A case of quick problem solving in birds string pulling in keas, Nestor notabilis.pdf

A case of quick problem solving in birds: string pulling
in keas, Nestor notabilis
DAGMAR WERDENICH*† & LUDWIG HUBER*
*Department for Behavior, Neurobiology and Cognition, University of Vienna
yKonrad Lorenz Institute for Ethology, Vienna
(Received 8 September 2004; initial acceptance 13 October 2004;
final acceptance 8 June 2005; published online 28 February 2006; MS. number: 8268R)
Keas are extremely inquisitive birds that explore and manipulate objects with their beak and feet when
foraging or at play. We investigated how captive keas face a problem that they have never encountered
before: food objects suspended from a long string. Using this well-probed test for means–end comprehen-
sion, our aim was to see whether, and how, the keas solved the task. The seven subjects showed immediate
interest in the string and, with the exception of a fledgling, reached the food by repeating combined ac-
tions of pulling up the string with the beak and holding loops of it against the perch with a foot. Four keas
completed the first trial within a few seconds, by showing only goal-directed behaviour, thus executing the
solution in a manner that could not be improved upon in nine further trials. The behaviour of the fledg-
ling shows that the crucial element of the string-pulling competence is the beak–foot coordination. We
then conducted four different two-string discrimination tests varying in the attachments and the spatial
relationship of the strings to assess the keas’ level of understanding of the functional properties of the
task. The subjects’ performance suggests that their attention is drawn very quickly to the relevant proper-
ties of the strings, that is, their connection between perch and food. These findings contribute to the on-
going debate on the distribution of higher cognitive skills in the animal kingdom by showing high levels
of sensorimotor intelligence in animals that do not use tools.
Ó 2006 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
For many decades, researchers have tried to get a glimpse
into the ‘folk physics’ of animals, that is, their common-
sense understanding of how the world works, as well as
why it works in the way it does (Köhler 1921; reviewed in
Beck 1980; Tomasello & Call 1997; Povinelli 2000; Matsu-
zawa 2001). Many researchers now believe that it is not
only humans but also many large-brained animals that de-
velop an increasing ability to understand causal relation-
ships with increasing experience, with language being
only a powerful accelerator of this process, rather than
a prerequisite (Rumbaugh et al. 2000).
Causal understanding requires assigning cause–effect or
means–end relationships to the physical or social world
through either observation or insight. Such causal re-
lationships between objects might lead to the representa-
tion of abstract concepts, such as ‘connectivity’ (Hauser
1997), or the construction of a chain of responses that
lead to a goal, or simply the generalization of responses
to similar situations. Only experimental work can eluci-
date which perceptual-motor abilities are involved and
which steps of the solution are achieved in cases where
animals appear at first sight to behave creatively or in-
sightfully. The apprehension of a cause–effect relationship
between two or more physical objects determines how ob-
jects are manipulated, which of several alternative objects
are used, and how the individual responses are assembled
into a coherent whole.
The ability to gain access to food that is out of reach has
long been regarded as a valid example of apprehension of
a cause–effect relationship (Köhler 1921; Piaget 1954).
Many experiments testing birds’ abilities to use a physical
object to obtain food that is out of reach have involved
the string-pulling task. Known since ancient times and
used in animal research for many decades (reviewed in
Seibt & Wickler, in press), this task involves the presenta-
tion of food suspended by a string or thread fixed to
a perch, not accessible from the ground or from flight
(Dücker & Rensch 1977). The most effective and intelli-
gent solution to this problem has been described by Hein-
rich (1995, 2000) with common ravens, Corvus corax. It
requires that the bird repeats the following sequence sev-
eral times: reaching down from the perch, grasping the
Correspondence: L. Huber, Department for Behavior, Neurobiology and
Cognition, Faculty of Life Sciences, University of Vienna, Biocenter,
Althanstrasse 14, A-1090 Vienna, Austria (email: ludwig.huber@
univie.ac.at).
855
0003–3472/06/$30.00/0 Ó 2006 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.
ANIMAL BEHAVIOUR, 2006, 71, 855–863
doi:10.1016/j.anbehav.2005.06.018

string with the beak, pulling the string up, placing the
pulled-up loop of string on the perch, then pressing
a foot down on the loop and letting go with the beak so
that the bird can reach down to pull up another loop
(Fig. 1). If a bird has achieved this solution spontaneously,
and the complete act has been accomplished in a rapid
and smooth manner without any fumbling, we would
probably agree with Thorpe (1956) in seeing nothing sug-
gestive of trial-and-error learning. However, out of more
than 15 species of birds initially tested with the string-
pulling task only ravens seem to understand the cause–
effect relationship between food, string and perch, use
effective techniques instantaneously, in contrast to a num-
ber of other species that do not, and generalize the solu-
tion to various situations (Heinrich 2000).
Besides corvids (Emery & Clayton 2004), parrots are es-
pecially well known for performing amazing cognitive
tasks. Recently, Funk (2002) reported some string-pulling
competence in yellow-crowned parakeets, Cyanoramphus
auriceps, and Pepperberg (2004) in language-trained grey
parrots, Psittacus erithacus. A parrot species that has re-
ceived public attention because of its alleged intelligence
and curiosity is the kea (Diamond & Bond 1999). This
bird’s explorative attitude, combined with its persistent
and rigorous destructiveness and almost total lack of neo-
phobia, makes it the ideal candidate for the study of
means–end comprehension and sensorimotor intelligence
in birds (Huber et al. 2001; Gajdon et al. 2004).
Despite the obvious problems of controlled testing in
field situations, Johnston (1999) successfully addressed
the string-pulling abilities of wild keas. Of 19 freely partici-
pating keas six solved the problem in their first trial, with
the median time to completion being 50 s. Other individ-
uals did not instantly master the task or never solved it,
suggesting that responses are not innate.
In our experiments we attempted to examine the keas’
sensorimotor intelligence under controlled laboratory con-
ditions by using a set-up that allows a distinction between
(1) finding the solution by chance, (2) executing innate
responses, (3) trial-and-error learning, and (4) means–
end comprehension. Because we were interested in the
goal directedness of the actions and were aware that keas
may learn from even a few errors, we performed a detailed
analysis of the initial steps they took to achieve success in
Figure 1. Six frames of a video clip (a–f) showing a kea performing string pulling in the first test.
ANIMAL BEHAVIOUR, 71, 4
856

their ﬁrst attempt to secure the food reward. To examine
the behavioural variability and the amount of learning we
compared the performance across trials. Furthermore,
following Heinrich’s (1995, 2000) comprehensive attempt
to examine the cognitive aspects of string pulling in ravens,
we devised four patterned string-pulling problems. By
varying shape, colour, length and weight of the strings as
well as their spatial relationship to each other we assessed
the keas’ level of understanding of the functional proper-
ties of the task.
METHODS
Study Animals
Seven captive keas at the Konrad Lorenz Institute for
Ethology in Vienna, Austria, participated in this study.
One group consisted of the female Kima (6 years) and the
males Bigo (4 years), John (3 years), Knut (2 years) and
Bruce (7 months) and the other of the female Tiffy
(6 years) and the male Mismo (3 years). All individuals
were born in captivity and reared by their parents. They
were fully habituated to humans who interacted with
them as feeders or experimenters.
The subjects were kept in two large aviaries, providing
outdoor conditions year round. The aviary for the large
group (15 10 m and 4 m high) and the aviary for Tiffy
and Mismo (5 4 m and 3 m high) could both be divided
into three compartments. Both aviaries were equipped
with tree trunks, rocks, pieces of wood, perches and
wooden shelters and the ground was covered with sand.
Chains holding heavy branches were present in the aviary,
but could not be lifted by the keas. Except for these items,
no string-like objects had been presented to them. Water
and food, which contained a mixture of vegetables, fruits,
seeds, margarine, ox heart and vitamin supplements every
day, were available ad libitum. Individual recognition of
the birds was facilitated by coloured leg bands. No licence
was required for this study.
Apparatus
We conducted the experimental tests in a visually
separated corner of the aviary, using a wooden perch
and different objects attached to one or more vertical
strings (Fig. 2). The perch (170 cm long, 5 5 cm in cross-
section) was supported by two poles (170 cm high) that
were screwed into wooden bases measuring 60 60 cm.
One or two plastic strings were nailed to the underside
of the perch. In the various tests we used strings that dif-
fered in colour, length and spatial relationships. We used
thin wire attached to the poles at a height of 115 cm to
cross the strings or to bring them into a slanted position
(Fig. 2). Fixed to the ends of the strings were objects of
different shape, texture and weight, such as a yellow toy
car (25 g), a black Kodak ﬁlm canister (20 g), a piece of
wood (50 g) and a stone (2000 g). Baiting one of the
objects with a mixture of butter and egg yolk provided
a food reward.
Procedure
The keas received six string-pulling tests. Before starting
the experiment we allowed the keas to investigate the
wooden perch without the string for 1 h on 3 consecutive
days. None of them had ever experienced rewards on
a string and they received no string-pulling training prior
to the experiments. They were tested individually with the
same number of trials in the same sequence of tests, but
we changed the order in which they participated every
day. Usually we conducted two sessions per day, one in
the morning (from 0900 hours) and one in the afternoon
(from 1500 hours). Video recording enabled us to analyse
the details of the keas’ behaviour.
Test 1
Test 5 Test 6
Test 3 Test 4
Test 2
Figure 2. Diagrams showing the single string in tests 1 and 6 and the
spatial relationships between the two strings in tests 2–5. The way in
which the crossed and the slanted strings were kept in position by
thin wires is also shown.
WERDENICH HUBER: STRING PULLING IN KEAS 857

Test 1: single string-pulling task
This initial task tested the keas’ ability to pull up a baited
toy that was suspended 70 cm from the horizontal perch
by a single yellow string. Before the experiment, the prese-
lected individual was lured into the test compartment and
the other individuals locked into another compartment of
the aviary, from which they could not see the test subject.
The subjects received 10 trials in immediate succession.
During the intertrial interval (20–30 s), the experimenter
(D.W.) baited the object on the string while the bird flew
to another perch in the aviary.
Test 2: object discrimination task
To see how the object on the string controlled the keas’
pulling behaviour, we presented in three consecutive
sessions of 10 trials each two strings, one with the baited
object attached, as before, and one with a piece of wood
attached. The three sessions differed in the baited object
(the toy of test 1 in the first session, a film canister in the
other two) and the colour of the strings (both yellow in
the first two sessions; the string with the baited object was
red in the third one). In this test, as in all other patterned
string problems, we determined the position of the
rewarded string for each session in an arbitrary manner
by tossing a coin. The subjects could investigate both
objects for about 10 s before the first trial. Test 2 followed
immediately after the last trial of test 1.
Test 3: crossed strings task
By crossing the strings we wanted to assess whether the
keas’ choice was determined by the spatial or the functional
relationship between string and reward, that is, if they
pulled at the string directly above the baited object as in the
40 trials before or based their choice on the functional
connection between food and string. The food container
was now suspended 20 cm laterally to where it was located
before (Fig. 2). We conducted two 30-trial sessions to exam-
ine improvement of performance. While in these sessions
we used strings of different colour to ensure that the birds
could follow the strings from one end to the other, we
used two yellow strings in a further 20-trial session.
Test 4: slanted strings task
In this task, the two yellow strings were bent off to the
same side (Fig. 2). Although the food container was again
directly below where the unbaited string was attached to
the perch, the strings did not cross. We conducted one
20-trial session.
Test 5: overload string task
In a 20-trial session, the keas could choose between two
yellow strings, one with the food container and the other
with a heavy (2000-g) stone baited with a large amount of
butter. This task tested whether the choices were influ-
enced by relevant physical properties of the attached
objects (their weight).
Test 6: overlength string task
In this 20-trial task we wanted to test the flexibility of
problem solving in keas. Over the preceding 160 trials, the
keas were trained to pull up the string from the perch they
were sitting on but in this task the length of the string was
140 cm, enabling the subject to reach the reward from the
ground.
RESULTS
Performance in the Single String-pulling Task
All keas showed immediate interest and flew directly to
the perch. None tried to obtain the reward by flying
directly at the toy or by jumping up to it from the ground.
With the exception of the fledgling (Bruce) they solved
the task in their first trial.
Table 1. Action list and action sequences in the keas’ first attempt to
secure the reward
Action
number
Efficient
actions N
Inefficient
actions N
Exploratory
actions
1 Lands on
perch
12 Tries to
place
string on
perch but
does not
succeed
17 Nibbles
string
2 Reaches for
string with
beak
13 Tries to
place
foot on
string but
does not
succeed
18 Nibbles
perch
3 Pulls up
string
with beak
14 Drops
string
after
having
pulled
it up
19 Touches
string
4 Places string
on perch
15 Sits on
perch
5 Uses one foot
to hold string
on perch
16 Flies off
perch
6 Uses both feet
to hold string
on perch
7 Pulls up string
until body is
upright and
therefore gains
more string
8 Grabs string
directly with
foot
9 Takes one or
more steps
sideways on
perch holding
string with beak
10 Lets go of string
with beak to
reach down
again
11 Gains
reward after
having pulled
up string
858

A detailed analysis of the actions of the keas in their first
attempt revealed that they fell into three subgroups with
regard to their goal directedness in the single string-
pulling task (Tables 1, 2). Four subjects showed mostly ef-
ficient, rather than exploratory, actions, including the key
elements of pulling up and stepping on the string (see
Table 1 for the categorization of actions). Given that the
birds were facing the problem for the first time, they
found this solution very rapidly (9–15 s; Table 3). Even
more striking, these ‘immediate solvers’ did not show
any improvement across the further nine trials of the
task. They also never attempted to fly off with the food ob-
ject once it was recovered.
Although taking the group as a whole, there was also no
significant difference in latency or number of pull-ups
between the first and the second trial (Wilcoxon test:
latency: T ¼ 7, N ¼ 7, P ¼ 0.24; pull-ups: T ¼ 10, N ¼ 7,
P ¼ 0.5) or the first and the 10th trial (latency: T ¼ 3,
N ¼ 7, P ¼ 0.06; pull-ups: T ¼ 1.5, N ¼ 7, P ¼ 0.06), it is ev-
ident that the two remaining subgroups behaved differ-
ently from the immediate solvers. The juvenile Knut and
the adult Tiffy showed many exploratory actions (nib-
bling), executed the efficient actions more slowly and
left the perch once or more before completing the task.
They dramatically improved their methods in the second
trial, but not thereafter (Table 3). The fledgling Bruce was
the only bird that was completely unable to secure the
food toy in his first attempts. Only after 1 month and
three further trials did he solve the string-pulling task.
One obvious measure of string-pulling skill is how often
the string is dropped (Table 3). In fact, the 7-month-old
fledgling’s (Bruce) bad performance in the first trial is re-
flected by the large number of drops. In his first trial he
pecked at the string and pulled it up with his beak a few
centimetres. Only twice was he able to hold the string
with his foot but he seemed unable to coordinate his
movements and to repeat the correct sequence several
times. After three unsuccessful sessions (each of 10 min
on 3 consecutive days) we stopped testing him for 1
month. When tested again he became successful only af-
ter three further 10-min trials. The obvious challenge
was to move the beak holding the string towards the
foot and thereby place the string directly underneath
the foot. When he managed to do this, he could hold
the string with the foot more accurately and repeat the
necessary movements until he reached the reward. In
Table 2. Action sequences of individual keas in their first attempt to secure the reward
Subjects Action sequence
Bigo 1-2-17-3-4-5-6-10-2-3-4-5-10-2-3-4-5-10-2-3-4-5-10-2-3-4-5-10-2-3-4-5-11
John 1-2-17-3-5-10-2-3-5-9-6-10-2-3-5-10-2-3-9-6-10-2-3-4-5-11
Kima 1-2-3-5-10-2-3-6-10-2-3-5-10-2-3-9-5-11
Mismo 1-2-17-3-9-5-10-2-3-9-5-10-2-3-9-5-10-2-3-5-10-2-3-5-11
Knut 1-2-17-17-10-18-2-17-17-10-2-17-2-3-5-10-2-3-5-10-2-3-9-5-10-2-3-10-14-2-17-10-2-3-5-10-2-3-9-5-10-14-16-1-2-3-9-5-
10-2-3-5-10-14-2-3-9-5-10-2-3-14-2-3-5-10-2-3-9-5-10-2-3-9-5-10-2-3-5-10-2-3-5-11
Tiffy 1-2-17-3-5-10-2-3-5-10-2-3-5-17-14-18-18-18-2-3-14-2-17-2-17-17-17-17-18-18-18-15-18-18-16-1-16-1-2-17-2-17-2-3-6-
10-14-2-19-2-3-6-10-2-3-5-3-4-17-3-9-5-10-2-3-8-10-17-10-14-18-15-2-17-3-10-14-2-3-6-10-2-3-9-5-10-2-3-5-10-2-3-5-10-
2-3-5-10-2-3-9-6-11
Bruce 1-2-17-17-3-13-14-2-17-3-3-12-14-2-3-3-12-13-14-2-3-3-13-7-13-14-18-18-18-18-18-2-3-4-5-17-14-2-3-13-14-2-3-14-2-3-
17-3-13-14-2-17-3-4-13-14-2-3-13-14-2-3-4-5-17-4-13-14-2-3-4-5-17-4-6-10-14-2-3-4-5-7-5-4-13-14-2-3-4-13-4-5-17-13-4-
6-17-8-17-7-13-14-2-3-4-5-7-13-14-2-3-7-13-7-9-4-5-10-2-3-3-9-4-13-14-2-3-3-7-9-4-5-7-5-7-14-2-3-3-7-14-2-3-3-7-9-4-
13-14-2-3-7-9-4-5-10-2-3-4-5-6-11
See Table 1 for list of actions. Exploratory and inefficient actions are in bold. The action sequence of Bruce was recorded during his second test
1 month after his first (unsuccessful) attempts (see text).
Table 3. Performance in the single string-pulling task
Trial
Subjects
Bigo Bruce* John Kima Knut Mismo Tiffy
T P D T P D T P D T P D T P D T P D T P D
1 12 5 0 369 30 19 15 4 0 9 4 0 120 16 4 13 6 0 330 17 5
2 10 6 0 24 4 2 11 7 0 21 7 1 18 9 1 26 9 1 32 11 2
3 9 5 0 17 3 0 4 5 0 15 5 0 10 5 0 13 5 1 17 6 0
4 17 7 1 33 4 1 6 5 0 11 5 0 10 6 0 16 5 1 15 8 0
5 7 3 0 40 8 4 6 6 0 16 6 0 11 6 1 26 5 1 14 6 1
6 10 4 0 13 4 2 13 5 1 13 5 0 13 6 1 24 6 3 29 11 2
7 22 6 1 120 10 3 6 7 0 17 7 0 17 8 1 15 3 1 18 5 0
8 13 8 0 13 1 0 8 7 0 17 7 1 9 5 0 8 3 0 14 4 0
9 10 4 0 86 8 5 9 5 0 11 5 0 10 5 0 14 4 2 25 10 0
10 13 5 0 65 21 7 6 6 0 12 6 0 10 5 0 6 1 0 11 5 0
T: time to completion (s); P: number of pulls; D: number of drops.
*For Bruce data are given not for his first session, but for the session including his first successful trial (see text).

his previous attempts, he held the string with the beak in
an upright position and tried to place the foot on the ver-
tically dangling string. While in this first successful trial
the step-on method emerged slowly, it was executed in
a quick and fluent manner in the next trial.
The female Tiffy also showed some improvement with
respect to the number of drops. In contrast, the remaining
keas showed small numbers of drops from the beginning
of the task (Table 3). These five individuals had no prob-
lem holding the string with their claws or by stepping
on it; four of them never dropped the string with the
food in the first trial.
Methods Used to Obtain the Reward
Although the task seemed to require straightforward
solutions, we found a considerable variation in the
techniques shown and in the frequency of their usage.
Taken together, the group of keas performed nine different
techniques when confronted with the various tests
(Table 4). Combinations of up to five different methods
per trial occurred. Only four methods, loop making, side
walking, turn and upright pull, were shown by all individ-
uals. The variation in methods used in alternation re-
mained consistently high across tasks.
Performance in the String Discrimination Tasks
An analysis of the subjects’ individual performances in
the four discrimination tests revealed some variation
between tests and subjects. We scored their choice in
a trial as ‘correct’ if they started with a clear pulling action
at the baited string. Note that at the onset of test 2, all
birds were proficient at the string-pulling task.
In test 2, which required a discrimination between the
baited and the unbaited object, all seven subjects per-
formed extremely well. Only two (Kima and Mismo) failed
in the first trial, but none in the following two trials. The
string with the baited toy was chosen by the group in 86%
of all 10 trials, in 83% of all further 10 trials with the food
container, and in 84% of the 10 trials with strings of
different colour.
In the crossed strings task, the birds faced much greater
problems. Only one bird (John) chose the correct string in
the first trial (as well as in 27 of the 29 following trials,
with the failures occurring in the second and 25th trials).
Three birds (Bruce, Knut and Mismo) learned quickly after
the failure in the first trial (with 24, 25 and 25 correct trials
thereafter), one bird only after a number of failures (Bigo,
with two correct choices in the first 13 trials and 14 correct
choices in the remaining 17 trials), and two birds (Kima
and Tiffy) remained at chance level throughout the 30
trials (12 and 16 correct trials). In the repetition phase
(further 30 trials), all birds were overwhelmingly correct,
with a minimum of 26 correct trials out of 30 trials (group
performance: 91%). When we used the same colour for
both strings, four birds failed in the first trial, and four
birds remained confused during 19 further trials. Only
Bruce, John and Knut performed significantly well (bi-
nomial test: P 0.02).
In the slanted strings task, all birds performed well.
Only Bigo failed in the first trial. The percentage of correct
choices in the first 10 trials of the whole group was 87%
and in the next 10 trials 86%. In the first trial of the
overload test, three birds (John, Knut and Tiffy) failed.
Some learning occurred thereafter, as the percentage of
correct choices in the first 10 trials of the whole group was
81% and in the next 10 trials 93%.
In the final overlength string task, the birds could easily
obtain the reward by flying to the ground near the food
container. However, after 160 trials in which they were
required to pull up the string from the perch, six birds
continued to use their pulling technique, although it
sometimes took them almost 1 min or 20 correct pulls.
Only one female (Tiffy) changed her behaviour. After 11
pulling trials she flew some distance away from the perch,
then returned to the apparatus by successively climbing
Table 4. The relative frequencies of methods used to solve the tasks
Subjects
Methods
Flip Foot grab Foot stretch Loop making Side walking Turn Upright pull Wrap Down
Bigo 0.4 0.7 0.3 58.9 32.4 1.1 6.2 0.0 0.0
Bruce 0.7 2.3 0.0 5.6 60.0 2.6 28.3 0.4 0.0
John 2.3 1.0 0.4 54.6 28.3 0.4 13.0 0.0 0.0
Kima 0.0 6.6 0.0 7.6 57.8 2.0 26.0 0.0 0.0
Knut 2.0 0.0 0.0 22.5 66.5 5.9 3.1 0.0 0.0
Mismo 0.0 0.8 0.0 9.7 76.6 3.8 9.0 0.0 0.0
Tiffy 0.5 3.8 3.3 23.2 32.4 7.1 27.7 0.0 2.1
Total 0.8 2.2 0.6 26.0 50.6 3.3 16.2 0.1 0.3
Flip: reaching down and flipping the string to the other side of the perch; Foot grab: grabbing the string with the foot and placing it on the
perch; Foot stretch: standing on the string with both feet, moving the foot with the loose end of the string laterally on the perch and thereby
pulling the string up; Loop making: reaching down, pulling up string with the beak, placing the foot on the string, letting go of the string with
the beak, remaining in place, reaching down again; Side walking: reaching down, pulling up the string with the beak, walking to the side of the
perch, placing the foot on the string, reaching down again; Turn: turning the whole body 180

while holding the string and stepping on the
additional string with the feet; Upright pull: pulling up the string till the body is in a completely upright position and thus gaining more string;
Wrap: pulling up the string, letting go at the other side of the perch, pulling up at the other side and by repeating these steps wrapping the
string around the perch; Down: taking the food container from the ground (only possible in test 6).
860

down branches and finally got close to the string on the
ground, from where she easily obtained the reward. In
all further trials she flew directly to the ground near the
food cup.
DISCUSSION
Inspired by Heinrich’s (1995, 2000) studies with ravens,
and in accord with findings obtained from wild keas
(Johnston 1999), our experiments with captive keas indi-
cate that these parrots are very efficient at finding a solu-
tion to a nontrivial technical problem and at
discriminating between stimuli based on the functionality
of an object and its relationship to other objects. So far,
only grey parrots and ravens have proved to behave simi-
larly well in such means–end object retrieval tasks (Pep-
perberg 2004; Heinrich Bugnyar 2005). Several
characteristics of our keas’ performance invite discussion
in comparison to other published experiments and to
the existing theories of the cognitive processes underlying
problem solving in animals.
First, all keas except the fledgling solved the string-
pulling task by coordinating beak and a foot in their first
trial. Four of those successful keas retrieved the food object
dangling on a string instantaneously and without show-
ing exploratory or ineffective methods (such as flying up
from the ground or pecking at the string as ravens did; see
Heinrich Bugnyar 2005). This is, to our knowledge, by
far the quickest manner in which the string-pulling prob-
lem has been solved by birds. Perhaps they had worked
out a solution to the retrieval problem even before their
first trial run.
Second, it is remarkable that from the first trial of the
initial single string-pulling task onwards there was no
significant improvement of the group across the test.
Neither the time to completion nor the number of pulls
or drops decreased in our whole bird sample in a signifi-
cant manner. In contrast, in all other birds tested so far,
including ravens (Heinrich 1995; Heinrich Bugnyar
2005), improvement in performance was on the whole
gradual and intermittent. We might therefore argue that
the overt construction, rather than the explicit training,
of a sequence of actions in order to bring about an effect
was the key achievement in the keas’ string-pulling
behaviour.
Third, the keas’ behaviour remained flexible, as they
used different techniques throughout all string-pulling
tests, that is, for about 180 trials. Although there was some
evidence of an increase in the efficiency of particular
action elements, the keas did not eliminate techniques.
We suggest that the elimination of methods is hampered
in keas by their strong inclination towards object explo-
ration and object play. Nevertheless, as is evident from the
short task completion times, the majority of their actions
were food directed and guided by some understanding of
the key functional features or affordances of the task. Only
in the test with the extra long string did they behave
stubbornly. However, 160 rewarded pull-ups of the string
might have narrowed their attention towards the end of
the string. Only tests with na€
ıve animals may elucidate
the true problem-solving capacities of keas in this task.
Finally, the keas showed very high success rates in the
discrimination tasks, thereby demonstrating attendance
to the end and the configuration of the strings. The two-
string choice experiments showed that the birds instan-
taneously distinguished between strings with baited and
unbaited objects. Such immediate solutions were obvi-
ously not achieved in the crossed and slanted strings tasks.
Except for one raven in Heinrich’s study (1995) there are
no reports to our knowledge of birds solving this problem.
Even for primates the solution is not always immediate
and may require an extended period of learning (reviewed
in Tomasello Call 1997; Povinelli 2000).
The most advanced solution in the crossed and slanted
strings tasks would require the possession of a ‘concept of
connectedness’ (Piaget 1954). However, without present-
ing in probe trials strings that vary in the visual salience
of connectedness, it remains unclear to what extent an
understanding of ‘contact’ is grounded in a genuine ap-
preciation of physical connection (Hauser 1997; Povinelli
2000). A less advanced although equally effective solution
would require attendance to the perceptually tangible cues
of the strings, particularly the continuity of the paths, as
opposed to causally imperceptible physical factors. How-
ever, tracing the paths of the strings is difficult when fac-
ing a crossing at some distance. The much better
performance in the slanted strings task than in the crossed
strings task might be counted as evidence that tracing the
paths of the strings was the critical element in the keas’
discrimination performance.
How does this study contribute to the ongoing discus-
sion about the evolution of cognition? In the past, the
string-pulling task has been used to examine whether
animals apply an evolutionarily ‘programmed’ solution, or
learn by trial and error (instrumental conditioning), or
show means–end behaviour. However, it is unlikely that
these alternatives constitute exclusive options for an
animal in the string-pulling situation. Obviously, the
execution of the motor acts involved rests on preprog-
rammed action elements. The real challenge in the string-
pulling task, of course, is the assembling of the actions
into a coherent whole. Gradual improvement was obvi-
ously not the dominant process underlying the keas’
performance. In addition, the collection of the necessary
proprioceptive experiences is arguably not sufficient for
the construction of an action sequence that is appropriate
for the task at hand. The early failures of the fledgling and
the great variability of techniques applied render an
explanation in terms of ‘preprogrammed’ solutions
unlikely.
Instead, there is increasing evidence that the apprehen-
sion of cause–effect relationships between two or more
physical objects, including the subject’s own body, guides
the covert assembly of the response sequence, the exact
manner in which objects are manipulated, and which of
several alternative objects are used (Hauser 1997; Heinrich
2000; Rumbaugh et al. 2000). Perhaps this is what Thorpe
(1956) called insight in contrast to trial-and-error learning,
in which actions are selected through reinforcement. We
believe that our keas’ string-pulling behaviour can be
best explained by such apprehension of a cause–effect re-
lationship and the adaptive reorganization of experience.

Our study supports the hypothesis that keas manage to
solve complex technical tasks by some kind of under-
standing of means–end relationships and by attending to
the functional properties of objects. Captive keas in
Vienna have shown an ability to manipulate group
members in a cooperation task (Tebbich et al. 1996), to
understand the affordances of locking devices through ob-
servation of a conspecific (Huber et al. 2001), and to solve
difficult technical tasks quickly (Gajdon et al. 2004). An
interesting feature of these studies is the striking differ-
ence in performance of wild and captive keas. When
tested under controlled conditions, wild keas failed to
show convincing forms of social learning and horizontal
transmission of technical innovations (Gajdon et al.
2004, in press), and also performed less convincingly in
the string-pulling task (Johnston 1999) than the birds in
the present study. However, we should not make strong
claims about this difference before determining the effects
of group arousal, neophobia and the presence of humans
in the wild. Furthermore, string-pulling behaviour, and
most probably the performance in all problem-solving
tasks, depends on the experience and skill of the individ-
ual acquired prior to the experiment and on social strate-
gies, as has been shown recently in grey parrots
(Pepperberg 2004). In that study, those birds with little
training in referential English requests succeeded in string
pulling, whereas birds that could request the suspended
item failed to obtain the object but engaged in repeated re-
questing. The wild keas in Johnston’s (1999) study might
also have been engaged in social strategies and therefore
might have shown less goal-directed string-pulling behav-
iour than our individually tested keas.
In conclusion, the findings from parrot studies have
shown that neither extensive tool use and manufacturing,
as in New Caledonian crows, Corvus moneduloides (Hunt
1996), nor food caching and feeding in the vicinity of
dangerous predators, as in ravens (Heinrich 1999; Bugnyar
Kotrschal 2002), are the evolutionary recipe for ad-
vanced intelligence in birds. For primates, numerous re-
searchers have proposed that intelligence is a correlate of
having a complex social system and a long life (the so
called ‘social function of intellect’ hypothesis put forward
first by Jolly 1966 and Humphrey 1976). Marler (1996)
and later Emery Clayton (2004) have suggested similar
selection processes for birds, and Pepperberg (1999) added
emphasis on the combination of intelligence and ad-
vanced communication skills in parrots. Indeed, parrots
may rival corvids and primates not only in relative brain
and telencephalic volumes (Iwaniuk et al. 2005), but
also in some cognitive abilities.
Acknowledgments
Financial support was provided by the Austrian Science
Foundation (FWF-grant P15027-BIO). We acknowledge
the hospitality of the Konrad Lorenz Institute for Ethol-
ogy, with special thanks to H. Winckler and D. Penn.
We are grateful for the assistance and support of C.
Grabmayr for the keas’ well-being. We thank S. Rechberger
for help in conducting the experiments, G. Gajdon, T.
Bugnyar and R. Johnston for valuable discussion and
comments on the manuscript, and C. Reid for improving
the English.
References
Beck, B. B. 1980. Animal Tool Behavior. The Use and Manufacture of
Tools by Animals. New York: Garland.
Bugnyar, T. Kotrschal, K. 2002. Observational learning and the
raiding of food caches in ravens, Corvus corax: is it tactical decep-
tion? Animal Behaviour, 64, 185–195.
Diamond, J. Bond, A. B. 1999. Kea, Bird of Paradox. The Evolution
and Behaviour of a New Zealand Parrot. Berkeley, California: Uni-
versity of California Press.
Dücker, G. Rensch, B. 1977. The solution of patterned string
problems by birds. Behavior, 62, 164–173.
Emery, N. J. Clayton, N. S. 2004. The mentality of crows: conver-
gent evolution of intelligence in corvids and apes. Science, 306,
1903–1907.
Funk, M. S. 2002. Problem solving skills in young yellow-crowned
parakeets (Cyanoramphus auriceps). Animal Cognition, 5, 167–
176. doi:10.1007/s10071-002-0149-4.
Gajdon, G. K., Fijn, N. Huber, L. 2004. Testing social learning in
a wild mountain parrot, the kea (Nestor notabilis). Learning and
Behavior, 32, 62–71.
Gajdon, G. K., Fijn, N. Huber, L. In press. Limited spread of inno-
vation in a wild parrot, the kea (Nestor notabilis). Animal Cognition.
Hauser, M. 1997. Artifactual kinds and functional design features:
what a primate understands without language. Cognition, 64,
285–308.
Heinrich, B. 1995. An experimental investigation of insight in
common ravens (Corvus corax). Auk, 112, 994–1003.
Heinrich, B. 1999. Mind of the Raven. New York: Harper Collins.
Heinrich, B. 2000. Testing insight in ravens. In: The Evolution of Cog-
nition (Ed. by C. M. Heyes L. Huber), pp. 289–305. Cambridge,
Massachusetts: MIT Press.
Heinrich, B. Bugnyar, T. 2005. Testing problem solving in ravens:
string-pulling to reach food. Ethology, 111, 962–976.
Huber, L., Rechberger, S. Taborsky, M. 2001. Social learning af-
fects object exploration and manipulation in keas, Nestor notabilis.
Animal Behaviour, 62, 945–954. doi:10.1006/anbe.2001.1822.
Humphrey, N. K. 1976. The social function of intellect. In: Growing
Points in Ethology (Ed. by P. G. Bateson R. A. Hinde), pp. 303–
317. Cambridge: Cambridge University Press.
Hunt, G. R. 1996. Manufacture and use of hook-tools by New
Caledonian crows. Nature, 379, 249–251.
Iwaniuk, A. N., Dean, K. M. Nelson, J. E. 2005. Interspecific
allometry of the brain and brain regions in parrots (Psittaciformes):
comparisons with other birds and primates. Brain, Behaviour and
Evolution, 65, 40–59.
Johnston, R. 1999. The kea (Nestor notabilis): a New Zealand prob-
lem or problem solver? M.Sc. thesis, University of Canterbury.
Jolly, A. 1966. Lemur social behavior and primate intelligence.
Science, 153, 501–506.
Köhler, W. 1921. The Mentality of Apes. London: Methuen.
Marler, P. 1996. Are primates smarter than birds? Current Ornithol-
ogy, 13, 1–32.
Matsuzawa, T. 2001. Primate Origins of Human Cognition and
Behavior. Tokyo: Springer.
Pepperberg, I. M. 1999. The Alex Studies. Cognitive and Communica-
tive Abilities of Gray Parrots. Cambridge, Massachusetts: Harvard
University Press.
Pepperberg, I. M. 2004. ‘Insightful’ string-pulling in grey parrots
(Psittacus erithacus) is affected by vocal competence. Animal Cog-
nition, 7, 263–266.
862

Piaget, J. 1954. The Construction of Reality in the Child. New York:
Ballantine Books.
Povinelli, D. J. 2000. Folk Physics for Apes. New York: Oxford Univer-
sity Press.
Rumbaugh, D. M., Beran, M. J. Hillix, W. A. 2000. Cause–effect
reasoning in humans and animals. In: The Evolution of Cognition
(Ed. by C. M. Heyes L. Huber), pp. 221–238. Cambridge, Mas-
sachusetts: MIT Press.
Seibt, U. Wickler, W. In press. Individuality in problem solving:
string-pullingintwoCarduelisspecies(Aves:Passeriformes).Ethology.
Tebbich, S., Taborsky, M. Winkler, H. 1996. Social manipulation
causes cooperation in keas. Animal Behaviour, 52, 1–10.
Thorpe, W. H. 1956. Learning and Instinct in Animals. London:
Methuen.
Tomasello, M. Call, J. 1997. Primate Cognition. New York: Oxford
University Press.

A case of quick problem solving in birds string pulling in keas, Nestor notabilis.pdf

More Related Content

Similar to A case of quick problem solving in birds string pulling in keas, Nestor notabilis.pdf (20)

More from Christine Maffla (20)

Recently uploaded (20)

A case of quick problem solving in birds string pulling in keas, Nestor notabilis.pdf