Anim. Behav., 1998, 55, 1063–1069
Five primate species follow the visual gaze of conspecifics
MICHAEL TOMASELLO, JOSEP CALL & BRIAN HARE
Department of Psychology
and
Yerkes Regional Primate Research Center, Emory University
(Received 28 April 1997; initial acceptance 18 June 1997;
final acceptance 8 August 1997; MS. number:
7804
)
Abstract. Individuals from five primate species were tested experimentally for their ability to follow the
visual gaze of conspecifics to an outside object. Subjects were from captive social groups of
chimpanzees, Pan troglodytes, sooty mangabeys, Cercocebus atys torquatus, rhesus macaques, Macaca
mulatta, stumptail macaques, M. arctoides, and pigtail macaques, M. nemestrina. Experimental trials
consisted of an experimenter inducing one individual to look at food being displayed, and then
observing the reaction of another individual (the subject) that was looking at that individual (not the
food). Control trials consisted of an experimenter displaying the food in an identical manner when
the subject was alone. Individuals from all species reliably followed the gaze of conspecifics, looking to
the food about 80% of the time in experimental trials, compared with about 20% of the time in control
trials. Results are discussed in terms of both the proximate mechanisms that might be involved and the
adaptive functions that might be served by gaze-following.
1998 The Association for the Study of Animal Behaviour
The ability to follow the direction of conspecifics’
visual gaze would seem to be a social skill with
immediate adaptive benefits. Following the gaze
of others might help individuals perceive import-
ant entities in the environment such as food,
predators, and certain kinds of social interactions
among group-mates. This social cue might be
especially helpful for highly social species, such as
many primates, that need constant information
about group-mates’ activities. A number of field
primatologists have made informal observations
suggesting that some primate species may follow
the gaze of conspecifics: for hamadryas baboons,
Papio hamadryas (Kummer 1967); for longtail
macaques, Macaca fascicularis (de Waal et al.
1976); for savannah baboons, Papio cynocephalus
(Byrne & Whiten 1992); and for chimpanzees, Pan
troglodytes (Plooij 1978).
The problem is that in each of these cases there
may have been some external stimulus that both
primate individuals perceived and oriented to and
that the human observer did not perceive, raising
the possibility that gaze direction was not in fact
the effective cue. Menzel (1973, 1974) reported a
series of experiments with chimpanzees in which
this could not have been the case. In all of these
studies, one captive chimpanzee knew the location
of food while its group-mates did not. When
allowed to search as a group, the group-mates
were able to use the behaviour of the knowledge-
able individual to determine successfully the
food’s location. In this situation, however, the
knowledgeable individual provided the group-
mates with a host of cues to the food’s location,
perhaps most importantly, direction of travel.
There was thus no way in these experiments to
know whether gaze direction by itself was respon-
sible for the group-mates’ success in finding the
food.
The few experimental studies with primates
conducted on this issue all have used human
experimenters as the animate being whose gaze
was being followed. For example, Povinelli &
Eddy (1996, study 1) had six 5- to 6-year-old
chimpanzees individually enter a room, where-
upon they encountered a human experimenter
facing them. As soon as the subject had spied her,
Correspondence: M. Tomasello, Department of Psy-
chology, Emory University, Atlanta, GA, 30322, U.S.A.
(email: [email protected]).
0003–3472/98/041063+07 $25.00/0/ar970636 1998 The Association for the Study of Animal Behaviour
1063
the human oriented both her eyes and head to a
distinct location, in some cases to the corners of
the room above and behind the subject (see also
Povinelli & Eddy, in press). In another condition,
the experimenter moved only her eyes toward
these same locations. In both of these situations,
the chimpanzees looked where the human was
orienting more than in a baseline condition in
which the human looked directly at the subject.
This result held even when the chimpanzee
encountered the human already looking at a loca-
tion, thus demonstrating that movement of the
head and eyes was not a critical social cue.
In a similar study with a different experimental
paradigm, Anderson et al. (1995) found negative
results with a pair of capuchin monkeys, Cebus
apella. In this study, an experimenter secretly
baited one of two food wells and then presented
them to the subject. Using several social cues over
different trials, direction of eye gaze (including
head direction) was not an effective cue in orient-
ing subjects to the hidden food (but touching the
baited food well was). Itakura & Anderson (1996)
trained a single capuchin monkey to follow
human eye gaze to hidden food in a similar
situation, but it took them over 120 trials to do so,
suggesting the possibility that gaze direction was
learned as a straightforward discriminative cue.
In the only study to investigate multiple species
within the same paradigm, Itakura (1996) investi-
gated 11 primate species (two species of lemur,
two species of cebus monkey, one species of
squirrel monkey, four species of macaque, and
two species of great ape). For all individuals a
human experimenter approached their cage and
tried to make eye contact, at which point he
then fixed his gaze behind the subject, either to
the right or to the left. In some cases, the gaze
was accompanied by a pointing gesture. When
there was no pointing, only the single orangutan,
Pongo pygmaeus (out of 40 subjects total) reliably
oriented in the direction of the human’s gaze.
When there was pointing, the most frequent
response from most species was to ignore the
experimenter’s gaze and pointing. Given that they
did respond, however, subjects of 10 of the 12
species followed the gaze and pointing of the
human more than they looked in the other direc-
tion (the exceptions were the squirrel monkeys
and the pigtail macaques). One possible explana-
tion is that the pointing gesture simply caused
individuals to look at the hand as it was being
extended, and thus to begin orienting in the
direction in which the hand was moving, without
necessarily understanding the social significance
of the pointing or the gaze.
Overall, then, there is inconsistent evidence that
non-human primates can follow the gaze of
humans to specific locations. The only solid exper-
imental evidence, in the absence of other cues such
as pointing or direction of travel, comes from six
chimpanzees in one study and a single orangutan
in another study, thus suggesting the possibility
of ape–monkey differences of social cognition as
hypothesized by some primatologists (e.g. de
Waal & Luttrel 1988; Byrne & Whiten 1992; but
see Tomasello & Call 1994 foradifferent view).
Moreover, there is no experimental evidence, and
only scattered anecdotal reports for only a few
species, that non-human primates can follow the
gaze of conspecifics under any conditions. The
purpose of the current study, therefore, was to
investigate experimentally the ability of five pri-
mate species from three different genera to follow
the gaze of conspecifics to a relatively distal object
within a relatively natural social setting.
METHODS
Subjects were housed in social groups in relatively
large enclosures at the Yerkes Regional Primate
Research Center Field Station. Enclosures ranged
from about 1515 m to about 3030 m. The
group sizes were as follows: 15 rhesus macaques,
Macaca mulatta, 38 stumptail macaques, Macaca
arctoides, 44 pigtail macaques, Macaca nemest-
rina, 18 chimpanzees, P. troglodytes, and 28 sooty
mangabeys, Cercocebus atys torquatus. All groups
were composed of individuals of both genders;
ages ranged from juveniles to adults. No feeding
or other caretaking activities were modified for
the current study.
Each group was observed by an experimenter
from a 6- to 8-m high observation tower overlook-
ing the group’s enclosure (observation distances
were approximately 8–30 m). An experimental
trial was as follows. The experimenter identified a
situation in which two individuals were in prox-
imity to one another, one facing away from the
tower (the subject) and one at least partially facing
the tower (the conspecific). The experimenter then
held up a preferred food item (orange) in an
attempt to gain the attention of the conspecific.
Animal Behaviour, 55, 4
1064
The behaviour of the subject was then observed
for the next 10 s. Control trials were identical to
experimental trials, except that there was no con-
specific present in the immediate vicinity of the
subject. That is, the experimenter identified an
individual subject not facing the tower, held up
the food, and observed its behaviour for the next
10 s (Fig. 1). We videotaped all trials of both types
from the tower for subsequent analysis.
There were an average of 72.4 trials across
species (range=120 trials for the mangabeys and
44 trials for the stumptail macaques). For two
species (chimpanzees and stumptail macaques) the
experimenter could reliably identify individuals.
For chimpanzees, 15 individuals were observed as
subjects (two to nine trials per subject); for stump-
tails, 17 individuals were observed as subjects (one
to four trials per subject). In the other species, we
took care to sample from as many different indi-
viduals as possible; that is, for all species, the
experimenter was careful to sample from different
individuals across adjacent trials and to sample
across various sectors of the compound over time.
Comparisons between sectors revealed that for no
species was any one sector predominant (for
example, in no case were trials taken from the
most frequent sector more than twice as frequent
as the nearest competitor).
Scoring was conducted by two independent
observers viewing the videotapes. For control
trials, each observer viewed each trial and deter-
mined whether the subject looked to the food
within the 10-s window (as in all studies of gaze-
following, head direction served as the main
operationalization of gaze direction). Because the
food was held up directly above the camera, this
determination (look, no look) was relatively
straightforward. For experimental trials, ob-
servers also determined whether the conspecific
looked up as it was supposed to and, when it did,
whether the subject noticed the conspecific’s look
(as quality checks on the experimental manipula-
tion). Thus, we categorized trials intended to be
experimental as: ‘No gaze by conspecific’ (the
conspecific did not look, so the experimental cue
was not available to the subject); ‘Gaze not per-
ceived by subject’ (the conspecific looked but the
subject did not notice, and so the experimental cue
was potentially available but not used by the
subject); and ‘Gaze perceived by subject’ (con-
specific looked and subject noticed: the only truly
experimental trials). For all three of these types of
trial, we also recorded the subject’s response
within the 10-s window (look, no look). The
‘No gaze’ and the ‘Gaze not perceived’ trials may
be thought of as secondary control conditions,
because they are identical to the experimental
trials except that, respectively: (1) the con-
specific did not provide the gaze cue; or (2) the
subject did not perceive the gaze cue given by the
conspecific.
Across all species, inter-observer reliability was
computed by means of a Cohen’s kappa between
the two observers (Bakeman & Gottman 1986).
Observers agreed whether or not a subject looked
on 97% of the control trials, yielding a kappa of
0.89. We categorized each experimental trial into
one of four mutually exclusive and exhaustive
categories created by crossing the two determina-
tions: experimental versus secondary control (No
gaze and Gaze not perceived) and Subject look
versus Subject no look. The two observers agreed
Figure 1. (a) A stumptail macaque conspecific (top)
looks towards food with a subject (bottom) watching.
(b) The subject follows the conspecific’s look towards
food.
Tomasello et al.: Primate gaze-following
1065
on the category for 92% of the experimental trials,
yielding a kappa of 0.83.
RESULTS
Subjects in all five species looked at the food
significantly more often in the Gaze perceived
trials than in the control trials (rhesus: ÷
2
1
=17.8,
N=42, P<0.001; pigtails: ÷
2
1
=16.0, N=54,
P<0.001; stumptails: ÷
2
1
=32.6, N=40, P<0.001;
chimpanzees: ÷
2
1
=15.8, N=56, P<0.001; manga-
beys: ÷
2
1
=30.2, N=88, P<0.001; Fig. 2). Subjects
in all five species also looked at the food signifi-
cantly more often in the Gaze perceived trials than
in either of the secondary control conditions.
They looked more in the Gaze perceived trials
than in the No gaze trials (rhesus: ÷
2
1
=10.7,
N=42, P<0.001; pigtails: ÷
2
1
=5.4, N=28, P<0.05;
stumptails: ÷
2
1
=34.0, N=34, P<0.001; chimpan-
zees: ÷
2
1
=10.0, N=29, P<0.01; mangabeys:
÷
2
1
=11.1, N=36, P<0.001). They also looked
more in the Gaze perceived trials than in the
Gaze not perceived trials (rhesus: N=26, P<0.03,
Fisher’s exact test; pigtails: ÷
2
1
=23.4, N=41,
P<0.001; stumptails: N=26, P<0.001, Fisher’s
exact test; chimpanzees: ÷
2
1
=15.2, N=36,
P<0.001; mangabeys: ÷
2
1
=16.7, N=50, P<0.001).
None of the three control conditions differed
significantly from one another.
In characterizing the behaviour of subjects,
latency to respond to the conspecific’s look
seemed to be important. Latency to look was thus
determined for each trial, with a mean and median
value computed for each species. The mean values
(in seconds) were: rhesus: 0.92, stumptails: 1.04,
pigtails: 0.46, mangabeys: 1.68, and chimpanzees:
1.00. The median response time for four of the
species was 0 s (i.e. less than 1 s); rhesus had a
median value of 0.50. Thus, all species responded
almost immediately after spying the gaze cue.
Vocalizations almost never occurred in the trials,
either from the conspecific or subject. Oranges are
a food all five of the subject species like very
much, but this fruit did not elicit an extreme
response because subjects ate oranges almost
daily.
An argument could be made that the Gaze not
perceived trials should be grouped with the exper-
imental trials, because the conspecific looked, and
thus provided a potential cue for the subject, even
though the subject did not notice this cue as
reliably judged by two independent observers.
When we grouped subjects in this way, results still
100
0
No gaze
Condition
Percentage of trials
80
60
40
20
Control Gaze not perceived Gaze perceived
Rhesus
Chimpanzees
Mangabeys
Stumptails
Pigtails
*
*
*
*
*
Figure 2. Percentage of trials in which subjects of different species looked towards food as a function of experimental
condition. *P<0.05.
Animal Behaviour, 55, 4
1066
corroborate the previous findings in that sub-
jects in four of the five species still looked at the
food significantly more often in the new exper-
imental condition than in the control condition
(rhesus: ÷
2
1
=11.7, N=52, P<0.001; stumptails:
÷
2
1
=23.2, N=44, P<0.001; chimpanzees: ÷
2
1
=5.0,
N=74, P<0.03; mangabeys: ÷
2
1
=13.3, N=120,
P<0.001; Fig. 3). Pigtail macaques also showed
the same pattern, although these results
were not statistically significant (÷
2
1
=3.3, N=72,
P<0.08).
Even though subjects looked at the food in the
vast majority of the Gaze perceived trials across
all species, they did not do so in all trials. We
therefore explored the possibility that some
behaviour patterns in the conspecific may have
facilitated the subjects’ gaze-following. We looked
at two behaviour patterns specifically: (1) the
presence of a distinctive head movement; and
(2) other salient behaviour (for example, head
bobbing) that might have attracted the subjects’
attention. We thus noted the presence of these
behaviour patterns for each Gaze perceived trial
and related this to the outcome of that trial.
Neither the conspecific’s head movement nor any
other salient behaviour was reliably associated
with subjects’ looks to the food (÷
2
, ); all five
species showed a similar pattern of results.
DISCUSSION
In the current study we found that five primate
species from three different genera reliably fol-
lowed the gaze of conspecifics to external objects
over 80% of the time when the conspecific pro-
vided a clear gaze cue that was noticed by the
subject. Subjects of all five species did this in an
experimental setting in which the possibility of
their noticing the external objects for themselves
was ruled out, as evidenced by the small amount of
looking (less than 20% of the time) in the control
conditions (control, No gaze, and Gaze not per-
ceived). When subjects did perceive the gaze of
conspecifics, they looked in the same direction
almost immediately thereafter (less than 1 s).
Although primate field workers have long sus-
pected that primates use the gaze direction of
others as a cue to the presence of external entities,
the current findings are the first to document this
fact experimentally. Our findings further suggest
that studies in which primates do not follow the
gaze direction of humans (e.g. Itakura 1996)may
reflect more of a motivational problem than a
competence problem. Primates are much more
interested in where conspecifics are looking than
in where humans are looking. The main possibility
for ambiguity in the current findings is how the
Control
100
0
Condition
Percentage of trials
80
60
40
20
Gaze perceived +
Gaze not perceived
Rhesus
Chimpanzees
Mangabeys
Stumptails
Pigtails
**
*
**
**
P < 0.07
Figure 3. Percentage of trials in which subjects of different species looked towards food as a function of experimental
condition, with Gaze not perceived and Gaze perceived trials combined. *P<0.05; **P<0.001.
Tomasello et al.: Primate gaze-following
1067
Gaze not perceived trials were treated. But for
four of the five species it made no statistical
difference whether these trials were included in the
experimental trials (and for the other species
P<0.08). This finding suggests that not only can
primates use gaze direction as a cue when they
notice it, they often do use it when it is produced
in their immediate vicinity.
We did not attempt to determine precisely what
behavioural cues were being used by the subjects
in the current study. In particular, we do not
know whether individuals used head direction in
general, or something more specifically with the
eyes. However, body orientation in general could
not have been the cue, because in both the No
gaze and Gaze not perceived conditions the
conspecific was oriented toward the food. We do
not know whether the affective state of the con-
specific while gazing was a part of the cue, but
the conspecifics almost never vocalized (which
they often do if highly desirable food is present)
and we did not notice many obvious signs of
emotional arousal during the looks. Neverthe-
less, if the conspecifics in this study had spied an
object that frightened them, it is possible that the
subjects would have noticed a different affective
component to their gaze behaviour and behaved
differently. Further research is needed to deter-
mine whether these additional affective cues are
reliably used by primates in their everyday social
interactions in which they follow conspecific gaze.
The current results provide no evidence for a
difference between monkey and ape social behav-
iour or cognition. Both monkeys and apes used
the gaze direction of conspecifics to locate food,
and they did so equally well. This is not surprising
from a functional point of view, because there are
no differences in the social organization or behav-
iour of the species of monkeys and apes tested
that would lead us to suspect different processes of
social behaviour or social cognition (Tomasello
& Call 1994, 1997). All of the species in the
current study are capable of using something like
the ‘attentional structure’ of conspecifics to gather
information about external events (Chance 1967).
These results say nothing about how primates
understand the significance of conspecifics’ gaze in
terms of the attentional or mental states involved.
The fact that primates follow the gaze of con-
specifics may have either a ‘rich’ interpretation,
in terms of primates’ understanding of the experi-
ence or mental states of conspecifics, or a ‘lean’
interpretation, in terms of primates’ use of social
cues in their creation of social and other behav-
ioural strategies (Povinelli & Eddy, in press). Only
further research can answer the question of the
most appropriate level of interpretation. To be
ecologically valid, such research should investi-
gate primates’ understanding of the visual gaze,
affective states, and possibly mental states, not of
human beings but of conspecifics.
ACKNOWLEDGMENTS
This investigation was supported in part by a
grant from the National Science Foundation
to M.T. (IBN-9507418) and in part by NIH grant
RR-00165 from the National Center for Research
Resources to the Yerkes Regional Primate
Research Center. B.H. was supported by an REU
supplement to the NSF grant and by a Hughes
Foundation Summer Research Fellowship. The
Yerkes Center is fully accredited by the American
Association of Laboratory Animal Care.
REFERENCES
Anderson, J. R., Sallaberry, P. & Barbier, H. 1995. Use
of experimenter-given cues during object-choice tasks
by capuchin monkeys. Anim. Behav., 49, 201–208.
Bakeman, R. & Gottman, J. 1986. Observing Inter-
action: An Introduction to Sequential Analysis.
Cambridge: Cambridge University Press.
Byrne, R. W. & Whiten, A. 1992. Cognitive evolution in
primates: evidence from tactical deception. Man, 27,
609–627.
Chance, M. 1967. Attention structure as the basis of
rank order. Man, 2, 503–518.
Itakura, S. 1996. An exploratory study of gaze-
monitoring in non-human primates. Jpn Psychol.
Res., 38, 174–180.
Itakura, S. & Anderson, J. 1996. Learning to use
experimenter-given cues during an object-choice task
by a capuchin monkey. Curr. Psychol. Cogn., 15,
103–112.
Kummer, H. 1967. Tripartite relations in hamadryas
baboons. In: Social Communication Among Primates
(Ed. by S. A. Altmann), pp. 63–71. Chicago: The
University of Chicago Press.
Menzel, E. W., Jr. 1973. Leadership and communication
in young chimpanzees. In: Precultural Primate Behav-
iour (Ed. by E. W. Menzel, Jr), pp. 192–225. Basel:
Karger.
Menzel, E. W., Jr. 1974. A group of young chimpanzees
in a one-acre field: leadership and communication. In:
Behaviour of Non-Human Primates (Ed. by A. M.
Schrier & F. Stollnitz), pp. 83–153. New York:
Academic Press.
Animal Behaviour, 55, 4
1068
Plooij, F. X. 1978. Some basics traits of language in wild
chimpanzees? In: Action, Gesture and Symbol (Ed. by
A. Lock), pp. 111–131. London: Academic Press.
Povinelli, D. J. & Eddy, T. J. 1996. Chimpanzees: joint
visual attention. Psychol. Sci., 7, 129–135.
Povinelli, D. J. & Eddy, T. J. In press. Specificity of
gaze-following in young chimpanzees. Br. J. devl.
Psychol.
Tomasello, M. & Call, J. 1994. Social cognition of
monkeys and apes. Ybk phys. Anthropol., 37, 273–305.
Tomasello, M. & Call, J. 1997. Primate Cognition.
Oxford: Oxford University Press.
de Waal, F. B. M. & Lutrell, L. M. 1988. Mechanisms of
social reciprocity in three primate species: symmetrical
relationship characteristics or cognition? Ethol. Socio-
biol., 9, 101–118.
de Waal, F. B. M., van Hooff, J. A. R. A. M. & Netto,
W. J. 1976. An ethological analysis of types of ago-
nistic interaction in a captive group of Java-monkeys
(Macaca fascicularis). Primates, 17, 257–290.
Tomasello et al.: Primate gaze-following
1069
