Content uploaded by Alex H Taylor
Author content
All content in this area was uploaded by Alex H Taylor on Nov 29, 2019
Content may be subject to copyright.
Focus Article
Corvid cognition
Alex H. Taylor∗
Numerous myths and legends across the world have suggested that corvids are
intelligent. However, it is only in the last two decades that their cognition has
become the subject of serious scientific investigation. Here I review what we
currently know about the temporal, social, and physical cognition of this group. I
argue that, while the work to date establishes corvids as one of the most intelligent
groups of animals on the planet, the real scientific potential of the Corvidae has
yet to be realized. However, a novel ‘signature-testing’ experimental approach
is required if we want to unlock this group’s promise and gain insights into the
evolution of human and animal minds. ©2014 John Wiley & Sons, Ltd.
How to cite this article:
WIREs Cogn Sci 2014. doi: 10.1002/wcs.1286
INTRODUCTION
Many old stories contain a kernel of truth.
Some do not. One of the most interesting
developments in animal cognition over the last
20 years is the claim that the intelligence attributed
to corvids in myth and folklore has at least some basis
in reality. Aesop suggested in his fable ‘The Crow and
the Pitcher’ that these birds are highly inventive.1In
Native American mythology the raven is seen as the
creator of light and as a trickster2–4while in Western
Bering folklore the raven is seen as ‘the transformer of
the world’, who teaches humans the skills they need
to survive in the world.5Thus the raven caws to make
people speak and invents the fire-drill by twisting his
forefinger (the drill) into the base of his other foot5
(the drill base). Nordic mythology tells the story of
two ravens Huginn (thought) and Muninn (memory),
who are sent out each dawn by the God Odin to fly
all over the world.6They return at dusk to whisper
to Odin what they have seen, so ensuring Odin is
wise.7Similarly, Celtic art depicts crows speaking
into the ears of men,8as in Celtic mythology crows
were considered to have oracular powers.8,9 These
myths and stories, then, suggest that corvids are not
only intelligent but, through their intelligence, have
something to teach us. Here I review recent advances
∗Correspondence to: alexander.taylor@auckland.ac.nz
School of Psychology, University of Auckland, Auckland, New
Zealand
Conflict of interest: The authors have declared no conflicts of
interest for this article.
in our scientific knowledge of corvid cognition to
examine whether this is really the case.
WHAT IS A CORVID?
The family Corvidae consists of the crows, magpies,
jays, and their allies. In total this family contains 117
species, with the crows (genus Corvus) comprising
around 40 species. This family is part of the ‘core
Corvoidea’, a group of some 750 oscine passerine
species that originated in Papua New Guinea around
50 million years ago (mya) and then radiated from
there across the globe.10 The crows themselves (genus
Corvus) arose around 11 mya (range 9–15 mya) in
the Palaearctic.11
WHAT MAKES THE CORVIDAE
INTERESTING?
The Corvidae have two particularly intriguing
characteristics. The first is the frequency with which
this group produces complex behaviors. Overall,
the species within the Corvidae are responsible
for a higher number of foraging innovations and
instances of tool use than any other bird group,12–14
though psittcaids (parrots) come to a close second.
Individually, certain species exhibit various complex
behaviors (Figure 1). For example, tool-making New
Caledonian crows are the only species, beside humans,
chimpanzees and orang-utans, to craft their tools into
a particular three-dimensional form using a sequence
of behaviors,15,16 and the only species except humans
©2014 John Wiley & Sons, Ltd.
Focus Article wires.wiley.com/cogsci
FIGURE 1|Examples of complex behaviors in the Corvidae. From left to right: a Clark’s nutcracker caching (Photo credit Dave McShaffrey), a
New Caledonian crow with a hook tool (Photo credit Gavin Hunt), and rooks bill-twining after a fight (Photo credit Julia Leijola).
that manufacture hook tools in the wild.17 Clark’s
nutcrackers bury up to 33,000 pine seeds in the ground
before winter and recover them months later.18 Both
rooks and ravens exhibit similar conflict management
behaviors to those seen in primates: ravens make up
with each other after fights19 (reconciliation), while
rooks will bill-twine with their partners after fighting
another rook20 (third-party affiliation).
The second key feature of the corvids is their
degree of brain encephalization (ratio of brain size
to body size). Remarkably, corvids have an encephal-
ization ratio similar to that of chimpanzees,21,22 and
larger than any other bird group, save the psittcaids23
(parrots). The most enlarged areas of the corvid
brain are the nidopallium and mesopallium,24–26
which are thought to be analogous to the areas of
mammalian brain used in complex cognitive processes
(the prefrontal cortex). While there is ongoing debate
on how closely encephalization correlates with
intelligence,27–29 the degree of encephalization in
this group is particularly striking, given the energetic
demands imposed by flight and by brain tissue. Flight
exerts a strong selective pressure to maintain low
body weight and keep energetic costs down.30,31
Despite this selective pressure, and the incredibly high
energetic costs of maintaining brain tissue,32 corvids
have evolved relatively large brains. For this to have
occurred these brains have to be (literally) pulling
their weight and providing large adaptive advantages
in the environment.
SO WHAT DO WE KNOW ABOUT
CORVID COGNITION?
The Corvidae have bigger relative brain sizes and
produce more complex behaviors than any other
bird group. The increased brain size suggests this
group has the potential to think in more sophisticated
ways than other birds. The high rate of behavioral
innovation suggests the same: if all birds thought
as the corvids do, then one would expect them to
produce the same amount of complex behaviors. Yet
they do not.12,14 However, only carefully controlled
behavioral experiments can show us what cognitive
mechanisms members of this group actually possess.
Here I highlight the key findings in corvid cognition
to date.
Imagining the Past and Future
Corvids use sophisticated cognition when remember-
ing the past and planning for the future. Clayton and
Dickinson33 found that when allowed to cache both
peanuts and worms, scrub jays preferred to recover
their favorite food, the worms, if only a short time has
passed (4 h). However, when a longer time had passed
(124 h) the jays preferentially recovered the nuts they
had stored, as worms degrade after 124 h and so are
not worth eating. In contrast, jays given experience
of non-degrading ‘magic’ worms, through the replace-
ment of degraded worms with living ones by the
experimenter, still preferred to recover worms after
124 h. Thus the jays used their experience of how
food degrades to guide their cache-recovery behav-
iors. This work suggests that jays encode ‘what’,
‘where’, and ‘when’ (WWW) during caching. That
is, they remember whether they have cached peanuts
or worms, where they cached the object, and whether
4 or 124 h has passed, though there is continued
debate as to how the ‘when’ aspect of their memory is
encoded.34,35 The presence of WWW memory in jays
suggests that their behavior can approximate episodic
memory in humans, thus this corvid species has been
described as having episodic-like memory.33 The crit-
ical difference is that in humans episodic memory is
based on a conscious re-experiencing of the past36,37
(autonoetic consciousness). Given limitations in our
©2014 John Wiley & Sons, Ltd.
WIREs Cognitive Science Corvid cognition
Eats maintenance diet
Prefed either Food A or B
Caches 30:70 mix of Food A and B at Location 1
Caches 50:50 mix of Food A and B at Location 2Caches 50:50 mix of Food A and B at Location 2
Prefed Food B so desires Food A
Has 50:50 mix of Food A and B to retrieve at
Location 2, but wants to eat more of A due
to prefeeding
Has 50:50 mix of Food A and B to retrieve at
Location 1, but wants to eat more of B due to
prefeeding
Prefed Food A so desires Food B
Caches 70:30 mix of Food A and B at Location 2
Trial 1
Caching
Trial 1 Maintenance prefeeding
Trial 1
Preference
Prefeeding
Trial 1 Food retrieval
Trial 2 Preference
Prefeeding
Trial 2
Caching 1
ABA
2
2
2
AB
1
AB
1
B
FIGURE 2|Experimental outline of planning for two future desire states study in Eurasian jays. Note on any one trial a jay would receive either
the conditions on the left or the right column of the diagram.
understanding of consciousness, we have little idea
whether jays are also capable of self-projection into
the past.38
While several studies have found intriguing
evidence for future planning in corvids,39,40 there are
alternative explanations for these results.41 Perhaps
the most convincing evidence comes from a recent
study on Eurasian jays by Cheke and Clayton.42 In
humans, the more you eat of one food, the less you
want of it compared to other foods. Similarly, jays
show such specific satiety: after eating a particular
food they then want less of it, but are happy to eat
new food types. After being fed their normal food,
these jays were allowed to cache two food types (A
and B) at two different locations (locations 1 and 2).
Initially the jays stored an equal amount of each food
type at each location. After 3 h they were then allowed
to eat as much as they wanted of one of the two food
types (specific satiation), and then were allowed to
recover their caches from one of the locations they
had cached in. Thus at this point the jays only wanted
to recover caches of the food they had not been just
fed, though they had previously filled the cache site
with both types of food. A day later the jays were fed
the other food type to satiation and then were allowed
to retrieve food from the second food location (where
they had cached food 27 h before). Again the jays only
wanted to eat the food type they had not just been fed,
though they had already cached both food types. On
the second trial of the experiment the jays changed
their caching behavior to reflect the future feeding
events they had experienced on trial 1 (Figure 2). Even
though they were fed to satiety on one of the food
types before caching began, when caching they hid
more of the food type in each tray that they knew they
would not be fed on later. Thus they remembered that
in 3 h they would be fed one food type before being
allowed to retrieve their caches and in 27 h they would
be fed the other before being allowed to retrieve their
cache. This work indicates that these jays can actually
©2014 John Wiley & Sons, Ltd.
Focus Article wires.wiley.com/cogsci
plan for two different desires at two different time
points in the future while ignoring their own current
desires. As with work on episodic memory it is unclear
whether jays consciously project themselves into the
future in order to plan for different desires at different
times.38,42
Social Cognition
Several ancient Greek writers suggested that jackdaws
could be trapped using only a dish of oil.43 The
trap was effective because these birds were highly
social and so became sufficiently absorbed by the
reflected image that they would fall into the dish.
What is not clear from this anecdote is what aspect
of the jackdaws’ social lives led them to fall in. Did
each jackdaw recognize itself as a distinct individual
separate from the other jackdaws in its life and thus
fall into the oil in effort to get a closer look at itself?
Or did they simply see another jackdaw in the oil
they wanted to interact with? Work by Prior et al. on
mirror recognition in magpies suggests that either of
these explanations could have driven the jackdaws’
fall.44 When initially confronted by a mirror, magpies
performed social behaviors such as aggressive displays
and jumping toward the mirror. However, over time
several of the magpies changed their behavior and
began to perform behaviors such as looking behind
the mirror. The magpies were then given the mark test.
Here, either a black dot or a colored dot was placed
underneath their throat. While neither dot could be
seen by the animal without a mirror, only the colored
dot could be seen in the mirror, due to the black dot
blending into the magpies’ black feathers. When the
colored dot was placed on the magpies, throats and a
mirror was present, two of the six individuals tested
performed a high number of self-directed behaviors
toward it. For example they attempted to scratch the
dot off with their leg, rather than toward the reflection
of the spot in the mirror. When a black dot was
placed on the magpies’ throats, or the mirror was not
present, these behaviors were performed significantly
less often. Thus, just like great apes,45 elephants,46
and dolphins,47 a proportion of the magpies tested by
using mirror reflections to notice changes to their own
bodies. Recent work suggests that passing the mirror
test indicates an animal can differentiate between itself
and others and has a restricted sense of self-awareness,
based on an ability to generate and compare multiple
mental models of its own physical appearance.48 This
raises the possibility that the jackdaws in Greek legend
fell into the oil because they became so absorbed by
looking at the reflection of their own body, though it
does seem more plausible they jumped in as part of
an aggressive social display. What is still a complete
mystery is whether corvids have a concept of self as
humans do.
Corvids go beyond discriminating between their
bodies and other conspecifics. Recent work by Boeckle
and Buygnar has shown that ravens remember for
over 3 years not only whether particular individuals
are familiar or unfamiliar, but also the valence of
their relationship with these individuals49 (whether
they get on with them or not). After housing a
group of juvenile ravens together as a non-breeding
group and monitoring their social behavior, the group
was spilt into breeding pairs. After 3 years, calls of
various past group members and of unknown birds
were played back to the ravens in each pair. Birds
changed the calls they made to these recordings
based not only on whether individuals were familiar
or not, but also based on their past relationships
with familiar individuals (whether they affiliated
with these individuals or not as juveniles in the
non-breeding group). This work demonstrates that
ravens have long-term memory for social interactions,
though it is unclear whether the ravens remembered
experiences with specific individuals, or remembered
categories of individual such as affiliates and non-
affiliates. Other work by Marzluff et al. has shown
that corvids’ memory for social interactions extend
beyond within-species interactions. American crows
caught and released by humans wearing a unique
mask scolded any human wearing the mask for up
to at least 2.7 years after the initial catching event.50
Thus, this corvid species is capable of discriminating
between humans using facial characteristics and so can
remember masks, and likely faces, over long periods
of time.
Corvids seem to be able to use their own
experiences in the world in order to predict the
future behavior of their conspecifics. One of the key
problems faced by caching scrub jays is theft. If a com-
petitor can observe where a jay has hidden his food, he
can get an easy meal simply by digging up the cache.
It therefore pays to recache: to dig up hidden food and
hide it again, particularly if the first hiding event was
observed. Emery and Clayton manipulated the expe-
rience of two groups of hand-raised scrub jays.51 One
group had previous experience of stealing other birds’
caches. The second group had no experience of being
a thief. Both groups were then given the opportunity
to cache when alone and when observed by a conspe-
cific. Only the jays that had experience of being a thief
recached their food when the initial hiding event was
observed. Thus, not only can scrub jays remember
whether they were being watched during caching,
but they may also engage in experience projection,
©2014 John Wiley & Sons, Ltd.
WIREs Cognitive Science Corvid cognition
where they use their experience of being a thief to
create cache protection strategies, possibly by using
their experience to mentally simulate the viewpoint
of a potential thief. This work raises the possibility
that corvids may have some type of theory of mind:
the ability to attribute mental states, such as beliefs,
knowledge, perspectives, and desires, to conspecifics.
Several studies have focused more explicitly on
whether corvid species have theory of mind. Scrub
jays remember who watched them during caching
andusethisinformationtodecide,whenbeing
subsequently watched by either knowledgeable or
uninformed competitors, whether to rehide their food
or leave it alone.52 As this behavior does not depend
on the behavior of the competitor, it raises the
possibility that these jays can attribute knowledge
and ignorance to conspecifics. However, alternate
explanations exist.53 While a recent model based
on stress driving recaching54 has been disproven,55
another possibility is that, on seeing a particular
individual watch a caching event, a jay could form
a negative association between this individual and
the location it is caching in, rather than attributing
‘knowledge’ of the item’s location to the individual.
The jay would then recache often in this location when
watched by individual it has ascribed this ‘negative’
tag to. Thus the content associated with a particular
individual is unclear: jays need not think as humans
do and so decide that a particular individual knows
where the food is.52,53,56,57 A similar critique applies to
work on ravens. This corvid will rush to steal hidden
food only when it is possible for a visible competitor
to also see the food being hidden.58 While this work
suggests that ravens not only remember who was
watching, but also whether this individual’s viewpoint
was blocked or not, it does not require the ravens to
have attributed knowledge to a particular competitor.
Instead, simply observing that a competitor has a line
of sight with food currently being hidden could have
been associated in the past with losing food, leading
to the raven learning to rush to steal the food before
the competitor.
Recent work by Ostoji´
c et al. suggests that
Eurasian jays might attribute desire to conspecifics.59
Male Eurasian jays choose food to give to their mates
during courtship. It would therefore be advantageous
if the males could predict what the females actually
wanted to eat, given that it would reduce the cost
of foraging for unwanted food items and potentially
make them a more attractive mate. Just as in Cheke
and Clayton,42 this experiment was based on specific
satiety: once jays have eaten a large quantity of one
food type, such as meal worm larvae, they want less
of it, but will still eat a new food type, such a wax
worms. Males in one set of trials were allowed to see
their mates eating a large portion of one food type
and then were given the choice of what to feed their
mate next (seen condition). In a second set of trials
the males did not see that the female had eaten a large
portion of one food type, but could still potentially use
her body language after the eating event to judge what
she wanted (unseen condition). In the seen condition,
males did not give females the food they had just
been eating, while in the unseen condition the males
gave both food types. This indicates it was the males’
observation of the females eating a food type that
drove his decision about what to share with his
mate, rather than the females’ subsequent behavior.
Importantly, the males were able to dissociate their
desires from those of the females. After watching the
females eat a large portion of one food type, the
males did not then choose to eat the other food type
themselves. Thus, observing a female eat lots of wax
worms led to the male giving mealworm larvae to the
female, but continuing to have a healthy appetite for
wax worms himself (Figure 3).
These results suggest that male Eurasian jays
can attribute a desire for a particular food type to a
female jay, given that the female’s behavior alone
did not guide giving and that males differentiate
between the food they want and food their mates
want. Their behavior may also be further evidence of
experience projection, as males may have used their
own experiences with eating lots of one particular
food type to predict that females would not want to
eat more of a food they had already gorged themselves
on. However, further work is required to substantiate
these conclusions. One key behavioral criteria put
forward for the explicit attribution of mental states
to others is that animals should be able to judge
the similarity between perceptually disparate behavior
patterns linked to an unobservable mental state.60 This
criterion has been championed by leading skeptics in
the field.53 In Ostoji´
c et al.’s experiment the sight
of females eating lots of one food type drove the
males’ provisioning behavior. If these jays really do
attribute desires to each other one would expect that
a perceptually disparate behavior pattern would also
drive provisioning behavior. Thus a male should react
the same if he sees his mate enter a room with lots
of worms and then later leave an empty room. Such
flexibility in the inputs used for desire attribution
would be highly adaptive, as it would be highly costly
for provisioning males to be constantly monitoring
what their mate is eating. If a male can remember
what food he has given his mate, and what is left
when he returns from a foraging trip, he can make
inferences about what the female has eaten, and thus
©2014 John Wiley & Sons, Ltd.
Focus Article wires.wiley.com/cogsci
Male watches female eat a large
amount of one food type
Male watches female after she has
eaten a large amount of one food
type
Male gives female both food types
Male still desires the food the has seen
the female eating
Male watches female eat a large
amount of one food type
Control 2
Control 1
Experiment
Male gives female the food type
different from that she has just eaten
FIGURE 3|Experimental outline of desire attribution study in Eurasian jays.
her desire state, without keeping her within his line of
sight.
Tool Use and Causal Understanding
A third line of work on corvid cognition has examined
what this group of birds think about tools and the
causality underpinning events in the world. Much
of this work has focused on New Caledonian crows,
because of their tool-making abilities in the wild15,17,61
and rooks, which produce impressive tool behaviors
in captivity.62 New Caledonian crows can choose
tools of the right length for the task in hand,63 can
make tools of the right diameter for a problem,64
and will risk their tools, rather than their beaks
when faced with hazardous objects such as a model
snake65,66 (context-dependent tool use). Rooks have
been shown to choose between functional and non-
functional tools, to spontaneously use sticks as tools
and to spontaneously modify sticks so they can be used
as tools.62 Both species have also produced meta-tool
behaviors62,67–69 where they use tools to gain access
to other tools which can be used to get food. One of
the most impressive behaviors produced by both these
corvid species is the bending of man-made material
into hooks.62,70 After using a wooden hook to pull
a bucket by its handle from a tube, Betty, a New
Caledonian crow, subsequently bent a straight piece
of wire into a hook shape in order to do the same.70
Rooks, given similar experience using a wooden hook
to pull a bucket from a tube, also then bent wire to
make a functional hook.62
More recent work has shown that New
Caledonian crows will use stones as tools after
limited experience pushing a platform with their
beaks71 and that rooks, crows, and Eurasian jays72–74
can, like children over the age of 7,75 learn the
functional properties of stones and stone-like tools
when dropping these objects into a water-filled tube.
Finally, both rooks and New Caledonian crows have
been able to solve the trap-tube problem, where an
animal has to pull food from within a horizontal
tube while avoiding a trap. One rook, Guillem, was
able to solve various transfer tasks that required
him to switch between treating the same cues as
negative or positive.76 Three New Caledonian crows
solved several transfer tasks, including the trap-table
problem, a problem with an identical causal structure
to the trap-tube task, in that an animal must avoid
pulling food into a hole, but with very different
perceptual elements.77,78 In contrast, chimpanzees
fail to transfer from the trap-tube to the trap-table
problem when required to use tools to do so.79
However, the cognitive mechanisms behind the
impressive performance outlined above have not been
pinpointed. While these studies show that the crows’
behavior is not guided by simple learning mechanisms
alone, there are also instances where their impressive
performances include odd (but informative) mistakes
©2014 John Wiley & Sons, Ltd.
WIREs Cognitive Science Corvid cognition
that humans would not make. The ability to generalize
tool use from out-of-reach food objects to out-of-
reach tools when solving metatool problems62,67,
and to generalize from the trap-tube to trap-table77,
shows that these birds are not bound by the law
of stimulus generalization80, but can instead form
abstract categories not tied to specific perceptual
features.67,77 Similarly, the ability to invent novel
wire tools,62,70 novel stone dropping behaviors,71
and novel metatool sequences62,67,68 shows that these
crows are not bound by learning phenomenon such as
resurgence, chaining, and conditional reinforcement
when creating novel behaviors.67 Finally, both the
trap-tube and stone-dropping studies73,74,76,77 show
that several corvid species have the ability to quickly
learn the functional properties of objects, but not
to learn arbitrary properties rewarded with the
same consistency. This demonstrates that the crows’
learning is based on more than co-variation levels
between objects in the environment and outcomes.81
However, these corvids clearly make errors in
their problem-solving performances that it seems
unlikely humans would make. After inventing stone-
tool use, one New Caledonian crow dropped a feather
onto a platform in an attempt to collapse it, despite the
feather’s weight being insufficient for the platform to
be collapsed in this way.71 When wire bending, New
Caledonian crows and rooks often make use of the
non-functional end of a tool first, despite this course
of action being highly unlikely to succeed.62,70,82 New
Caledonian crows make a variety of errors when
solving metatool problems,66,68,69 which brings into
question whether they create hierarchically structured
plans to solve these tasks. Finally, while New
Caledonian crows can transfer from the trap-tube to
the trap-table, they failed to solve a trap-tube which
contained two holes, only one of which has a base
and so worked as a functional trap. Thus these crows
cannot transfer from using the holes in a trap-tube
to using the base of a trap as the critical feature that
predicts success.77
This paradox in corvids’ understanding of
causality is neatly exemplified by two recent studies
on New Caledonian crows. These corvids can, in a
matter of seconds, spontaneously solve the string-
pulling paradigm, where a string must be pulled and
then stepped on repeatedly in order to bring hanging
food within reach.83,84 However, studies by Taylor
et al. have found that when New Caledonian crows
are not given feedback on the effect of their actions on
the hanging food85 or feedback is interrupted,84 they
stop producing coherent performances, even if they
have done so in the past.84 Thus, when faced with
novel string-pulling problems involving man-made
objects, corvids seem to use the perceptual-motor
feedback created from observing the effects of their
actions on the world to drive their performance, rather
than mentally imagining a plan of action toward the
string and then executing it.85,86 In contrast, there is
evidence of mental simulation by this species when
it thinks of hidden causal agents. The ability to
mentally simulate the potential causal interactions
of a hidden animate object is highly adaptive. For
example, understanding that the rustling of the leaves
in a forest canopy indicates the presence of a predator
would allow an animal to make evasion decisions
before the predator is seen. This ability is also a
precursor to a number of highly complex human
cognitive abilities, including theory of mind, scientific
reasoning, and religious reasoning.87–90 Taylor et al.
recently examined whether New Caledonian crows
could infer that the backward and forward movement
of a stick from a hide was caused by a human they
had seen enter the hide.91 The crows seemed able
to ‘join the dots’ and infer that the hidden human
was the cause of the stick’s movement. Thus when
the human left the hide, the crows reasoned that
the stick could not move again and so it was safe
to forage in close proximity to the hide. Though
associative accounts for the crows’ behavior have been
suggested92,93 they do not seem able to explain the
results in their entirety.94,95 These results are therefore
in stark contrast to those found with string pulling
paradigms (Figure 4).
SO HOW DO CORVIDS REALLY
THINK?
The studies outlined above suggest that corvid species
use sophisticated cognition when thinking about
temporal events, conspecifics, tools, and the causality
of the world. The field of corvid cognition has,
therefore, made substantial progress in demonstrating
that the two hallmarks of intelligence seen in this
group, (large relative brain size and behavioral com-
plexity), are because of this group actually thinking in
sophisticated ways. In fact, in several of the cognitive
spheres discussed here, the evidence for intelligence
rivals that found for primates,23,42,56,59,77,96 which
is highly impressive, particularly given that primate
cognition has been studied for far longer. Because of
this, the corvids are now established as one of the
most intelligent groups of animals on the planet.
However, a number of key questions remain
unanswered. Certain corvid species seem to remember
the past and plan for the future in a flexible way that
goes beyond the Bischof-Kohler hypothesis, which
states that an animal cannot plan for a future desire
©2014 John Wiley & Sons, Ltd.
Focus Article wires.wiley.com/cogsci
FIGURE 4|Experimental outline of reasoning about hidden causal agents study in New Caledonian crows. Crows infer that the movement of this
stick is caused by the hidden human and so that the stick will not move again when the human leaves.
state.35,97 However beyond this, it is unclear whether
any corvid can mentally imagine itself in the past
or future, as humans do.37 When thinking about
themselves there is evidence that at least one corvid
species has a limited form of self-awareness,44,48 but
we has little idea how similar the corvid concept
of self is to that of humans.48 When thinking of
others a similar situation exists: work on experience
projection51 and desire-state attribution59 raises the
possibility that both Eurasian and scrub jays have
some kind of theory of mind, but further work
is required, particularly given how controversial
this area of science is.53,56,57 The mix of results
surrounding tool use and causality show several
corvid species are not limited by simple psychological
mechanisms but do make mistakes that humans
would not. This suggests that, when problem solving,
corvids use forms of cognition intermediate between
humans thought and simple learning. However, few,
if any, theoretical cognitive mechanisms have been
proposed which could predict the mix of success and
failures seen to date. Similarly, while recent work
raises the possibility that one corvid species, the New
Caledonian crow, is able to mentally simulate the
actions of hidden humans,91 other work shows that
such simulation is not used with novel man-made
objects.85 This suggests that mental simulation in
this corvid species is highly dependent on context
and experience, and so may be different from that
used in humans. What form it therefore takes, in
this species and others, is something of a mystery.
Finally, it is not yet known the degree to which the
species with the Corvidae share the same patterns of
thinking. We have little idea which cognitive traits
are ancestral, though it does appear the common
ancestor of the Corvidae was a moderate cacher.98 The
species in this group may think in very different ways,
depending on the evolutionary pressures they have
been subjected to, or they may all share a corvid mode
of thinking, using many of the cognitive mechanisms
discussed above.
CONCLUSIONS: WHAT NEXT?
Despite the many questions still surrounding corvid
cognition, this group has already taught us that
sophisticated types of intelligence are not just found
in ourselves and our closest relatives. However,
there is far more that we can learn from them. In
particular, pinpointing the cognitive mechanisms used
by this group has the potential to provide insights
into the evolution of not only the human mind,
but intelligence in general. By comparing the life
histories of various corvid species to the cognitive
mechanisms they possess, it should be possible to
discover whether particular selective pressures, such
as group living or tool manufacture, are sufficient for
the evolution of specific cognitive mechanisms that are
highly important to humans. Corvids, therefore, offer
us a way to peer into our own past and discover why
we think the way we do. Furthermore, examination of
these cognitive mechanisms in detail will show us the
extent to which the structure and form of cognitive
mechanisms evolve convergently. Thus work in this
area has the potential to allow us to build a deep
understanding of how intelligence evolves.
However, these exciting research possibilities
are predicated on one crucial point that behavioral
convergence does not equal cognitive convergence.
While it has been claimed that corvids are ‘feathered
apes’,23,96 it is currently difficult to know whether
corvids produce similar behaviors to apes because they
think in the same way, or because different cognitive
mechanisms have evolved to produce the same
behavioral output. Evidence of behavioral convergent
evolution in corvids and apes is not evidence of
cognitive convergent evolution: that these two groups
have actually evolved the same ways of thinking.
When Alan Turing proposed the Turing test
as a way of assessing machine intelligence he
made a crucial point about the machine’s cognitive
limitations.99 He noted that the machine would need
to mirror both the strengths and the flaws in human
cognition if it was to be attributed with human
©2014 John Wiley & Sons, Ltd.
WIREs Cognitive Science Corvid cognition
intelligence. If given an arithmetical problem, the
machine would need to not give an accurate answer
immediately, but instead take around 30 seconds to
reply, and occasionally make errors.99 Clearly, it is
more probable that a machine and a human have the
same cognitive mechanisms if they not only solve the
same problems, but also process problems at similar
speeds and exhibit similar error rates. These additional
points of similarity increase the chance that the
cognition being used is the same. Similarly, it is more
likely primates and corvids use the same cognition in
decision making if corvids mirror humans100,101 and
capuchins102 in exhibiting the bias of loss aversion
and it is more likely scrubs jays use the same memory
system as humans if they make the same retrieval
errors.34,103 Thus, if different minds produce the same
errors, have the same biases in information processing,
and have the same limitations, it is far more likely
they share the same cognitive mechanisms than if they
only produce the same solutions to a problem. Each
additional similarity found along these dimensions
increases our confidence that the underlying cognition
is the same.
Therefore, if we wish to use the Corvidae
to understand the evolution of intelligence, and so
realize this group’s full scientific potential, we need to
change our experimental practice. We should adopt
a ‘signature-testing’ approach, where experimenters
explicitly set out to search for the signatures of various
cognitive mechanisms in terms of their errors, biases
and limits, rather than a ‘success-testing’ approach
where experimenters simply examine whether a
problem can be solved or not. The use of brain
imaging techniques on corvids104 should greatly help
with this search. Only this signature-testing approach
can allow us to gain true insight into how our own
intelligence evolved and how intelligence may evolve
more generally on our planet and others. This could
lead to the intelligence of the Corvidae teaching us far
more than even ancient myths and legends suggest.
ACKNOWLEDGMENTS
The author thanks Mike Corballis, Thomas Suddendorf, Russell Gray, Corina Logan, Lucy Cheke and one
anonymous reviewer for helpful comments and advice. This work was supported by a University of Auckland
FRDF grant.
REFERENCES
1. Gibbs L. Aesop’s Fables. Oxford: Oxford University
Press; 2002.
2. Bierhorst J. The Mythology of North America.New
York: Quill; 1985.
3. Erdoes R, Ortiz A. American Indian Myths and
Legends. New York: Random House; 1984.
4. Reid B, Bringhurst R. The Raven Steals the light:
Native American Tales. New York: Random House;
1996.
5. Bogoras W. The Folklore of Northeastern Asia, as
compared with that of Northwestern America. Am
Anthropol 1902, 4:577–683.
6. Bellows HA. The Poetic Edda. New York: The
American-Scandinavian Foundation; 1923.
7. Hollander LLM. Heimskringla: History of the Kings
of Norway. Texas: University of Texas Press; 1964.
8. MacCulloch JA. The Religion of the Ancient Celts.
Edinburg: T. & T. Clark; 1911.
9. Green M. Animals in Celtic Life and Myth. London:
Routledge; 2002.
10. Jønsson KA, Fabre P-H, Ricklefs RE, Fjelds˚
aJ.Major
global radiation of corvoid birds originated in the
proto-Papuan archipelago. Proc Natl Acad Sci USA
2011, 108:2328–2333.
11. Jønsson KA, Fabre P-H, Irestedt M. Brains, tools,
innovation and biogeography in crows and ravens.
BMC Evol Biol 2012, 12:72.
12. Lefebvre L, Nicolakakis N, Boire D. Tools and brains
in birds. Behaviour 2002, 139:939–973.
13. Lefebvre L, Reader SM, Sol D. Brains, innovations
and evolution in birds and primates. Brain Behav Evol
2004, 63:233–246.
14. Lefebvre L, Whittle P, Lascaris E, Finkelstein A.
Feeding innovations and forebrain size in birds. Anim
Behav 1997, 53:549–560.
15. Hunt GR, Gray RD. The crafting of hook tools by
wild New Caledonian crows. Proc R Soc Lond B Biol
Sci 2004, 271:S88–S90.
16. Beck BB, Walkup K, Shumaker R. Tool use and
problem solving in animals. In: Seel NM, ed.
Encyclopedia of the Sciences of Learning. New York:
Springer; 2012, 3330–3333.
17. Hunt GR. Manufacture and use of hook-tools by New
Caledonian crows. Nature 1996, 379:249–251.
©2014 John Wiley & Sons, Ltd.
Focus Article wires.wiley.com/cogsci
18. Balda RP, Kamil AC, Bednekoff PA. Predicting
cognitive capacity from natural history. In: Nolan
V Jr, Ketterson ED, eds. Current Ornithology, vol. 13.
Springer; 1996, 33–66.
19. Fraser ON, Bugnyar T. Ravens reconcile after
aggressive conflicts with valuable partners. PloS One
2011, 6:e18118.
20. Seed AM, Clayton NS, Emery NJ. Postconflict third-
party affiliation in rooks, Corvus frugilegus.Curr Biol
2007, 17:152–158.
21. Jerison HJ. Evolution of the Brain and Intelligence.
New York: Academic Press; 1973.
22. Emery NJ. Cognitive ornithology: the evolution of
avian intelligence. Philos Trans R Soc B Biol Sci 2006,
361:23–43.
23. Emery NJ, Clayton NS. Comparing the complex
cognition of birds and primates. In: Rogers LJ, Kaplan
G, eds. Comparative Vertebrate Cognition. New York:
Springer; 2004, 3–55.
24. Reiner A, Perkel DJ, Bruce LL, Butler AB, Csillag A,
Kuenzel W, Medina L, Paxinos G, Shimizu T, Striedter
G. Revised nomenclature for avian telencephalon and
some related brainstem nuclei. J Comp Neurol 2004,
473:377–414.
25. Rehk¨
amper G, Frahm H, Zilles K. Quantitative
Development of Brain and Brain Structures in Birds
(Galliformes and Passeriformes) Compared to that in
Mammals (Insectivores and Primates)(Part 1 of 2).
Brain Behav Evol 1991, 37:125–134.
26. Jarvis ED, Yu J, Rivas MV, Horita H, Feenders G,
Whitney O, Jarvis SC, Jarvis ER, Kubikova L, Puck AE.
Global view of the functional molecular organization
of the avian cerebrum: Mirror images and functional
columns. J Comp Neurol 2013, 521:3614–3665.
27. Williams MF. Primate encephalization and intelli-
gence. Med Hypotheses 2002, 58:284–290.
28. Deaner RO, Isler K, Burkart J, van Schaik C. Overall
brain size, and not encephalization quotient, best
predicts cognitive ability across non-human primates.
Brain Behav Evol 2007, 70:115–124.
29. Alba DM. Cognitive inferences in fossil apes (Primates,
Hominoidea): does encephalization reflect intelligence.
J Anthropol Sci 2010, 88:11–48.
30. Dumont ER. Bone density and the lightweight
skeletons of birds. ProcRSocBBiolSci2010,
277:2193–2198.
31. Hone D, Dyke G, Haden M, Benton M. Body
size evolution in Mesozoic birds. J Evol Biol 2008,
21:618–624.
32. Aiello LC, Wheeler P. The expensive-tissue hypothesis:
the brain and the digestive system in human and
primate evolution. Curr Anthropol 1995, 36:199–221.
33. Clayton NS, Dickinson A. Episodic-like memory
during cache recovery by scrub jays. Nature 1998,
395:272–273.
34. Suddendorf T, Busby J. Mental time travel in animals?
Trends Cogn Sci 2003, 7:391–396.
35. Suddendorf T, Corballis MC. The evolution of
foresight: What is mental time travel, and is it unique
to humans? Behav Brain Sci 2007, 30:299–312.
36. Tulving E. How many memory systems are there? Am
Psychol 1985, 40:385.
37. Tulving E. Episodic memory: From mind to brain.
Annu Rev Psychol 2002, 53:1–25.
38. Cheke LG, Clayton NS. Mental time travel in animals.
Wiley Interdiscip Rev Cogn Sci 2010, 1:915–930.
39. Correia SP, Dickinson A, Clayton NS. Western
scrub-jays anticipate future needs independently of
their current motivational state. Curr Biol 2007,
17:856–861.
40. Raby C, Alexis D, Dickinson A, Clayton NS. Planning
for the future by western scrub-jays. Nature 2007,
445:919–921.
41. Suddendorf T, Corballis MC. New evidence for animal
foresight? Anim Behav 2008, 75:e1–e3.
42. Cheke LG, Clayton NS. Eurasian jays (Garrulus
glandarius) overcome their current desires to anticipate
two distinct future needs and plan for them
appropriately. Biol Lett 2012, 8:171–175.
43. Thompson SDAW. A Glossary of Greek Birds.
London: Oxford University Press; 1936.
44. Prior H, Schwarz A, G ¨
unt ¨
urk ¨
un O. Mirror-induced
behavior in the magpie (Pica pica): evidence of self-
recognition. PLoS Biol 2008, 6:e202.
45. Gallup GG. Chimpanzees: self-recognition. Science
1970, 167:86–87.
46. Plotnik JM, de Waal FBM, Reiss D. Self-recognition
in an Asian elephant. Proc Nat Acad Sci USA 2006,
103:17053–17057.
47. Reiss D, Marino L. Mirror self-recognition in the
bottlenose dolphin: a case of cognitive convergence.
Proc Nat Acad Sci USA 2001, 98:5937–5942.
48. Suddendorf T, Butler DL. The nature of visual self-
recognition. Trends Cogn Sci 2013, 17:121–127.
49. Boeckle M, Bugnyar T. Long-term memory for
affiliates in ravens. Curr Biol 2012, 22:801–806.
50. Marzluff JM, Walls J, Cornell HN, Withey JC, Craig
DP. Lasting recognition of threatening people by wild
American crows. Anim Behav 2010, 79:699–707.
51. Emery NJ, Clayton NS. Effects of experience and social
context on prospective caching strategies by scrub jays.
Nature 2001, 414:443–446.
52. Dally JM, Emery NJ, Clayton NS. Food-caching
western scrub-jays keep track of who was watching
when. Science 2006, 312:1662–1665.
53. Penn D, Povinelli D. The comparative delusion: the
behavioristic/mentalistic dichotomy in comparative
theory of mind research. In: Samuels R, Stich S,
eds. Oxford Handbook of Philosophy and Congnitive
Science. Oxford: Oxford University Press; 2009.
©2014 John Wiley & Sons, Ltd.
WIREs Cognitive Science Corvid cognition
54. van der Vaart E, Verbrugge R, Hemelrijk CK. Corvid
re-caching without ‘Theory of mind’: A model. PLoS
One 2012, 7:e32904.
55. Thom JM, Clayton NS. Re-caching by Western Scrub-
Jays (Aphelocoma californica) Cannot Be Attributed
to Stress. PLoS One 2013, 8:e52936.
56. Penn DC, Holyoak KJ, Povinelli DJ. Darwin’s mistake:
Explaining the discontinuity between human and
nonhuman minds. Behav Brain Sci 2008, 31:109–129.
57. Penn DC, Povinelli DJ. On the lack of evidence
that non-human animals possess anything remotely
resembling a ‘theory of mind’. Philos Trans R Soc B
Biol Sci 2007, 362:731–744.
58. Bugnyar T. Knower–guesser differentiation in ravens:
others’ viewpoints matter. Proc R Soc B Biol Sci 2011,
278:634–640.
59. Ostoji´
c L, Shaw RC, Cheke LG, Clayton NS. Evidence
suggesting that desire-state attribution may govern
food sharing in Eurasian jays. Proc Natl Acad Sci USA
2013, 110:4123–4128.
60. Whiten A. When does smart behaviour-reading
become mind-reading?. In: Carruthers P, Smith PK,
eds. Theories of Theories of Mind. Cambridge, UK:
Cambridge University Press; 1996, 277.
61. Hunt GR, Gray RD. Diversification and cumulative
evolution in New Caledonian crow tool manufacture.
Proc R Soc Lond B Biol Sci 2003, 270:867.
62. Bird CD, Emery NJ. Insightful problem solving
and creative tool modification by captive nontool-
using rooks. Proc Natl Acad Sci USA 2009,
106:10370–10375.
63. Chappell J, Kacelnik A. Tool selectivity in a
non-primate, the New Caledonian crow (Corvus
moneduloides). Anim Cogn 2002, 5:71–78.
64. Chappell J, Kacelnik A. Selection of tool diameter by
New Caledonian crows Corvus moneduloides. Anim
Cogn 2004, 7:121–127.
65. Wimpenny JH, Weir AAS, Kacelnik A. New
Caledonian crows use tools for non-foraging activities.
Anim Cogn 2011, 14:459–464.
66. Taylor AH, Hunt GR, Gray RD. Context-dependent
tool use in New Caledonian crows. Biol Lett 2011,
8:205–207.
67. Taylor AH, Elliffe D, Hunt GR, Gray RD. Complex
cognition and behavioural innovation in New
Caledonian crows. Proc R Soc B Biol Sci 2010,
277:2637–2643.
68. Taylor AH, Hunt GR, Holzhaider JC, Gray RD.
Spontaneous metatool use by New Caledonian crows.
Curr Biol 2007, 17:1504–1507.
69. Wimpenny JH, Weir AAS, Clayton L, Rutz C, Kacelnik
A. Cognitive processes associated with sequential tool
use in New Caledonian crows. PLoS One 2009,
4:e6471.
70. Weir AAS, Chappell J, Kacelnik A. Shaping of hooks
in New Caledonian crows. Science 2002, 297:981.
71. von Bayern AMP, Heathcote RJP, Rutz C, Kacelnik
A. The role of experience in problem solving and
innovative tool use in crows. Curr Biol 2009,
19:1965–1968.
72. Bird CD, Emery NJ. Rooks use stones to raise the
water level to reach a floating worm. Curr Biol 2009,
19:1410–1414.
73. Cheke LG, Bird CD, Clayton NS. Tool-use and
instrumental learning in the Eurasian jay (Garrulus
glandarius). Anim Cogn 2011, 14:441–455.
74. Taylor AH, Elliffe DM, Hunt GR, Emery NJ, Clayton
NS, Gray RD. New Caledonian crows learn the
functional properties of novel tool types. PLoS One
2011, 6:e26887.
75. Cheke LG, Loissel E, Clayton NS. How do children
solve aesop’s fable? PLoS One 2012, 7:e40574.
76. Seed AM, Tebbich S, Emery NJ, Clayton NS.
Investigating physical cognition in rooks, Corvus
frugilegus.Curr Biol 2006, 16:697–701.
77. Taylor A, Hunt G, Medina F, Gray R. Do New
Caledonian crows solve physical problems through
causal reasoning? Proc R Soc B Biol Sci 2009,
276:247–254.
78. Taylor A, Roberts R, Hunt G, Gray R. Causal
reasoning in New Caledonian crows: ruling out spatial
analogies and sampling error. Commun Integr Biol
2009, 2:311–312.
79. Martin-Ordas G, Call J, Colmenares F. Tubes, tables
and traps: great apes solve two functionally equivalent
trap tasks but show no evidence of transfer across
tasks. Anim Cogn 2008, 11:423–430.
80. Shepard RN. Toward a universal law of gener-
alization for psychological science. Science 1987,
237:1317–1323.
81. Pavlov IP. Conditioned Reflexes. New York: Dover
Publications; 1927.
82. Weir AAS, Kacelnik A. A New Caledonian crow
(Corvus moneduloides) creatively re-designs tools by
bending or unbending aluminium strips. Anim Cogn
2006, 9:317–334.
83. Heinrich B. An experimental investigation of insight in
common ravens (Corvus corax). Auk 1995:994–1003.
84. Taylor AH, Medina FS, Holzhaider JC, Hearne
LJ, Hunt GR, Gray RD. An investigation into the
cognition behind spontaneous string pulling in New
Caledonian crows. PLoS One 2010, 5:e9345.
85. Taylor AH, Knaebe B, Gray RD. An end to insight?
New Caledonian crows can spontaneously solve
problems without planning their actions. Proc R Soc
BBiolSci2012, 279:4977–4981.
86. Seed A, Boogert N. Animal cognition: an end to
insight? Curr Biol 2013, 23:R67.
©2014 John Wiley & Sons, Ltd.
Focus Article wires.wiley.com/cogsci
87. Gopnik A, Meltzoff AN. Words, Thoughts, and
Theories, vol. 1. Cambridge, MA: MIT Press; 1997.
88. Atran S. In Gods We Trust: The Evolutionary
Landscape of Religion. Oxford, UK: Oxford
University Press; 2002.
89. Atran S, Norenzayan A. Religion’s evolutionary land-
scape: Counterintuition, commitment, compassion,
communion. Behav Brain Sci 2004, 27:713–729.
90. Wolpert L. Causal belief and the origins of technology.
Philos Trans R Soc A Math Phys Eng Sci 2003,
361:1709–1719.
91. Taylor AH, Miller R, Gray RD. New Caledonian
crows reason about hidden causal agents. Proc Natl
Acad Sci USA 2012, 109:16389–16391.
92. Boogert NJ, Arbilly M, Muth F, Seed AM. Do crows
reason about causes or agents? The devil is in the
controls. Proc Natl Acad Sci USA 2013, 110:E273–E.
93. Dymond S, Haselgrove M, McGregor A. Clever crows
or unbalanced birds? Proc Natl Acad Sci USA 2013,
110:E336–E.
94. Taylor AH, Miller R, Gray RD. Reply to Dymond
et al.: Clear evidence of habituation counters
counterbalancing. Proc Natl Acad Sci USA 2013,
110:E337.
95. Taylor AH, Miller R, Gray RD. Reply to Boogert et al.:
The devil is unlikely to be in association or distraction.
Proc Natl Acad Sci USA 2013, 110:E274–E.
96. Emery NJ, Clayton NS. The mentality of crows:
convergent evolution of intelligence in corvids and
apes. Science 2004, 306:1903–1907.
97. Kohler W. The mentality of apes. 1924.
98. De Kort SR, Clayton NS. An evolutionary perspective
on caching by corvids. Proc R Soc B Biol Sci 2006,
273:417–423.
99. Turing AM. Computing machinery and intelligence.
Mind 1950, 59:433–460.
100. Kahneman D, Knetsch JL, Thaler RH. Anomalies: The
endowment effect, loss aversion, and status quo bias.
J Econ Perspect 1991, 5:193–206.
101. Kahneman D, Knetsch JL, Thaler RH. Experimental
tests of the endowment effect and the Coase theorem.
J Polit Econ 1990:1325–1348.
102. Chen MK, Lakshminarayanan V, Santos LR. How
basic are behavioral biases? Evidence from capuchin
monkey trading behavior. J Polit Econ 2006,
114:517–537.
103. Loftus EF. When a lie becomes memory’s truth:
Memory distortion after exposure to misinformation.
Curr Dir Psychol Sci 1992, 1:121–123.
104. Marzluff JM, Miyaoka R, Minoshima S, Cross DJ.
Brain imaging reveals neuronal circuitry underlying
the crow’s perception of human faces. Proc Natl Acad
Sci USA 2012, 109:15912–15917.
©2014 John Wiley & Sons, Ltd.
