ÐÓ°ÉÔ­´´

For anyone interested in how humans think, there is something especially
fascinating about other primates. The social behaviour of nonhuman primates
offers a glimpse of minds that have similarities with our own but fall short
in important respects. Examining the workings of these ‘almost minds’ may
shed light on the origins of human language, cognition and self-awareness.
It may ultimately tell us how, in the course of our own evolution, some
minds gained an advantage over others. This evolutionary perspective draws
attention to the many apparent similarities between human and nonhuman primate
behaviour.

The goal of our own research, however, has been to establish where these
similarities between us and nonhuman primates break down. We approach the
study of primate intelligence from a practical, functional perspective.
What problems do monkeys face in their daily lives? What do they need to
know, and how might one method of obtaining and storing knowledge give certain
individuals a reproductive advantage over others? Eventually, this approach
could reveal not only how human intelligence evolved, but why.

To document what monkeys and apes do in their natural habitats, we observed
East African vervet monkeys (Cercopithecus aethiops) in the Amboseli National
Park in southern Kenya, and compared our own observations with studies of
other primate species. Our aim was to consider what sorts of underlying
mental operations might possibly account for the monkeys’ behaviour. From
observing a particular event, do the monkeys make the same deductions that
we do? Do they understand kinship and social rankings? Or are they acting
out complex strategies without being in any sense aware of what they are
doing? To probe further into the minds of our subjects, we conducted experiments
designed to test whether the patterns of social behaviour we humans see
might also exist in the minds of the monkeys themselves.

We can think of social groups of monkeys and apes as being composed
of many different long-term and short-term alliances among related and unrelated
animals. To gain a social and reproductive advantage over others, individuals
must be able not only to predict each other’s behaviour, but also to assess
relationships. It is not enough to know who is dominant or subordinate to
oneself; an individual must also know who is allied to whom and who is likely
to come to an opponent’s aid. For this reason, we might predict that individual
monkeys would be sensitive to other animals’ relationships in any species
in which alliances are common.

Monkeys do seem to recognise the social relationships that exist among
other group members, and judgments about these relationships seem to underlie
much of their behaviour. Males assess the closeness of bonds between other
males and their females before attempting a takeover. Females assess the
ranks of others when competing for grooming partners. And females and juveniles
apparently recognise the ways in which monkeys related through their mother
(matrilineal kin) act together in unison, as they direct reconciliation
or retaliation not only to individual opponents, but to their opponents’
kin as well. In fact, vervets sometimes trigger fights in apparent efforts
to retaliate against a slight received by another family member. If, for
example, juvenile female Shelley observes her sister involved in a fight
with juvenile male Trollope, it is likely that Shelley will threaten a close
relative to Trollope.

Knowledge of other animals’ social relationships can only be obtained
by observing the behaviour of others and making appropriate deductions.
Primates may not be limited to an egocentric view of the world: they may
be able to step outside their own immediate experience to make judgments
about the experiences of others. Monkeys, in this respect, seem to differ
from Anthony Powell’s infamous character Widmerpool, who was ‘one of those
persons capable of envisioning others only in relation to himself’.

The assessments that monkeys make of one another, moreover, are not
simple, but seem to occur along at least two dimensions simultaneously.
They classify individuals on the basis of kinship or close association at
the same time as they recognise differences in rank. Thus two individuals
may sometimes be lumped together as members of the same family, while at
other times they may be considered separately, one ranking higher than the
other. A female who recognises that sisters B and C are closely associated
may threaten B after being involved with a fight with C. Nevertheless, the
female is also capable of distinguishing B’s and C’s relative ranks, and
in other contexts may groom the higher ranking B in preference to C.

We can make no precise statements about the mechanisms that underlie
the monkeys’ knowledge of each others’ relationships. For example, even
if monkeys do have abstract concepts such as ‘closely bonded’, their knowledge
of these bonds may derive principally from noticing that certain individuals
spend a lot of time with each other. Similarly, we cannot be sure that monkeys
constructing a dominance hierarchy engage in ‘transitive inference’ – in
other words, that having seen that A is dominant to B and B is dominant
to C, they infer that A is dominant to C. There is evidence that they do,
but it is also possible that they simply memorise the outcome of every interaction
between other individuals.

The ability to classify others into abstract categories such as ‘closely
bonded’ would have at least two functional advantages. First, it would allow
individuals to identify types of relationships quickly, and to predict the
behaviour of others based on partial information. As a result, a monkey
who joins a new group, or whose group receives an influx of migrants, could
make accurate predictions about behaviour without having to observe the
interactions of every individual with every other. Secondly, as group size
increases, the ability to form categories and to make judgments based on
these categories would provide an increasingly efficient method for recognising
the characteristics of relationships and predicting what specific individuals
are likely to do next.

If we grant that a monkey has concepts, can we also conclude that the
monkey recognises the existence of concepts in others? Does a monkey’s knowledge
of its social companions include knowledge – or at least a theory – about
its companions’ mental states?

On 1 October 1972, the Sunday Times printed the obituary of Flo, an
adult female chimpanzee who had lived in the Gombe National Park in Tanzania,
and who had been studied by Jane Goodall for more than 11 years. An excerpt
from this obituary read as follows: ‘Flo has contributed much to science.
She and her large family have provided a wealth of information about chimpanzee
behaviour,infant development, family relationships, aggression, dominance,
sex . . . But this should not be the final word. It is true that her life
was worthwhile because it enriched human understanding. But even if no one
had studied the chimpanzees at Gombe, Flo’s life, rich and full of vigour
and love, would still have had a meaning and a significance in the pattern
of things.’

This final sentence highlights an important question for biologists.
If no human observer had ever interpreted Flo’s life, could we still say
that she had knowledge, motives, beliefs and desires, and that her life
was full of vigour and love? Do such mental states really exist in the mind
of any animal? Or are they artefacts, invented by ethologists as the best
means of describing what they have seen? When we watch nonhuman primates
in the wild and analyse their social behaviour, do we have their minds or
ours under the microscope?

Stategic observations

Watching monkeys, one is tempted to treat them like tiny humans not
only because they look rather like us but also because features of their
social organisation, for example close bonds among kin and status-striving,
look like simplified versions of our own. More important, anthropomorphising
works: attributing motives and strategies to animals is often the best way
for an observer to predict what an individual is likely to do next.

Descriptions of social behaviour in anthropomorphic terms do not, however,
constitute an explanation. After all, one way to describe and predict the
behaviour of a cash dispenser is to assume that it wants to help you to
do your banking, even though this motive plays no part in the machine’s
operation. So a key goal of our research has been to dissect the knowledge
and motives that make monkeys do what they do. Can we prove that the mechanisms
that govern behaviour and communication in vervet monkeys are similar to
our own? If they are not, what are the differences? What can we do that
monkeys cannot, and how does this make their lives and their view of the
world different from our own?

Vervet monkeys, like many other animals, recognise individuals and take
note of the interactions that occur among them. The monkeys may also create
in their minds a num-ber of representations that describe different sorts
of social relationships: for example, mother-offspring relationships, relationships
among kin and friendships between males and females. Experiments with captive
monkeys have shown that primates compare relationships according to the
types of bond they exemplify and not just according to the individuals involved.
They judge mother-offspring pairs as the same even if one involves a mother
and her adult daughter and the other involves a mother and her infant son.
Monkeys appear to make use of such representations in their daily lives.
A vervet monkey is more likely to threaten another female if one of her
own close relatives and one of her opponent’s close relatives have recently
fought. Such behaviour is difficult to explain unless vervets recognise
some similarity between their own close bonds and the close bonds of others.

Any mental representations monkeys have of social relationships are
probably not as abstract and flexible as our own. Language allows us to
label different types of relationships, to specify the criteria by which
to include a bond in one class or another, and to discuss types of relationships
independently of individuals, in general, abstract terms. There is as yet
no evidence that monkeys’ knowledge of social relationships is accessible
to them in a similar way or that their classification of social bonds is
abstract enough to include the identification of unfamiliar individuals
or social structures. We suspect that a monkey raised among families A,
B, C, D and E could tell us that the bond between Mother A and Infant A1
is the same sort of bond as that between Mother C and Infant C₁.
We do not know, however, if the same monkey could deduce that bonds just
like this occur in other vervet groups, let alone in other species.

Humans not only classify social relationships into types but also examine
the motives and strategies of others in an attempt to explain why some relationships
are alike and others different. When trying to understand behaviour, humans
often use introspection as a guide. Knowing that our own actions are often
caused by particular mental states, we look for the same processes in others.

We have as yet no evidence that monkeys are aware of their own knowledge
or attribute mental states to others. Monkeys are doubtless excellent at
monitoring and predicting each other’s actions, and they probably have little
difficulty recognising that monkey X’s actions can have a particular effect
on Y’s behaviour. It seems unlikely, however, that monkeys take into account
each other’s thoughts, motives or beliefs when they assess what other individuals
are likely to do next.

For example, although their vocalisations certainly function to alert
others to the presence of food, danger or each other, monkeys do not adjust
their calls according to whether or not their audience is ignorant or informed,
and in this sense cannot be said to communicate with the intention of changing
a listerner’s mental state. Similarly, while monkeys are clearly able to
acquire novel skills from others through observation, social enhancement
and trial and error learning, there is little evidence that they teach or
even imitate each other, perhaps because they cannot distinguish between
their own states of mind and the states of mind of others.

Even in the case of deception, monkeys’ attempts at manipulation seem
aimed more at altering their rivals’ behaviour than at affecting their rivals’
thoughts. Deception, in the human sense, requires the attribution of mental
states to oneself and to others. When we lie to someone we recognise a distinction
between our own and another person’s thoughts, and we depend on the fact
that a person’s beliefs can affect subsequent behaviour. If we are correct
in arguing that monkeys live in a world without attribution, it would also
seem to follow that they live in a world without deception.

The issue, however, is more complicated than this. Many animals do act
in ways that serve to deceive others. Great tits, for example, give apparently
deceptive alarm calls at feeding perches, and they vary their false alarm
calls depending upon who is nearby. If the birds at the feeding perch are
lower ranking than the signaller, false alarm calls are rarely given, presumably
because the caller can simply supplant the rivals by approaching. But when
the birds that are feeding are of higher rank and so cannot be supplanted,
the lower-ranking birds do give false alarm calls. We have no evidence,
however, that the birds use any other signals to deceive each other, or
that they use deceptive signals in any other social context.

Such deception in animals is less flexible than human deception. We
suspect that many of the differences between human and nonhuman deception
ultimately derive from the failure of most animal species to attribute mental
states to others. Human deception rests on the assumption that other individuals
have mental states that can be manipulated to one’s own benefit. Armed with
this general theory, humans can modify deceptive behaviour widely, within
a single context or from one context to the next. By contrast, animal deception
seems to rest primarily on the recognition of certain behavioural contingencies:
if I do this, he will do that. It must be learnt and relearnt from one circumstance
to the next.

We do not yet know how flexible deception is in monkeys, nor indeed
how flexible the detection of deception is. For example, a subordinate male
who wishes to mate with a sexually receptive female will often attempt to
‘deceive’ a more dominant male by leading the female behind a bush or a
rock before he mates with her. If a dominant male learnt to be wary of a
male who attempted to sneak copulations in this way, would he also begin
to distrust the male’s alarm calls? We suspect that he would not, because
to do so would demand that the dominant male would have some understanding
of the other male’s motives, something he may be incapable of doing if he
cannot recognise mental states in others. The dominant male might continue
to be distrustful of his subordinate rival whenever his rival approached
a female, but still continue to run into trees whenever his rival uttered
an alarm call.

Apes may be considerably better than monkeys at attributing mental states
to others. Chimpanzees, for example, do seem to recognise thoughts as agents
of actions, and much of their behaviour seems designed to alter or control
other individuals’ states of mind. They deceive each other in more ways
and in more contexts than monkeys, and experiments with captive subjects
have suggested that chimpanzees can learn to distinguish between ignorant
and knowledgeable trainers. More intriguingly, there is also evidence from
the Ivory Coast that chimpanzees may actively teach their infants how to
crack open palm nuts, suggesting that mothers recognise and attempt to rectify
their infants’ ignorance.

Even apes, however, seem to have difficulty attributing specific mental
states to others. Consider empathy, for example. When his mother Flo died,
the young male chimpanzee Flint exhibited many of the behavioural patterns
we associate with grief in humans. He avoided others, stopped eating and
spent many hours a day sitting in a hunched posture, rocking back and forth.
After some days, he died. It is clear from this and other descriptions of
death in chimpanzees that these animals experience grief and a sense of
loss when an individual close to them dies. Equally striking, however, is
the absence of sympathy among other chimpanzees: no chimpanzee has ever
been reported to have consoled a grieving companion. Although chimpanzees
have mental states and grieve at the loss of close friends, they do not
seem to recognise the same mental states in others. As a result, they are
unable to share another’s sorrow or show empathy towards it.

When we study the social behaviour of monkeys we are tempted to anthropomorphise
and treat them as if they were human. There are, however, many ways in which
a vervet’s view of the world is very different from our own. Monkeys see
the world as composed of things that act, not things that think and feel.
Like the primatologists who study them, vervet monkeys observe social interactions
and draw generalisations about the types of relationships that exist among
individuals. But there is no evidence that the monkeys have a ‘theory of
mind’ that allows them to recognise their own knowledge and attribute mental
states to others.

While vervet monkeys are acutely sensitive to other animals’ behaviour,
they know little about what causes them to do what they do. A monkey may
make use of abstract concepts and have motives, beliefs and desires, but
her mental states are not accessible to her: she does not know what she
knows.

Robert Seyfarth and Dorothy Cheney are at the Department of Psychology
of the University of Pennsylvania in Philadelphia.

Further reading: How Monkeys See the World, by Dorothy Cheney and Robert
Seyfarth, is published by the University of Chicago Press.

* * *

WORKING OUT WHAT A GRUNT MEANS

Like the players at Wimbledon, vervet monkeys frequently grunt to each
other during normal social interactions. The vervets’ grunts are harsh,
raspy signals that sound like a human clearing his throat with his mouth
open. As Thomas Struhsaker originally noted, grunts are given in at least
four distinct social circumstances. A female monkey may grunt as she approaches
a more dominant individual or a subordinate, as she watches another animal,
or as she herself initiates a group movement across an open plain, or when
she has just spotted the members of another group.

Even to an experienced human listener, there are no immediately obvious
differences between grunts, either from one context to another or between
individuals. Even when grunts are displayed on sound spectrograms, it is
not possible to detect any consistent differences in acoustic structure
from one context to the next. Although grunts are occasionally answered
by other group members, in most cases grunts evoke no behaviourial responses.
Changes in the direction of gaze, which are difficult to measure in the
wild, seem the only obvious response when one individual grunts to another.

From an observer’s perspective, watching monkeys grunt to each other
is very much like watching humans engaged in conversation without being
able to hear what they are saying. The creatures are saying something but
we have no idea what it is. Vervet monkey grunts are strikingly different
from their alarm calls: alarm calls given in response to different predators
are easily distinguished acoustically

The most obvious way of understanding what is happening is to adopt
the influential view of W. John Smith, professor of biology at the University
of Pennsylvania. He suggests that the information contained in an animal’s
vocalisation can be determined by both its acoustic properties and the context
in which it is given. According to Smith, animals have relatively small
repertoires of signals, each of which conveys a broad, general message.
A small repertoire of general signals can nevertheless elicit a variety
of different responses because of variation in the contexts in which calls
are given.

Applied to vervet grunts, this hypothesis predicts that vervets are
using a single vocalisation in a variety of different circumstances. The
grunt itself provides general information about the vocaliser’s identity,
location or subsequent behaviour. Variation in the responses evoked by different
grunts is accounted for by variation that the receiver perceives in the
context in which they are given. Alternatively, it may be that within what
seems to a human observer to be a single type of grunt there are a number
of different grunts, each conveying quite specific information that depends
more on a call’s acoustic properties than on the context in which it is
given.

To test these hypotheses, we designed the following set of experiments.
First, grunts from the same individual were tape-recorded in each of the
social contexts described above. Then, over a number of months, we played
each grunt to subjects from a concealed loudspeaker and filmed their responses.
For example, we might play Bokassa’s ‘grunt to a dominant’ to Duvalier on
one day and then, three or more days later, play Bokassa’s ‘grunt to another
²µ°ù´Ç³Ü±è’.

Throughout these trials we allowed social context to vary freely. Tests
were conducted, for instance, when there were dominant or subordinate animals
nearby, when the group was foraging or resting, or when animals were at
the centre or the edge of their range. We reasoned that if the grunts were
really one vocalisation whose meaning was largely determined by context,
subjects should show no consistent differences when responding to different
calls; responses should be determined by the contexts in which calls were
presented. On the other hand, if each of the grunts was different, and if
each carried specific information that was relatively independent of context,
we should find consistent differences in responses to each grunt type, regardless
of the varying circumstances in which it was played.

Eighteen animals were played the recording of a grunt that had originally
been given to a dominant animal; then a few days later, they were played
a grunt given to another group. Subjects responded in many different ways,
but two responses appeared consistently, and were consistently different
across the two grunt types. ‘Grunts to a dominant’ caused subjects to look
towards the loudspeaker, while ‘grunts to another group’ caused subjects
to look out, towards the horizon, in the direction the loudspeaker was pointing.
So ‘grunts to another group’ directed the listener’s attention away from
the speaker, and in the direction toward which, under normal conditions,
the vocaliser would have been facing.

By their behaviour, then, the monkeys seemed to be saying that although
their grunts sound more or less the same to us, they contain information.
In many cases, this information can include events external to the signalling
individual, such as the approach of another group or the movement of animals
into an open area.