will be higher when mother and handlerhave a closer established social bond.

METHODS

Study Site and Subjects

I collected data on members of a single social groupof white-faced capuchins in and just north of LomasBarbudal Biological Reserve, Guanacaste, Costa Rica

913MANSON: CAPUCHIN INFANT HANDLING

(10)30*N, 85)22*W). The area consists largely of highlyseasonal tropical dry deciduous forest (Frankie et al.1988). Observers followed the social group from dawn todusk every day for up to 25 sequential days per month.We easily identified adults by their scars, hair colourpatterns and facial contours. Infant subjects (i.e. less than90 days old) were recognized by association with theirmothers.

I collected data during two distinct periods, in whichslightly different protocols were used, and analysed thetwo data sets separately when testing predictions. I col-lected data on six infants aged 0–90 days between May1991 and July 1992 (study 1), and on four infants (includ-ing one set of twins) between 13 March and 19 June 1997(study 2). During study 1, independently locomotingjuveniles were not recognized until March 1992 when Imarked them by squirting Clairol ‘Born Blonde UltimateLightener’ on their backs at a range of 3–10 m via syringe.During study 2, I successfully marked only two smallindependently locomoting juveniles, whereas the otherscould not be recognized consistently. Over the course ofthe two studies, the group contained five to eight adultfemales, three to four adult males and 11–15 immatures.Maternal kinship is known only for individuals born in1990 or later. After study 1, data collection on female–female relationships in the group continued until May1993 (Perry 1996). Before study 2, I collected focal data onadult females beginning in December 1996. I collected nodata before the beginning of study 1 or after the end ofstudy 2.

Data Collection

In both studies, I collected data in the form of focalindividual follows (Altmann 1974) during which Inarrated all social behaviours and the identities of theinteractants onto microcassettes. Focal subjects weremothers of infants that were 0–90 days old. In study 1, allfollows were of 10-min duration. In study 2, in additionto 10-min follows, I conducted 20 40-min follows and 2060-min follows of the mother of the twins as part of an adhoc study of the first observed case of twins in wild Cebus(J. H. Manson, K. Atkins & S. Perry, unpublished data). Toensure comparability of data among females, I extractedthe first and last 10-min periods of each of these longerfollows and treated them as separate 10-min follows. I didnot analyse data from other portions of the longer followsfor this paper.

In study 1, I recorded the times of behavioural events towithin 1 min, except for onsets and ends of groomingbouts after 30 December 1991 (Perry 1996). I recordedfemales’ activities and neighbours at instantaneoussamples every 2.5 min. In study 2, I recorded the times ofall behavioural events to the nearest second. I recordedonly one instantaneous sample per follow at the end ofeach 10 min. Interobserver reliability was established asdescribed by Perry (1996).

Table 1 lists each mother–infant dyad that contributeddata to the analyses presented in this paper. I collected114 h of focal follows on mothers that had infants 0–90days old (55 h in study 1, 59 h in study 2). In calculating

rates of interactions between females and infants otherthan their own, I also used 115.5 data h on females thatdid not currently have infants 0–90 days old (84 h instudy 1, 31.5 h in study 2). Finally, to measure aspects offemale–female relationships when neither female had aninfant that was 0–90 days old, I used 60 h of focal femaledata collected in December 1996–April 1997 before study2, and results from Perry’s (1996) study of female–femalerelationships carried out during and after study 1. Duringstudy 1, adult females could be ranked in a linear domi-nance hierarchy based on avoidance and cowering indyadic interactions (Perry 1996). During study 2, the first-through the fourth-ranked females could be ordered in alinear hierarchy (unpublished data). Two females rankedbelow these four females but their dyadic dominancerelationship was unclear, so I assigned them a tied rank of5.5. Similarly, the two lowest-ranking females’ domi-nance relationship could not be resolved so they wereassigned a tied rank of 7.5.

Table 1. Mother–infant dyads

StudyMother’s

IDInfant’s

IDInfant

birthdateFocal h on

mother

1 TT FR ∼7 May 1991 8.171 SS BR 18 May 1991 10.171 NN LL 8 July 1991 9.171 AA RI 13 July 1991 7.831 WW YO 16 July 1991 7.671 TT EE 8 April 1992 11.672 AA KJ 11 March 1997 25.672 AA AT 21 March 1997 *2 TT FK 5 April 1997 18.332 NN BU 19 April 1997 15.00

*Second twin.

Behavioural Definitions

Except for behaviours specific to interactions involvinginfants, behavioural definitions used in this paper areidentical to those used by Perry (1995, 1996, 1997) inearlier work based on data collected on this group. I codedproximity using adult male monkey body lengths (minusthe tail) as a metric (approximately 40 cm; K. Glander,personal communication). In this paper, ‘approaches’,‘leaves’ and ‘proximity’ refer, respectively, to approachesto within one body length, leaves to outside one bodylength, and proximity within one body length. Groom-ing consisted of a monkey running its fingers throughanother monkey’s hair. Aggression included facial andvocal threats, lunging, hitting and biting. Coalitions weresignalled by joint facial or vocal threats towards acommon opponent, and often by specific forms of bodycontact (cheek to cheek, overlord; Oppenheimer &Oppenheimer 1973) between the coalition partners.

I distinguished among several varieties of interactionsbetween females and infants. An ‘inspect’ was scoredwhen a monkey brought its face within 15 cm of aninfant and peered at it. A ‘nuzzle’ consisted of a monkeyplacing its nose in contact with an infant. Grooming was

914 ANIMAL BEHAVIOUR, 57, 4

scored as described above. A ‘touch’ was scored when amonkey placed its hand(s) on the infant other than togroom it. Pulling on any part of an infant’s body wasscored as a distinct type of interaction. A ‘restraint’was scored when a monkey held on to an infant thatwas trying to leave. Allonursing consisted of an infantestablishing nipple contact with a female other than itsmother. A ‘carry away’ was scored when a female, withanother’s infant dorsal or ventral, left outside one bodylength of the infant’s mother. A ‘ride invite’ consistedof a monkey crouching within one body length of aninfant. In some analyses, when they were performed bya female towards another female’s infant, the abovebehaviours were lumped as ‘infant handling’. ’Machinegun vocalizations’ consisted of a rapid sequence of soft,low-pitched staccato noises. I did not classify them as aform of infant handling because when perfomed frombeyond 15 cm, they may have had little effect on infantsor mothers, and when performed from within 15 cm,they (by definition) accompanied some other form ofinfant handling. I did not regard a ‘climb on’ (an in-fant climbs on to a monkey other than its mother) asinfant handling, because the action was taken by theinfant itself. A ‘turn away’ consisted of a mother, with herinfant dorsal or ventral, moving and thereby movingher infant further away from a monkey that was withinone body length of her. I distinguished touch, inspectand nuzzle from each other consistently only in study 2,I coded machine gun vocalizations consistently only instudy 2, and I recorded turn away only during study 2.

Data Analysis

Except where explicitly indicated otherwise, I carriedout all analyses presented here using, as data points,dyads consisting of an adult female and another female’sinfant. However, most of the variables of interest arecharacteristics of the relationship between the infant’smother and the other female (henceforth, the handler).This approach reflects the questions raised by several ofthe hypotheses proposed to account for IH (see Introduc-tion). The twins born to female AA in March 1997 aretreated as one infant in these analyses. Study 1 included26 handler–infant dyads, and study 2 included 18handler–infant dyads, although not every dyad contri-buted to every data analysis (e.g. I excluded dyads with IHrates of zero from an analysis in which the dependantvariable was the proportion of IH acts comprised bytouching and rough handling). Using female–infantdyads, rather than individual females, as data pointscorrects for differential availability to would-be handlersof infants of varying characteristics (e.g. mother’sdominance relationship to the handler).

For the study 1 data, I estimated from instantaneoussamples the proportion of time within proximity (forfemale–female dyads) and the proportion of time dorsaland in nipple contact (for mother–infant dyads). For thestudy 2 data, I calculated these rates directly from actualdurations of these states. Dyadic IH rates from study 1 arepresented as event onsets per hour of focal observation ofthe two females while the infant in question was 0–90

days old. I calculated dyadic IH rates from study 2 using adifferent method, which takes advantage of the avail-ability of event onset times recorded to the exact second.I calculated the proportion of focal minutes, while theinfant in question was 0–90 days old, during which atleast one IH interaction of each type occurred for eachdyad. Denominators used in the calculations of theserates and proportions consisted of all the focal obser-vation time (of mother plus handler) during the infant’sfirst 90 days.

I used two techniques to measure associations betweenmaternal grooming of female potential handlers andlatency to maternal termination of IH episodes. For bothtechniques, I coded an IH episode as beginning when thehandler first inspected, nuzzled, touched or groomed aninfant while the infant was in contact with its mother,and ending when the mother left or turned away fromthe other female or when the handler left or the focalfollow ended. As the first technique, I calculated themedian duration of mother-terminated IH episodes foreach handler–infant dyad, and the correlation betweenthese medians and maternal grooming of the handler. Iexcluded from this analysis infant-handling episodes thatwere ended by the handler, or that were still in progresswhen a focal follow ended. This procedure’s strength isthat the data points (handler–infant dyads) are indepen-dent, while its weakness is that the exclusion ofnonmother-terminated IH episodes could introduce bias.Therefore, I also analysed the same data using a form ofsurvival analysis (Allison 1984), a set of techniques thatincorporate data from ‘censored’ intervals. For example,in clinical studies, survival analysis incorporates thesurvival times of patients still alive at the end of a study,which is a censoring event with respect to the event ofinterest (death). Because I was interested in maternaltermination of IH episodes, I regarded the termination ofIH episodes by the handler, and by the ends of focalfollows, as censoring events. I regarded only mother-terminated IH episodes as uncensored, and consideredthem to have ‘survived’ until terminated by the mother. Iused Peto & Peto’s (1972) generalized Wilcoxon test, aform of survival analysis for comparing two groups, todetermine whether prior maternal grooming of thehandler was associated with latency to maternaltermination of IH episodes. These episodes were notindependent data points, because each handler–infantdyad contributed multiple episodes to the analysis.

RESULTS

Relative Frequencies of Infant HandlingBehaviours

The four most common IH behaviours were inspect,nuzzle, groom and touch. Using the mean values andranges among the three mothers in study 2, thesebehaviours occurred at least once during 43% (30–56%),37% (31–44%), 36% (19–48%) and 35% (24–41%),respectively, of all minutes during which at least oneIH behaviour occurred. Rough handling was rare (seebelow).

915MANSON: CAPUCHIN INFANT HANDLING

Infant Handling and Infant Age

To ascertain whether IH was disproportionatelydirected towards newborns, I examined dyadic IH rates asa function of infant age using the study 2 data (study 1produced insufficient data per dyad per time period topermit this analysis). There was no tendency for IH topeak in the first 10 days of infant life (Fig. 1). Indeed, forthose handler–infant dyads that showed a change in IHrate between consecutive 10-day periods, seven of 11(63.6%) showed an increase between days 1–10 and11–20, and seven of nine (77.7%) showed an increasebetween days 11–20 and 21–30. However, the meanapproach rate to new mothers by other adult females,which may be a more sensitive measure of interest ininfants, did peak during days 1–10 for two of the threenew mothers (Table 2).

Abusive IH, Maternal Aggression andRestrictiveness

I calculated, for each dyad consisting of an adult orlarge juvenile female and another female’s infant, theproportion of all IH interactions that consisted of ‘roughhandling,’ which included pulling and all overt andsuspected aggressive acts by the handler towards theinfant (hitting, nipping and any contact that caused the

infant to yelp or jerk). In study 1, this proportion was zerofor 16 of 25 (64%) handler–infant dyads, the maximumproportion was 0.33, and the maximum hourly rate ofrough handling was 0.27 interactions per hour. In study2, this proportion was zero for 13 of 15 (86.7%) handler–infant dyads, the maximum proportion was 0.04, and themaximum proportion of observation minutes duringwhich rough handling was observed was 0.0007. Iobserved only one case of an infant pulled off its mother’sback by a female, and the infant returned to its motherwithin 1 min.

Mothers tolerated their infants’ attempts to contactother females, rarely behaved aggressively towardshandlers or would-be handlers, and never had difficultyretrieving their own infants. On 47 occasions duringstudy 1, and on 48 occasions during study 2, an infantassumed a dorsal or ventral position on a female otherthan its mother (sometimes with one or more limbs stillon its mother) by actively climbing on to the otherfemale. On 16 occasions during study 1 and 18 occasionsduring study 2, an infant left (i.e. moved beyond onebody length of) its mother. On only one occasion did amother restrain her infant from leaving or from climbingon to another female that was within one body length. Iobserved a total of five incidents (two during study 1,three during study 2) of mothers directing aggressiontowards females that were attempting to handle, or hadhandled within the past minute, their infants. Followingcarry aways (N=8), mothers were always able to retrievetheir infants by approaching the handler and directing aride invite towards the infant.

0.08

0

Infant age (days)

Infa

nt

han

dli

ng

rate

(p

rop

orti

on o

f m

in)

21–30

0.03

0.07

0.06

0.05

0.04

0.02

0.01

1–10 11–20

Figure 1. Infant handling rate as a function of infant age duringstudy 2. Each line represents one infant–handler dyad. -: dyadsincluding female AA’s twins. /: dyads including infant BU. 0: dyadsincluding infant FK.

Table 2. Female (N=7) approach rates (X±SE approaches/h) tomothers during days 1–30 postpartum

Mother ID

Days postpartum

1–10 11–20 21–30

AA 0.27±0.14 0.76±0.39 1.11±0.48NN 0.69±0.44 0.18±0.09 0.32±0.24TT 0.28±0.18 0.13±0.08 0.05±0.05

Correlates of Rate and Forms of IH

During study 1, five infants were born within 70 daysin May–July 1991, and a sixth infant (EE) was born tofemale TT in April 1992 (EE’s older sibling FR had dis-appeared in January 1992; Table 1). This schedule ofbirths permits a comparison of IH rates as a function ofthe availability of infants in the group that were 0–90days old. Specifically, I examined whether EE, who wasborn while the group contained no other small infants,received more IH, as would be expected if females cur-rently caring for their own small infants engage in less IHthan other females. Adult females’ handling rates of otherfemales’ infants (0–90 days old) were higher in April–July1992, when one infant was present, than in May–September 1991, when up to five infants were present(Mann–Whitney U test: U=10, N1=20, N2=4, P=0.02). Icompared the rates at which each of the four adultfemales, other than TT, handled TT’s 1991 and 1992infants (FR and EE, respectively). I also compared the ratesat which large juvenile females DD and VV, which werefirst recognized in March 1992, handled TT’s infant EE.These nulliparous females did not engage in more IHthan did parous females (Fig. 2). Because EE was thegroup’s only infant, which would strongly affect the rateat which he was handled, I excluded data on EE fromsubsequent analyses.

Because I did not individually identify nulliparousfemales, or even distinguish them from immature males,

916 ANIMAL BEHAVIOUR, 57, 4

1.8

Number of infants (0–90 days old) in group

Han

dle

s/h

of

fem

ale

TT

's i

nfa

nts

1

0.8

0.6

0.2

012–5

1.2

1.4

1.6

0.4

AA

NN

WW

SS

AA

NN

WW

SS

DD

VV

Figure 2. Handling rates of TT’s two infants born during study 1, asa function of the number of infants in the study group that were0–90 days old. Letters denote identity of handler.

during the first 90 days of any infant’s life other than EE,nulliparous females’ frequencies of handling otherinfants, relative to those of adult females, can onlybe inferred indirectly. The mean (&SE) adult female IHrate towards the five infants born in 1991 (averagedamong adult female potential handlers) was 0.37&0.06/h. Dividing the hourly IH rate of all juveniles collec-tively by the number of juveniles in the group yields amean (&SE) juvenile IH rate of 0.22&0.04/h towardsthese five infants. The difference between these ratesapproached statistical significance (Wilcoxon test: T=1.0,N=5, P=0.08). The mean (&SE) adult male IH ratetowards these five infants (averaged among adult malepotential handlers) was 0.15&0.09/h. Thus, infanthandling was predominantly an activity of adult females.

I examined whether IH interactions varied as a functionof the relative dominance ranks of the mother and thehandler. I categorized each handler–infant dyad bywhether the handler was lower ranking or higher rankingthan the infant’s mother. There was no difference inhandling rates between these two categories of dyadsin either study (Mann–Whitney U test: study 1, U=31.5,N1=N2=10, P=0.162; study 2, U=40.5, N1=12, N2=8,P=0.558).

Allonursing of infants 0–90 days old was rare. Of the136 IH interactions observed during study 1, two (1.5%)consisted of allonursing. In contrast, older infants andsmall juveniles engaged in more frequent allonursingduring the 1991–1993 observation period (Perry 1996).Combining data from study 1 with Perry’s (1996) findingsreveals that older infants were not allonursed by the samefemales that had handled them when they were younger:females in female–infant dyads that engaged in more IHduring the infant’s first 90 days showed no tendency toengage in more allonursing when the infant was older(Kendall’s ô= "0.08, N=16, P=0.68).

I examined whether parous females preferentiallyhandled the infants of females that handled their (thehandlers’) infants. Data from study 1, but not study 2,permitted a test of this prediction from the reciprocity

hypothesis. For all female–female dyads, the rate at whichone female (A) handled the other’s (B’s) infant wasplotted against the rate at which B handled A’s infant.These variables were not correlated (ô= "0.22, N=10,P=0.37).

Several measures of the mother–handler social relation-ship were associated with IH rate. In both studies, IH ratewas negatively correlated with absolute dominance rankdistance between the handler and the mother; femalestended to handle infants whose mothers were closer tothem in rank (study 1, ô= "0.35, N=20, P=0.03; study 2,ô= "0.33, N=21, P=0.03). Females also tended to handleinfants of females with which they groomed at higherrates. In study 1, IH rate was positively correlated withmother–handler grooming rate (sum of rates in bothdirections) between 30 December 1991 (when observersbegan to record grooming bout onsets and ends to thenearest second; see Perry 1996) and the end of obser-vations in May 1993 (ô=0.34, N=20, P=0.04). This corre-lation was stronger in dyads in which the infant’s motheroutranked the handler (ô=0.67, N=10, P=0.007) than indyads in which the handler ouranked the infant’s mother(ô=0.07, N=10, P=0.79). In study 2, mother–handlergrooming rate prior to the infant’s birth was positivelycorrelated with IH rate (ô=0.43, N=21, P=0.006). In con-trast to the results of study 1, this correlation was strongerin dyads in which the handler outranked the infant’smother (ô=0.61, N=8, P=0.03) than in dyads in which theinfant’s mother outranked the handler (ô=0.34, N=12,P=0.12). The correlation between grooming and IHrate in study 1 may be attributable in part to the effect offemale–female proximity, because mother–handlerproximity (proportion of time within one body length)between May 1991 and May 1993 was even more stronglycorrelated with IH rate (ô=0.44, N=20, P=0.0066) thanwas mother–handler grooming. However, in study 2,mother–handler proximity before the infant’s birth wasnot significantly correlated with IH rate (ô=0.30, N=21,P=0.06). I obtained rates at which mothers and handlersformed coalitions against other females throughoutthe 1991–1993 observations from Perry’s (1996) data set.Infant handling rate during study 1 was marginallypositively correlated with mother–handler coalitionformation rate (ô=0.31, N=20, P=0.05). Female–femalecoalition formation rate during 1991–1993 was positivelycorrelated with IH rate during study 2, which was carriedout 4 years later (ô=0.44, N=18, P=0.01; Fig. 3). Duringdata collection in 1997 before the births of infants, Iobserved only two female–female coalitions againstfemales and was therefore unable to calculate meaningfulcoalition formation rates. The rate at which the handlerformed coalitions with other females against the motherin 1991–1993 was not correlated with IH rate in eitherstudy 1 (ô=0.15, N=20, P=0.343) or study 2 (ô=0.02,N=18, P=0.92).

Using data from study 2, I calculated, for each female–infant dyad in which IH occurred, the proportion ofIH interactions that consisted of touching plus roughhandling. Because touching was the only apparentlybenign form of IH that sometimes (although rarely)elicited signs of infant distress such as jerking and

917MANSON: CAPUCHIN INFANT HANDLING

yelping, I reasoned that it was the most stressful form ofnonrough handling. The proportion of IH that comprisedtouching and rough handling was not correlated withmother–handler grooming rate prior to the infant’s birth(ô=0.35, N=15, P=0.07) or with mother–handler coalitionrate against females during 1991–1993 (ô=0.36, N=14,P=0.08).

Across handler–infant dyads in study 2, handlersdirected machine gun vocalizations to infants during amedian of 16% (range 0–50%) of observation minutesduring which they handled infants. Mothers were neverobserved to direct machine gun vocalizations towardstheir own infants.

Correlates of Mother-controlled IH Duration

I calculated durations of the 167 infant handlingepisodes observed during the first 45 days of infants’ livesduring study 2. I restricted analysis to this age rangebecause older infants began to leave their mothers(unpublished data), and I wished to focus on situations inwhich mothers had complete control of the durations ofIH episodes (i.e. a mother could terminate any IH episodeby leaving or turning away from the handler). Of theseepisodes, 87 were uncensored (i.e. they ended when themother left or turned away from the handler) and 80 werecensored (i.e. they ended when a focal follow ended orthe handler left the mother). Considering only theuncensored episodes, the rate at which the mothergroomed the handler before the infant’s birth was notcorrelated with median IH episode duration across dyads(ô=0.27, N=12, P=0.22). For the survival analysis, usingeach episode as a data point, I coded each episodedichotomously depending on the rate at which themother had groomed the handler before the infant’sbirth. I designated ‘high grooming’ episodes as episodes

involving mother–handler dyads in which, before thebirth of the infant, the mother had groomed the handlerabove the median rate of all mother–handler dyads. Idesignated all other intervals as ‘low grooming’ episodes.Peto & Peto’s (1972) generalized Wilcoxon test revealedthat high grooming episodes were significantly longerthan low grooming episodes (high grooming median=193 s, N=36; low grooming median=87 s; N=131; teststatistic= "2.17, P=0.015; Fig. 4). This difference was notattributable to a generalized tendency for females toterminate quickly proximity bouts with females that theyseldom groomed, irrespective of the presence of aninfant. For the same 12 female–female dyads that contrib-uted data to the analysis of IH interval durations, Icalculated durations of proximity bouts that occurredbefore either female had given birth during the 1997observations. I coded these bouts as uncensoredif they ended when the future mother left the futurehandler (N=33), and as censored if they ended whenthe future handler left the future mother or the focalfollow ended (N=38). An analogous Peto & Peto general-ized Wilcoxon test revealed no difference between thedurations of proximity bouts as a function of the rate atwhich the future mother groomed the future handler(high grooming median=133 s, N=37; low groomingmedian=128 s, N=34; test statistic= "0.56, P=0.288).

0.04

Mother–handler coalition rate, 1991–1993

Infa

nt

han

dli

ng

rate

(p

rop

orti

on o

f m

in),

199

7

0.01

0.02

0.015

0.01

0.005

0 0.0060.0040.002 0.008

0.025

0.03

0.035

Figure 3. Infant handling rate during study 2 (1997) as a function ofthe rate (proportion of 10-min focal follows) at which the handlerand the infant’s mother formed coalitions against other femalesduring 1991–1993 observations. Each data point is one infant–handler dyad.

1.0

Latency to maternal termination of IH episode (s)

Prop

orti

on o

f IH

ep

isod

es 's

urv

ivin

g'

700

0.8

0.6

0.4

0.2

0 500400300200100 600

Figure 4. Survivorship function of infant handling episodes based onthe Kaplan–Meier product-limit method (Kaplan & Meier 1958).Episodes involving mother–handler dyads in which the mother hadgroomed the handler above (—,—) and below (—.—) themedian rate before the infant’s birth.

DISCUSSION

Because infant handling is a heterogeneous phenomenon(Hrdy 1976; McKenna 1979; Maestripieri 1994a), datafrom one species cannot be used to evaluate, for primatesgenerally, any of the hypotheses advanced to explain it.However, for my C. capucinus group, results presented inthis paper fail to support several widely discussed hypoth-eses. Parous females performed a large proportion of allinfant handling, and when juvenile females were indi-vidually recognizable (N=1 infant, EE), parous femalesengaged in more IH than did nulliparous females, sug-gesting that practice for motherhood is not the sole

918 ANIMAL BEHAVIOUR, 57, 4

function of IH. The harassment hypothesis was not sup-ported, because (1) a small proportion (zero, for mostdyads) of IH interactions were classifiable as roughhandling, even using a very broad definition of this term,(2) IH was not performed disproportionately often bydominant females towards the infants of subordinates, (3)maternal restraint of infants and maternal aggressiontowards handlers were both very rare, indicating thatmothers did not perceive IH as a serious threat to theirinfants (see Maestripieri 1994a), and (4) handlers tendedto join coalitions with, but not against, the mother.Almost all transfers of infants between females wereinitiated by infants, and mothers never had difficultyretrieving their infants. However, although infants didnot appear to suffer costs as a result of their attractivenessto adult females, these data provide no support forthe assumption (required by both the kin selection andreciprocity hypotheses) that young white-faced capuchininfants benefit from IH. Allonursing of semi-independentinfants and juveniles in Cebus (O’Brien 1988; O’Brien &Robinson 1991; Perry 1996) probably represents altruisticcare of others’ offspring. However, because I found nocorrelation, among female–infant dyads, between IH rateduring the infant’s first 90 days and subsequent allo-nursing rate, it is unlikely that early IH functions toestablish handler–infant relationships that later includeallonursing. Handling of young infants appears to be adistinct phenomenon from that of allonursing of olderinfants. Similarly, O’Brien & Robinson (1991) report thatamong C. olivaceus female–infant dyads, allonursing fre-quency was not affected by the sociodemographic vari-ables (kinship, relative rank and handler’s age) thataffected the frequencies of other IH behaviours. It is notclear how infants’ welfare could be enhanced by theinspects and nuzzles that constituted most of thehandling directed, in my study, towards young infantsthat spent over 90% of their time (unpublished data)on their mothers. However, two cases of temporaryadoption, both of which probably saved the life of a 2- to3-month-old nursing infant, have been observed orinferred at Lomas Barbudal. One is reported by Perry(1996). The second incident occurred in June 1997.Female AA and one of her twins (AT) were observed apartfrom the group for at least 2 days. When the group waslocated again, both twins were alive and appearedhealthy. Presumably, females NN and/or TT nursedtwin KJ while AA was separated from the group. Theseincidents are consistent with the view that IH estab-lishes relationships that could provide an infant withalternative sources of care following the temporary orpermanent disappearance of its mother (Jay 1963).

Female white-faced capuchins tended to handle infantsof females with which they groomed and formed aggres-sive coalitions more frequently. These results are probablynot attributable solely to differences among female–female dyads in spatial proximity, because whereasfemale–female grooming rate was correlated with IH ratein both study periods, female–female proximity was cor-related with IH rate only in study 1. The association ofgrooming and coalition formation with IH can be inter-preted as support for the alliance-formation hypothesis

for IH. Some authors (e.g. Paul & Kuester 1996) derivefrom this hypothesis the predictions that females willdirect IH disproportionately up the dominance hierarchyand that high IH rates towards an infant will precedeimprovements in the handler’s relationship with themother. Neither prediction was supported by theseresults. Indeed, dominance and grooming relationshipsin adult female–female dyads were highly stable between1992 and 1997 (Perry 1996; unpublished data; but seebelow). However, a broader interpretation of the alliance-formation hypothesis could include the arguments thatsubordinates are sometimes valuable allies and that IHdoes not necessarily improve female–female cooperativerelationships, but is important merely to maintain theserelationships. Nevertheless, to provide a complete causalmodel of the evolution of infant handling, the alliance-formation hypothesis requires additional assumptionsthat are rarely made explicit. If tolerance or coalitionalsupport from high rankers is regarded as reciprocation forIH by low rankers (Fairbanks 1990), then IH must beshown to increase the mother’s reproductive success(e.g. by decreasing her next interbirth interval or increas-ing the infant’s survival chances). If IH does not pro-vide fitness benefits to the mother, the source of itshypothesized positive effect on the mother–handlerrelationship is considerably more obscure. For example,Kurland (1977) concluded for Japanese macaques, Macacafuscata, that IH provides little or no benefit to mothers orinfants, yet suggested that nulliparous handlers ‘solidifyintermatrilineal relationships between females’ merely byrefraining from harming their charges. Although it seemsobvious that a mother’s willingness to defend anotherfemale from aggression will rise while that female isholding her infant, it is much less obvious why thiswillingness would persist in any other context. As Nishida(1983) notes, grooming another female would appear, apriori, to be at least as effective as infant handling inbuilding female–female alliances.

In contrast, the bond-testing hypothesis explains theobservation that females preferentially handle the infantsof their close associates, without requiring that IH func-tion as infant care. Indeed, in this view, IH imposes asmall cost on the mother, but it is a cost she is willing tobear (and does not experience as unpleasant) when shevalues her bond with the handler. This tolerance bywhite-faced capuchin mothers seems on first consider-ation to falsify the assumption that IH is an imposition.However, as Zahavi & Zahavi (1997, page 112) argue, abond-testing imposition is expected to be escalatedgradually until the target individual experiences it asstressful and terminates it. The point at which this occursgives the tester rather precise information regarding thetarget individual’s current willingness to invest in thetester. If this argument is correct, then most bond-testinginteractions will end before they become noticeablystressful.

My results provided partial support (significant onlywhen IH episodes were treated as individual data points)for the prediction from the bond-testing hypothesis thatmothers will tolerate longer IH episodes from femalesthat they had groomed at higher rates before the birth of

919MANSON: CAPUCHIN INFANT HANDLING

their infant. One prediction generated from the bond-testing hypothesis was not supported in this study: theproportion of IH interactions that comprised touchingand rough handling was not significantly correlatedwith mother–handler grooming rate prior to the infant’sbirth. However, results based on a small sample ofmother–handler dyads showed a tendency (P<0.10) inthis direction.

All else being equal, social relationships characterizedby rapid change and opportunistic ‘shopping for profit-able partners’ (Noe 1992) should require more testingthan long-lasting kinship-based bonds. As an example ofthe former, Smuts & Watanabe (1990) proposed that maleolive baboons, Papio cynocephalus anubis, use greetingsinvolving intimate contact to test one another’s willing-ness to provide coalitional support (but see Silk 1994 foran alternative interpretation of similar greetings by malebonnet macaques, Macaca radiata). Females of a female-philopatric species like C. capucinus might seem unlikelycandidates for frequent users of bond-testing signals.However, recent findings from Lomas Barbual and thenearby study site of Santa Rosa (J. Manson, L. Rose,S. Perry & J. Gros-Louis, unpublished data) indicatethat female intergroup transfer in white-faced capuchinsis more common than previously believed and thatoccasional dominance struggles among females some-times involve shifting alliance patterns. For example,following the disappearance of beta female SS in late1997, alpha female AA became the target of coalitions ofsubordinates, and was peripheralized and possiblydeposed from alpha rank (J. Gros-Louis, personal com-munication). Thus, frequent bond testing could benefitC. capucinus females. A mother’s tolerance of long IHbouts might signal her short-term willingness (e.g. dur-ing the remainder of the same day) to provide coalition-ary support in response to the handler’s solicitations.Testing this prediction would require longer focal fol-lows than those carried out in this study, or all-occurrence sampling of coalitionary aggression.Maternal tolerance of long IH bouts, measured usingsurvival analysis techniques as in the current study,should predict a higher subsequent likelihood of themother aiding the handler. The alliance-formationhypothesis makes different predictions. If it assumesthat IH is a service by the handler to the mother, inexchange for the mother’s coalitionary support, thenthe length of time that the infant spends on the handlershould predict the mother’s likelihood of providingcoalitional support to the handler, irrespective ofwhether the mother or the handler terminates the IHbouts. In contrast to the bond-testing hypothesis, inter-acting with an infant while it is on its mother shouldnot predict maternal support of the handler, becausesuch interactions presumably provide no benefit tothe mother or the infant. If the alliance-formationhypothesis assumes that mere proximity of the handlerto the mother strengthens the bond between them,then duration of mother–handler proximity, regardlessof the frequency or type of interaction between handlerand infant, should predict the mother’s likelihood ofproviding coalitional support to the handler.

The bond-testing hypothesis predicts an interactionbetween kinship (not yet known for this study popu-lation) and strength of female–female social relationshipsin their effects on IH rate. Specifically, for a given level ofbond strength (measured by grooming and coalitionalsupport), female–female kin dyads should show lower IHrates than nonkin dyads, because kin should have lessneed than nonkin to test their social bonds.

During their first 90 days of life, C. capucinus infantsbegan to take an active role in interactions with adultfemales, primarily by climbing on them while remainingin contact with or in proximity to their mothers. Theseinteractions suggest that infant behaviour towardspotential handlers reflects infants’ evolved socialstrategies. The logic of bond testing could apply to infantbehaviour as well as to handler behaviour: infants mayacquire, from the reactions of adult females on whichthey attempt to climb, reliable information about thosefemales’ propensities to invest in them. Although I neverobserved females actively resisting infants’ attempts toclimb on, their reactions may have varied in ways toosubtle for my data collection protocol to detect.

Although the bond-testing hypothesis has somepromise for explaining C. capucinus infant handling, it isprobably unable to explain IH in species in which IH isdirected predominantly down the female dominancehierarchy (e.g. bonnet macaques: Silk 1980; yellowbaboons, Papio cynocephalus cynocephalus: Wasser 1983) oris demonstrably beneficial to infants and is directedpreferentially towards kin (e.g. vervets, Cercopithecusaethiops: Fairbanks 1988, 1990).

The nonadaptive hypothesis for IH is supported by myfinding that when infant EE was the only infant in thegroup that was 0–90 days old, he seemed to ‘absorb’ theinfant handling that was distributed among the fiveinfants born in a cluster during the previous year.Mothers’ general tolerance of IH, even though it did notappear to confer benefits on infants, could also be viewedas support for the nonadaptive hypothesis. However, thebehavioural components of C. capucinus IH differedsomewhat from the behaviour of mothers towards theirown infants. Mothers very rarely inspected or nuzzledtheir own infants, and were never observed to directmachine gun vocalizations towards them. This could beconsidered evidence against the view that IH is merely aby-product of selection for maternal behaviour (Emlenet al. 1991). However, shifting the unit of analysis fromspecific behaviours to evolved goals (the goal of inter-acting with infants) would support the nonadaptivehypothesis by questioning the relevance of the behav-ioural differences between IH and maternal behaviour fordeciding whether the former is merely a by-product ofselection for the latter. The subhypothesis that newbornsare attractive simply because capuchins are curious aboutnovel objects was not supported by an examination of thetemporal patterning of IH rates themselves, which didnot peak during the first week of life. However, IH ratesreflect an interaction between motivation to handleothers’ infants and maternal tolerance of handling,whichmay increase over time. Female approach rates to two ofthree new mothers in study 2 did peak during the first

920 ANIMAL BEHAVIOUR, 57, 4

week of infant life, as would be expected if newborns weremore attractive than slightly older infants.

Variables proposed to affect the form and frequency ofIH may interact in ways that I did not consider in detail.For example, high-ranking mothers could restrict accessto their infants to kin, whereas low-ranking mothersmight be constrained to tolerate IH by higher-rankingnonkin (O’Brien & Robinson 1991). However, this inter-action would lead to the expectation that mother–handler grooming rate would be less strongly associatedwith down-hierarchy IH than with up-hierarchy IH, apattern that I found in only one of two study periods.Alternatively, IH might occur disproportionately oftenamong kin only in high-ranking females, while those oflow ranks handle the infants of higher-ranking nonkin tocurry favour with them (see Seyfarth 1977, 1980 for asimilar argument regarding grooming). Kinship couldalso interact with handler age if mothers toleratehandling by nulliparous relatives but not by unrelatednullipara.

Finally, these findings represent a platyrrhine challengeto the common primatological view (McKenna 1979;Maestripieri 1994a) that rigid female dominancehierarchies and a despotic dominance style are associatedwith rough IH and maternal restrictiveness. Female–female relationships in this study group were character-ized by a stable dominance hierarchy over a 5-year period(Perry 1996; J. Manson, L. Rose, S. Perry & J. Gros-Louis,unpublished data), mutual coalitional support by highrankers against low rankers and spatial peripheralizationof the lowest-ranking female (Perry 1996), and noobserved tendency to reconcile following fights (Perry1995). Nevertheless, IH was rarely rough, kidnapping wasnot observed, and even low-ranking females practised alaissez-faire maternal style.

Acknowledgments

Kathryn Atkins, Todd Bishop, Julie Gros-Louis, SusanPerry and Laura Sirot participated in data collection.Study 1 was funded by the following awards to SusanPerry: National Geographic Society, two grants from theLeakey Foundation, an NSF graduate fellowship, SigmaXi, the University of Michigan Alumnae Society, aRackham Research Partnership grant (with BarbaraSmuts), a Rackham predoctoral fellowship, a Rackhamdissertation grant, NSF awards (to Barbara Smuts andJohn Mitani), and grants from the Evolution and HumanBehavior (EHB) Program of the University of Michigan.Susan Perry transcribed and computerized the data fromstudy 1 with help from Meta Landys, Sarah Newman,Laura Sirot and Michelle Tomaszycki. Norma Amaya A.,Marvin Cedillos A., Gordon and Jutta Frankie, KennethGlander and the Rosales family provided logistical assist-ance in Costa Rica. The Department of Anthropology,University of Alberta, provided office space and otherlogistical support while I carried out some of the dataanalyses presented here. The Costa Rican Servicio deParques Nacionales, Ministerio del Ambiente y Energiaand the Area de Conservacion Tempisque granted

permission to work in Lomas Barbudal Biological Reserve.Permission to work in adjoining areas was granted by LuisEcheverri Trujillo and the community of San Ramon deBagaces. I thank Lynn Fairbanks, Melissa Gerald, WarrenHolmes, Dario Maestripieri, Susan Perry, John Robinson,Joan Silk and an anonymous referee for helpful com-ments on early drafts of the manuscript, and ElizabethKaldor for discussions. Ken Glander kindly providedinformation on capuchin body size. The research pre-sented here was described in Animal Research ProtocolNo. 3081 approved on 29 October 1990 by the Com-mittee on Use and Care of Animals of the University ofMichigan, and Animal Research Protocol No. 96-122-02approved on 9 September 1996 by the Chancellor’sAnimal Research Committee of the University ofCalifornia, Los Angeles.

References

Allison, P. D. 1984. Event History Analysis: Regression for LongitudinalEvent Data. Newbury Park, California: Sage Publications.

Altmann, J. 1974. Observational study of behavior: samplingmethods. Behaviour, 49, 227–265.

Baldwin, J. D. 1969. The ontogeny of social behavior of squirrelmonkeys (Saimiri sciureus) in a semi-natural environment. FoliaPrimatologica, 11, 35–79.

Di Fiore, A. & Rendall, D. 1994. Evolution of social organization:a reappraisal for primates by using phylogenetic methods.Proceedings of the National Academy of Sciences, U.S.A., 91,9941–9945.

Emlen, S. T. 1991. The evolution of cooperative breeding in birdsand mammals. In: Behavioural Ecology: An Evolutionary Approach(Ed. by J. Krebs & N. Davies), pp. 301–337. Oxford: BlackwellScientific.

Emlen, S. T., Reeve, H. K., Sherman, P. W., Wrege, P. H.,Ratnieks, F. L. W. & Shellman-Reeve, J. 1991. Adaptive versusnonadaptive explanations of behavior: the case of alloparentalhelping. American Naturalist, 138, 259–270.

Fairbanks, L. A. 1988. Vervet monkey grandmothers: interactionswith infant grandoffspring. International Journal of Primatology, 9,425–441.

Fairbanks, L. A. 1990. Reciprocal benefits of allomothering forfemale vervet monkeys. Animal Behaviour, 40, 553–562.

Fedigan, L. M. 1993. Sex difference and intersexual relations inadult white-faced capuchins (Cebus capucinus). InternationalJournal of Primatology, 14, 853–877.

Frankie, G. W., Vinston, S. B., Newstrom, L. E. & Barthell, J. F.1988. Nest site and habitat preferences of Centris bees in theCosta Rican dry forest. Biotropica, 20, 301–310.

Goldizen, A. W. & Terborgh, J. 1989. Demography and dispersalpatterns of a tamarin population: possible causes of delayedbreeding. American Naturalist, 134, 208–224.

Hrdy, S. B. 1976. Care and exploitation of nonhuman primateinfants by conspecifics other than the mother. In: Advances in theStudy of Behavior (Ed. by J. S. Rosenblatt, R. A. Hinde, E. Shaw &C. Beer), pp. 101–158. New York: Academic Press.

Jay, P. C. 1963. Mother–infant relations in langurs. In: MaternalBehavior in Mammals (Ed. by H. L. Rheingold), pp. 282–304. NewYork: J. Wiley.

Kaplan, E. L. & Meier, P. 1958. Nonparametric estimation fromincomplete observations. Journal of the American StatisticalAssociation, 53, 457–481.

Kurland, J. A. 1977. Kin Selection in the Japanese Monkey. Basel:Karger.

921MANSON: CAPUCHIN INFANT HANDLING

Lancaster, J. 1971. Play-mothering: the relations between juv-enile females and young infants among free-ranging vervetmonkeys (Cercopithecus aethiops). Folia Primatologica, 15, 161–182.

McKenna, J. J. 1979. The evolution of allomothering behavioramong colobine monkeys: function and opportunism inevolution. American Anthropologist, 81, 818–840.

Maestripieri, D. 1994a. Social structure, infant handling, andmothering styles in group-living Old World monkeys. InternationalJournal of Primatology, 15, 531–553.

Maestripieri, D. 1994b. Influence of infants on female socialrelationships in monkeys. Folia Primatologica, 63, 192–202.

Mitchell, C. L., Boinski, S. & van Schaik, C.P. 1991. Com-petitive regimes and female bonding in two species of squirrelmonkeys (Saimiri oerstedi and S. sciureus). Behavioral Ecology andSociobiology, 28, 55–60.

Nicolson, N. A. 1987. Infants, mothers, and other females. In:Primate Societies (Ed. by B. B. Smuts, D. L. Cheney, R. M. Seyfarth,R. W. Wrangham & T. T. Struhsaker), pp. 330–342. Chicago:University of Chicago Press.

Nishida, T. 1983. Alloparental behavior in wild chimpanzees of theMahale mountains, Tanzania. Folia Primatologica, 41, 1–33.

Noe, R. 1992. Alliance formation among male baboons: shoppingfor profitable partners. In: Coalitions and Alliances in Humans andOther Animals (Ed. by A. H. Harcourt & F. B. M. de Waal),pp. 285–321. Oxford: Oxford University Press.

O’Brien, T. G. 1988. Parasitic nursing behavior in the wedge-capped capuchin monkey (Cebus olivaceus). American Journal ofPrimatology, 16, 341–344.

O’Brien, T. G. & Robinson, J. G. 1991. Allomaternal care byfemale wedge-capped capuchin monkeys: effects of age, rank andrelatedness. Behaviour, 119, 30–50.

Oppenheimer, J. R. & Oppenheimer, E. C. 1973. Preliminaryobservations of Cebus nigrivittatus (Primates: Cebidae) on theVenezuelan Llanos. Folia Primatologica, 19, 409–436.

Paul, A. & Kuester, J. 1996. Infant handling by female Barbarymacaques (Macaca sylvanus) at Affenberg Salem: testing func-tional and evolutionary hypotheses. Behavioral Ecology andSociobiology, 39, 133–145.

Perry, S. 1995. Social relationships in wild white-faced capuchinmonkeys, Cebus capucinus. Ph.D. thesis, University of Michigan,Ann Arbor, Michigan.

Perry, S. 1996. Female–female social relationships in wild white-faced capuchin monkeys, Cebus capucinus. American Journal ofPrimatology, 40, 167–182.

Perry, S. 1997. Male–female social relationships in wild white-facedcapuchins (Cebus capucinus). Behaviour, 134, 477–510.

Peto, R. & Peto, J. 1972. Asymptotically efficient rank in-variant procedures. Journal of the Royal Statistical Society, 135,185–207.

Quiatt, D. 1979. Aunts and mothers: adaptive implications ofallomaternal behavior of nonhuman primates. American Anthro-pologist, 81, 310–319.

Riedman, M. L. 1982. The evolution of alloparental care andadoption in mammals and birds. Quarterly Review of Biology, 57,405–435.

Robinson, J. G. & Janson, C. H. 1987. Capuchins, squirrel monkeys,and Atelines: socioecological convergence with Old Worldprimates. In: Primate Societies (Ed. by B. B. Smuts, D. L. Cheney,R. M. Seyfarth, R. W. Wrangham & T. T. Struhsaker), pp. 69–82.Chicago: University of Chicago Press.

van Schaik, C. P. 1989. The ecology of social relationships amongstfemale primates. In: Comparative Socioecology: The BehaviouralEcology of Humans and Other Mammals (Ed. by V. Standen & R. A.Foley), pp. 195–218. Oxford: Blackwell Scientific.

Scollay, P. A. & DeBold, P. 1980. Allomothering in a captive colonyof Hanuman langurs (Presbytis entellus). Ethology and Sociobiology,1, 291–299.

Seyfarth, R. M. 1977. A model of social grooming among adultfemale monkeys. Journal of Theoretical Biology, 65, 671–698.

Seyfarth, R. M. 1980. The distribution of grooming and relatedbehaviours among adult female vervet monkeys. AnimalBehaviour, 28, 798–813.

Silk, J. B. 1980. Kidnapping and female competition among captivebonnet macaques. Primates, 21, 100–110.

Silk, J.B. 1994. Social relationships of male bonnet macaques: malebonding in a matrilineal society. Behaviour, 130, 271–291.

Smuts, B. B. & Watanabe, J. 1990. Social relationships and ritual-ized greetings in adult male baboons (Papio cynocephalus anubis).International Journal of Primatology, 11, 147–172.

Stanford, C. B. 1992. Costs and benefits of allomothering in wildcapped langurs (Presbytis pileata). Behavioral Ecology and Socio-biology, 30, 29–34.

Strier, K. B. 1994. Myth of the typical primate. Yearbook of PhysicalAnthropology, 37, 233–271.

Tardif, S. D., Carson, R. L. & Gangaware, B. L. 1992. Infant-carebehavior of non-reproductive helpers in a communal-careprimate, the cotton-top tamarin (Saguinus oedipus). Ethology, 92,155–167.

Trivers, R. L. 1971. The evolution of reciprocal altruism. QuarterlyReview of Biology, 46, 35–57.

Wasser, S. K. 1983. Reproductive competition and cooperationamong female yellow baboons. In: Social Behavior of FemaleVertebrates (Ed. by S. K. Wasser), pp. 349–390. New York:Academic Press.

Wasser, S. K. & Barash, D. P. 1981. The selfish ‘allomother’: acomment on Scollay and DeBold (1980). Ethology and Socio-biology, 2, 91–93.

Williams, L., Gibson, S., McDaniel, M., Bazzel, J., Barnes, S. &Abee, C. 1994. Allomaternal interactions in the Bolivian squirrelmonkey (Saimiri boliviensis boliviensis). American Journal ofPrimatology, 34, 145–156.

Zahavi, A. 1977. The testing of a bond. Animal Behaviour, 25,246–247.

Zahavi, A. & Zahavi, A. 1997. The Handicap Principle: A Missing Pieceof Darwin’s Puzzle. New York: Oxford University Press.