Theory of mind in animals is the ability of nonhuman animals to attribute mental states to themselves and others
Theory of mind in animals is an extension to non-human animals of the philosophical and psychological concept of theory of mind (ToM), sometimes known as mentalisation or mind-reading. It involves an inquiry into whether animals have the ability to attribute mental states (such as intention, desires, pretending, knowledge) to themselves and others, including recognition that others have mental states that are different from their own.
To investigate this issue experimentally, researchers place animals in
situations where their resulting behavior can be interpreted as
supporting ToM or not.
The existence of theory of mind in animals is controversial. On
the one hand, one hypothesis proposes that some animals have complex
cognitive processes which allow them to attribute mental states to other
individuals, sometimes called "mind-reading". A second, more
parsimonious, hypothesis proposes that animals lack these skills and
that they depend instead on more simple learning processes such as associative learning; or in other words, they are simply behaviour-reading.
Several studies have been designed specifically to test whether
animals possess theory of mind by using interspecific or intraspecific
communication. Several taxa have been tested including primates, birds
and canines. Positive results have been found; however, these are often
qualified as showing only low-grade ToM, or rejected as not convincing
by other researchers.
History and development
Much of the early work on ToM in animals focused on the understanding chimpanzees have of human knowledge
The term "theory of mind" was originally proposed by Premack and Woodruff in 1978.
Early studies focused almost entirely on studying if chimpanzees could
understand the knowledge of humans. This approach turned out not to be
particularly fruitful and 20 years later, Heyes, reviewing all the
extant data, observed that there had been "no substantial progress" in
the subject area.
A 2000 paper approached the issue differently by examining competitive foraging behaviour between primates of the same species (conspecifics). This led to the rather limited conclusion that "chimpanzees know what conspecifics do and do not see".
Next, brain activity in higher primates was studied and as a result, a
2003 study of the human brain suggested that the a functioning ToM
system activated three major nodes, the medial prefrontal, superior
temporal sulcus, and inferior frontal: the medial prefrontal node
handles the mental state of the self, that the superior temporal sulcus
detects the behaviour of other animals and analyzes the goals and
outcomes of this behaviour, and the inferior frontal region maintains
representations of actions and goals.
In 2007, Penn and Povinelli wrote "there is still little
consensus on whether or not nonhuman animals understand anything about
unobservable mental states or even what it would mean for a non-verbal
animal to understand the concept of a 'mental state'." They went on
further to suggest that ToM was "any cognitive system, whether
theory-like or not, that predicts or explains the behaviour of another
agent by postulating that unobservable inner states particular to the
cognitive perspective of that agent causally modulate that agent's
behaviour".
In 2010, an article in Scientific American acknowledged that dogs are considerably better at using social direction cues (e.g. pointing by humans) than are chimpanzees.
In the same year, Towner wrote, "the issue may have evolved beyond
whether or not there is theory of mind in non-human primates to a more
sophisticated appreciation that the concept of mind has many facets and
some of these may exist in non-human primates while others may not."
Horowitz, working with dogs, agreed with this and suggested that her
recent results and previous findings called for the introduction of an
intermediate stage of ability, a rudimentary theory of mind, to describe
animals' performance.
In 2013, Whiten reviewed the literature and concluded that
regarding the question "Are chimpanzees truly mentalists, like we are?",
he stated he could not offer an affirmative or negative answer.
A similarly equivocal view was stated in 2014 by Brauer, who suggested
that many previous experiments on ToM could be explained by the animals
possessing other abilities. They went on further to make reference to
several authors who suggest it is pointless to ask a "yes or no"
question, rather, it makes more sense to ask which psychological states
animals understand and to what extent.
At the same time, it was suggested that a "minimal theory of mind" may
be "what enables those with limited cognitive resources or little
conceptual sophistication, such as infants, chimpanzees, scrub-jays and
human adults under load, to track others' perceptions, knowledge states
and beliefs."
In 2015, Cecilia Heyes,
Professor of Psychology at the University of Oxford, wrote about
research on ToM, "Since that time [2000], many enthusiasts have become
sceptics, empirical methods have become more limited, and it is no
longer clear what research on animal mindreading is trying to find" and
"However, after some 35 years of research on mindreading in animals,
there is still nothing resembling a consensus about whether any animal can ascribe any
mental state" (Heyes' emphasis). Heyes further suggested that "In
combination with the use of inanimate control stimuli, species that are
unlikely to be capable of mindreading, and the 'goggles method' [see
below], these approaches could restore both vigour and rigour to
research on animal mindreading."
Methods
Specific
categories of behaviour are sometimes used as evidence of animal ToM,
including imitation, self-recognition, social relationships, deception,
role-taking (empathy), perspective-taking, teaching and co-operation, however, this approach has been criticised.
Some researchers focus on animals' understanding of intention, gaze,
perspective, or knowledge, i.e. what another being has seen. Several
experimental methods have been developed which are widely used or
suggested as appropriate tests for nonhuman animals possessing ToM. Some
studies look at communication between individuals of the same species (intraspecific) whereas others investigate behaviour between individuals of different species (interspecific).
Knower-Guesser
The
Knower-Guesser method has been used in many studies relating to animal
ToM. Animals are tested in a two-stage procedure. At the beginning of
each trial in the first discrimination training stage, an animal is in a
room with two humans. One human, designated the "Guesser," leaves the
room, and the other, the "Knower," baits one of several containers. The
containers are screened so that the animal can see who does the baiting,
but not where the food has been placed. After baiting, the Guesser
returns to the room, the screen is removed, and each human points
directly at a container. The Knower points at the baited container, and
the Guesser at one of the other three, chosen at random. The animal is
allowed to search one container and to keep the food if it is found.
Competitive feeding paradigm
The
competitive feeding paradigm approach is considered by some as evidence
that animals have some understanding of the relationship between
"seeing" and "knowing".
At the beginning of each trial in the paradigm, a subordinate
animal (the individual thought to be doing the mind-reading) and a
dominant animal are kept on opposite sides of a test arena which
contains two visual barriers. In all trials, a researcher enters the
enclosure and places food on the subordinate's side of one of the visual
barriers (one baiting event), and in some trials the researcher
re-enters the enclosure several seconds later and moves the food to the
subordinate's side of the other visual barrier (second baiting event).
The door to the subordinate's cage is open during any baiting by the
researcher. The conditions vary according to whether the dominant's door
is open or closed during the baiting events, and therefore whether the
subordinate individual can see the dominant. After baiting, both of the
animals are released into the test arena, with the subordinate being
released several seconds before the dominant. If the animals possess
ToM, it is expected that subordinates are more likely to gain the food,
and more likely to approach the food under several circumstances: (1)
When the dominant's door is closed during trials with a single baiting
event; (2) when the dominant's door is open during a first baiting event
but closed during a second; (3) in single baiting event trials with the
dominant's door open, subordinates are more likely to get the food when
they compete at the end of the trial with a dominant individual who did
not see the baiting.
Goggles Method
In
one suggested protocol, chimpanzees are given first-hand experience of
wearing two mirrored visors. One of the visors is transparent whereas
the other is not. The visors themselves are of markedly different
colours or shapes. During the subsequent test session, the chimpanzees
are given the opportunity to use their species-typical begging behaviour
to request food from one of the two humans, one wearing the transparent
visor and the other wearing the opaque. If chimpanzees possess ToM, it
would be expected they would beg more often from the human wearing the
transparent visor.
False Belief Test
A
method used to test ToM in human children has been adapted for testing
non-human animals. The basis of the test is to track the gaze of the
animal. One human hides an object in view of a second human who then
leaves the room. The object is then removed. The second human returns
whereupon they will mistakenly look for the object where they last saw
it. If the animal stares first and longest at the location where the
human last saw the object, this suggests they expect him to believe it
is still hidden in that place.
In nonhuman primates
Many ToM studies have used nonhuman primates (NHPs). One study that examined the understanding of intention in orangutans (Pongo pygmaeus), chimpanzees (Pan troglodytes) and children showed that all three species understood the difference between accidental and intentional acts.
Chimpanzees
There is controversy over the interpretation of evidence purporting to show ToM in chimpanzees.
William Field and Sue Savage-Rumbaugh have no doubt that bonobos have evolved ToM and cite their communications with a captive bonobo (Pan paniscus), Kanzi, as evidence.
However, empirical studies show that chimpanzees are unable to follow a human's gaze, and are unable to use other human-eye information. Attempts to use the "Goggles Method" (see above) on highly human-enculturated chimpanzees failed to demonstrate they possess ToM.
In contrast, chimpanzees use the gaze of other chimpanzees to gain information about whether food is accessible.
Subordinate chimpanzees are able to use the knowledge state of
dominant chimpanzees to determine which container has hidden food.
If chimpanzees can see two opaque boards on a table and are
expecting to find food, they do not choose a board lying flat because if
food was under there, it would not be lying flat. Rather, they choose a
slanted board, presumably inferring that food underneath is causing the
slant. Chimpanzees appear able to know that other chimpanzees in the
same situation make a similar inference. In a foraging game, when their
competitor had chosen before them, chimpanzees avoided the slanted board
on the assumption that the competitor had already chosen it.
In a similar study, chimps were provided with a preference box with
two compartments, one containing a picture of food, the other containing
a picture of nothing (the pictures had no causal relation to the
contents). In a foraging competition game, chimpanzees avoided the
chamber with the picture of food when their competitor had chosen one of
the chambers before them. The authors suggested this was presumably on
the assumption that the competitor shared their own preference for it
and had already chosen it.
One study tested another sensory mode of ToM. In a food
competition, a human sat inside a booth with one piece of food to their
left and one to their right. The food could be withdrawn from the
competing chimpanzee's reach when necessary. In the first experiment,
the chimpanzee could approach either side of the booth unseen by the
human, but then had to reach through either a transparent or opaque tube
to get the food. In a second experiment, both were transparent and the
human was looking away, but one of the tubes made a loud rattle when it
was opened. Chimpanzees reached through the opaque tube in the first
experiment and the silent tube in the second. The chimpanzees
successfully concealed their food-stealing from their human competitor
in both cases.
Chimpanzees have passed the False Belief Test (see above)
involving anticipating the gaze of humans when objects have been
removed.
Other primates
_in_Kinnarsani_WS,_AP_W_IMG_5792.jpg)
Rhesus
macaques selectively steal grapes from humans who are incapable of
seeing the grape compared to humans who can see the grape.
In one approach testing monkeys, rhesus macaques (Macaca mulatta) are able to "steal" a contested grape
from one of two human competitors. In six experiments, the macaques
selectively stole the grape from a human who was incapable of seeing the
grape, rather than from the human who was visually aware. The authors
suggest that rhesus macaques possess an essential component of ToM: the
ability to deduce what others perceive on the basis of where they are
looking.
Similarly, free ranging rhesus macaques preferentially choose to steal
food items from locations where they can be less easily observed by
humans, or where they will make less noise.
A comparative psychology
approach tested six species of captive NHPs (three species of great
apes: orangutans, gorillas, chimpanzees, and three species of old-world
monkeys: lion-tailed macaques (Macaca silenus), rhesus macaques and collared mangabeys (Cercocebus torquatus))
in a "hide and seek" game in which the NHPs played against a human
opponent. In each trial, the NHP has to infer where food has been hidden
(either in their right or left hand) by the human opponent. In general,
the NHPs failed the test (whereas humans did not), but surprisingly,
performances between the NHP species did not reveal any inter-species
differences. The authors also reported that at least one individual of
each of the species showed (weak) evidence of ToM.
In a multi-species study, it was shown that chimpanzees, bonobos and orangutans passed the False Belief Test (see above).
In 2009, a summary of the ToM research, particularly emphasising an extensive comparison of humans, chimpanzees and orang-utans, concluded that great apes do not exhibit understanding of human referential intentions expressed in communicative gestures, such as pointing.
In birds
Parrots
Grey parrots (Psittacus erithacus) have demonstrated high levels of intelligence. Irene Pepperberg did experiments with these and her most accomplished parrot, Alex, demonstrated behaviour which seemed to manipulate the trainer, possibly indicating theory of mind.
Ravens
Ravens adjust their caching behaviour according to whether they have been watched and who was watching them.
Ravens are members of the corvidae family and are widely regarded as having complex cognitive abilities.
Food-storing ravens cache
(hoard) their food and pilfer (steal) from other ravens' caches. They
protect their caches from being pilfered by conspecifics using
aggression, dominance and re-caching. Potential pilferers rarely
approach caches until the storing birds have left the cache vicinity.
When storers are experimentally prevented from leaving the vicinity of
the cache, pilferers first search at places other than the cache sites.
When ravens (Corvus corax) witness a conspecific making caches,
to pilfer those caches they (1) delay approaching the cache only when in
the presence of the storer, and (2) quickly engage in searching away
from the caches when together with dominant storers. These behaviours
raise the possibility that ravens are capable of withholding their
intentions, and also providing false information to avoid provoking the
storer's aggression to protect its cache. Ravens adjust their pilfering
behaviour according to when the storers are likely to defend the
caches. This supports the suggestion that they are deceptively
manipulating the other's behaviour.
Other studies indicate that ravens recall who was watching them during
caching, but also know the effects of visual barriers on what
competitors can and can not see, and how this affects their pilfering.
Ravens have been tested for their understanding of "seeing" as a
mental state in other ravens. It appears they take into account the
visual access of other ravens, even when they cannot see the other
raven.
In one study, ravens were tested in two rooms separated by a
wooden wall. The wall had two functional windows that could be closed
with covers; each cover had a peephole drilled into it. In the next
familiarization step, the ravens are trained to use a peephole to
observe and pilfer human-made caches in the adjacent room. Under test
conditions, there was no other raven present in the adjacent room,
however, a hidden loudspeaker played a series of sounds recorded from a
competitor raven. The storing raven generalized from their own
experience when using the peephole to pilfer the human-made caches and
predicted that the audible (raven) competitors could potentially see
their caches through the peep-hole and took appropriate action, i.e. the
storing ravens finished their caches more quickly and they returned to
improve their caches less often. The researchers pointed out that this
represented "seeing" in a way that cannot be reduced to the tracking of
gaze cues – a criticism leveled at many other studies of ToM.
The researchers further suggested that their findings could be
considered in terms of the "minimal" (as opposed to "full-blown") ToM
recently suggested.
Using the Knower-Guesser approach, ravens observing a human
hiding food are capable of predicting the behaviour of bystander ravens
that had been visible at both, none or just one of two baiting events.
The visual field of the competitors was manipulated independently of the
view of the test-raven. The findings indicate that ravens not only
remember whom they have seen at caching but they also take into account
that the other raven's view was blocked.
Scrub jays
Western scrub jays may show evidence of possessing theory of mind
Scrub jays are also corvids. Western scrub jays (Aphelocoma californica)
both cache food and pilfer other scrub jays' caches. They use a range
of tactics to minimise the possibility that their own caches will be
pilfered. One of these tactics is to remember which individual scrub jay
watched them during particular caching events and adjust their
re-caching behaviour accordingly.
One study with particularly interesting results found that only scrub
jays which had themselves pilfered would re-cache when they had been
observed making the initial cache.
This has been interpreted as the re-caching bird projecting its own
experiences of pilfering intent onto those of another potential
pilferer, and taking appropriate action.
Another tactic used by scrub jays is if they are observed caching,
they re-cache their food when they are subsequently in private. In a
computer modeling study using "virtual birds", it was suggested that
re-caching is not motivated by a deliberate effort to protect specific
caches from pilfering, but by a general motivation to simply cache more.
This motivation is brought on by stress, which is affected by the
presence and dominance of onlookers, and by unsuccessful recovery
attempts.
In dogs
Dogs can use the pointing behaviour of humans to determine the location of food.
Domestic dogs (Canis familiaris)
show an impressive ability to use the behaviour of humans to find food
and toys using behaviours such as pointing and gazing. The performance
of dogs in these studies is superior to that of NHPs, however, some have stated categorically that dogs do not possess a human-like ToM.
The Guesser-Knower approach has been used with ToM studies in
dogs. In one study, each of two toys was placed on the dog's side of
two barriers, one opaque and one transparent. In experimental
conditions, a human sat on the opposite side of the barriers, such that
they could see only the toy behind the transparent barrier. The human
then told the dog to 'Fetch' without indicating either toy in any way.
In a control, the human sat on the opposite side but with their back
turned so that they could see neither toy. In a second control, the
human sat on the same side as the dog such that they could see both
toys. When the toys were differentiable, dogs approached the toy behind
the transparent barrier in experimental as compared to "back-turned" and
"same-side" condition. Dogs did not differentiate between the two
control conditions. The authors suggested that, even in the absence of
overt behavioural cues, dogs are sensitive to others' visual access,
even if that differs from their own.
Similarly, dogs preferentially use the behaviour of the human Knower
to indicate the location of food. This is unrelated to the sex or age
of the dog. In another study, 14 of 15 dogs preferred the location
indicated by the Knower on the first trial, whereas chimpanzees require
approximately 100 trials to reliably exhibit the preference.
Human infants (10 months old) continue to search for hidden
objects at their initial hiding place, even after observing them being
hidden at another location. This perseverance of searching errors is at
least partly contributed to by behavioural cues from the experimenter.
Domestic dogs also commit more search errors in communicative trials
than in non-communicative or non-social hiding trials. However,
human-encultured wolves (Canis lupus)
do not show this context-dependent perseverance in searching. This
common sensitivity to human communication behaviour may arise from convergent evolution.
Dogs which have been forbidden to take food are more likely to
steal the food if a human observer has their back turned or eyes closed
than when the human is looking at them. Dogs are also more likely to beg
for food from an observer whose eyes are visible compared to an
observer whose eyes are covered by a blindfold.
In a study of the way that dogs interact, play signals were sent
almost exclusively to forward-facing partners. In contrast,
attention-getting behaviors were used most often when the other dog was
facing away, and before signaling an interest to play. Furthermore, the
type of attention-getting behaviour matched the inattentiveness of the
playmate. Stronger attention-getting behaviours were used when a
playmate was looking away or distracted, less forceful ones when the
partner was facing forward or laterally,
In pigs
An experiment at the University of Bristol
found that one out of ten pigs was possibly able to understand what
other pigs can see. That pig observed another pig which had view of a
maze in which food was being hidden, and trailed that pig through the
maze to the food. The other pigs involved in the experiment did not.
In goats
A
2006 study found that goats exhibited intricate social behaviours
indicative of high-level cognitive processes, particularly in
competitive situations. The study included an experiment in which a
subordinate animal was allowed to choose between food that a dominant
animal could also see and food that it could not; those who were subject
to aggressive behaviour selected the food that the dominant animal
could not see, suggesting that they are able to perceive a threat based
on being within the dominant animal's view – in other words, visual
perspective taking.
