Metacognition in dogs: Do dogs know they could be wrong?

Article (PDF Available)inLearning & Behavior 46(4) · November 2018with 637 Reads
DOI: 10.3758/s13420-018-0367-5
Cite this publication
Abstract
In the current study, we investigated the question of whether dogs were sensitive to the information that they themselves had or had not acquired. For this purpose, we conducted three consecutive experiments in which dogs had to find a reward that was hidden behind one of two V-shaped fences with a gap at the point of the V. This setup allowed us to distinguish between selecting one of the fences by walking around it and seeking additional information by checking through the gap in the fence. We varied whether dogs had visual access to the baiting procedure or not. In addition, we manipulated the type and quality of reward as well as the time delay between baiting and choosing to analyze if the dogs’ searching behavior was affected. Our results were partly consistent with the findings of Call (Animal Cognition, 13 (5), 689–700, 2010) with great apes, on whose findings we based our experiments. We found that dogs checked more often through the corner of the V-shaped fence when they had not seen where the reward was hidden. Interestingly, dogs rewarded with toys selected the correct fence more often than dogs rewarded with food. Even though dogs’ performance was not affected by the food quality condition, dogs were significantly faster in fetching a high-quality food reward as opposed to a low-quality food reward. When testing whether forgetting and checking would increase as a function of delay, we found that although dogs slightly decreased in their success in finding the food when time delays were longer, they were not more likely to check before choosing. We show that – similar to apes – dogs seek additional information in uncertain situations, but their behavior in uncertain situations is less flexible compared to great apes. Electronic supplementary material The online version of this article (10.3758/s13420-018-0367-5) contains supplementary material, which is available to authorized users.
Metacognition in dogs: Do dogs know they could be wrong?
Julia Belger
1,2
&Juliane Bräuer
1,2
#The Author(s) 2018
Abstract
In the current study, we investigated the question of whether dogs were sensitive to the information that they themselves had or
had not acquired. For this purpose, we conducted three consecutive experiments in which dogs had to find a reward that was
hidden behind one of two V-shaped fenceswith a gap at the point of the V. This setup allowed us to distinguish between selecting
one of the fences by walking around it and seeking additional information by checking through the gap in the fence. We varied
whether dogs had visual access to the baiting procedure or not. In addition, we manipulated the typeand quality ofreward as well
as the time delay between baiting and choosing to analyze if the dogssearching behavior was affected. Our results were partly
consistent with the findings of Call (Animal Cognition, 13 (5),689700, 2010) with great apes, on whose findings we based our
experiments. We found thatdogs checked more often through the corner of the V-shaped fence when they had not seen where the
reward was hidden. Interestingly, dogs rewarded with toys selected the correct fence more often than dogs rewarded with food.
Even though dogsperformance was not affected by the food quality condition, dogs weresignificantly faster in fetching a high-
quality food reward as opposed to a low-quality food reward. When testing whether forgetting and checking would increase as a
function of delay, we found that although dogs slightly decreased in their success in finding the food when time delays were
longer, they were not more likely to check before choosing. We show that similar to apes dogs seek additional information in
uncertain situations, but their behavior in uncertain situations is less flexible compared to great apes.
Keywords Metacognition .Domestic dog .Seeking information .Comparative psychology
General introduction
While in recent years much attention has been given to what
animals understand about each other, only little is known
about what animals understand about their own mental pro-
cesses. Moreover, the evolutionary origins of metacognition
the ability to access, monitor, and control onesown
perceptual and cognitive processes and, thus, know about
ones own cognitive potentials as well as limitations (Flavell,
1979; Hertzog & Hultsch, 2000; Smith, Shields, &Washburn,
2003;Zohar,1999)are still widely debated (Rosati &
Santos, 2016). The question therefore arises whether human-
like forms of metacognition exist in other species (Carruthers,
2008; Crystal & Foote, 2011;Hampton,2009;Kornell,2009;
Smith, Beran, Couchman, & Coutinho, 2008). Thus, the gen-
eral issue we raise here is whether animals have access to what
they have seen and what they know, and whether they seek
additional information in situations of uncertainty.
However, the question is not only whether animals share
humanscapacity for metacognition (Foote & Crystal, 2007;
Smith, Shields, & Washburn, 2003; Smith, 2009), but also
what the best methods are for studying non-linguistic behavior
for evidence of metacognition in animals. Comparative psy-
chologists have conducted cognitive tests on non-human ani-
mals to determine whether they possess knowledge of their
own cognitive states by using memory and food concealment
as well as perceptual and information-seeking paradigms
(Kornell, 2014). It seems that some animals make certain
judgments in similar ways to humans, although not by directly
Electronic supplementary material The online version of this article
(https://doi.org/10.3758/s13420-018-0367-5) contains supplementary
material, which is available to authorized users.
*Julia Belger
jubelger@cbs.mpg.de; http://doglab.shh.mpg.de/dog-
cognition_de.php
Juliane Bräuer
http://www2.uni-jena.de/svw/allgpsy/team/braeuer-j.htm
1
Max Planck Institute for the Science of Human History, Department
of Linguistic and Cultural Evolution, Dogstudies, Kahlaische Strasse
10, 07745 Jena, Germany
2
Department for General Psychology and Cognitive Neuroscience,
Friedrich Schiller University Jena, Jena, Germany
Learning & Behavior
https://doi.org/10.3758/s13420-018-0367-5
accessing their memories but rather by drawing inferences
based on cues like ease of processing and reaction time
(Kornell, 2014).
Griffin (2004) has emphasized that all animals regularly
face uncertain situations, not only when they have to read
social signals but also when they have to make a determina-
tion about the presence of a predator or available food. It is
essential for survival to evaluate ambiguous information.
Therefore, it is clearly advantageous to differentiate between
certain and uncertain situations, paying the cost of seeking
extra information only when it is really necessary (Griffin,
2004). In a number of different experiments, it has been
shown that humans, dolphins, monkeys, and rats refuse to
complete trials that are difficult, such as at a threshold in an
auditory discrimination task. In other words, when the task is
difficult, the risk of failing at a task, and therefore not receiv-
ing a reward, is so high that it might not be worth the cost of
trying. In some cases, a wrong choice could, additionally,
result in a time-out (rhesus macaques: Hampton, 2001;
Smith et al., 2006; rats: Foote & Crystal, 2007;orangutans:
Suda-King, 2008; also, see Smith et al., 2003 and Smith, 2009
for reviews). Additionally, it has been argued that subjects
perform better in tests when they have the option to decline
trials as compared to when they are forced to make a decision
(Foote & Crystal, 2007;Hampton,2001). These results can be
interpreted as evidence for the fact that these species know
what they remember (but see Browne, 2004 and Carruthers,
2008 for different interpretations).
Several researchers have criticized such methods by argu-
ing that the results of these tests could be interpreted in an
associative-behaviorist way (Smith, Zakrzewski, & Church,
2016). More precisely, in a more difficult trial an uncertainty
state is created when a perceptual threshold is exceeded. By
using the uncertainty response, the animal will know about
knowing or not know whether it will successfully pass trials
that are at the perceptual threshold (Smith, Beran, Couchman,
Coutinho, & Boomer, 2009). However, studies using uncer-
tainty responses have been criticized because the animals
behavior might solely be based on learned responses to a
specific stimulus (Carruthers, 2008;Crystal&Foote,2009).
Call and Carpenter (2001) introduced a novel and different
approach to the question of metacognition: the information-
seeking paradigm that does not require extensive training or
prior knowledge. The key features of this more naturalistic
approach are that animals can seek additional information
when needed, which enables them to respond accordingly as
soon as they have gathered the relevant information (Call &
Carpenter, 2001). The experimental set-up in Call and
Carpenters study consisted of two parallel tubes that chim-
panzees, orangutans, and 2.5-year-old children observed. The
tubes were placed on a platform with their openings oriented
towards the subjects. Then the experimenter placed a piece of
food inside one of the tubes while ensuring that the subject
was aware of the baiting procedure. In order to receive the
reward, they had to touch the baited tube containing the bait
on the first attempt. They introduced two conditions: in one
condition, the subjects witnessed the baiting process (Seen
condition), while in the other condition, baiting took place
behind an opaque occlude that blocked the subjectsvisual
access to the bait (Unseen condition). All subject groups spon-
taneously bent down more often to look inside the tubes be-
fore making a decision during the Unseen condition. The au-
thors concluded that subjects had access to their own mental
states (Call & Carpenter, 2001).
The information-seeking paradigm has been subject to crit-
icism on the grounds that animals could just engage in a rou-
tine by looking for information instead of applying
metacognitive abilities (Call, 2010). Call (2012) pointed out
two sorts of alternative explanations of a non-metacognitive
nature. One alternative non-metacognitive approach is the
broad-beam explanation, which states that a non-
metacognitive construct actually accounts for the observed
results in studies on animal metacognition and not on moni-
toring processes of knowledge states. The second approach is
the narrow-beam hypothesis (see Call, 2012), which claims
that subjects who lack information about a rewardslocation
engage in search behavior until they find it. Many animals are
presumably engaged in this so-called Bsearch, locate, retrieve
routine,^which might be an alternative explanation for the
results in the hidden and visible trials. To address this issue,
Call (2010) introduced five conditions to test the flexibility of
the information-seeking behavior in great apes. He referred to
the so-called BPassport Effect,^i.e., that in humans as well as
in other animals, whether an individual will search for extra
information depends on various factors, such as the value of
the Breward^(i.e., a passport is more valuable than a tram
ticket) and the time delay between hiding and searching (i.e.,
re-checking for the passport when itwas packed yesterday, but
not 5 min ago).
In his study, Call (2010) introduced five conditions, manip-
ulating (1) whether subjects had visual access to the baiting,
(2) costs associated with seeking information, (3) food quality,
(4) additional information offered regarding the foodsloca-
tion, and (5) the time delay between baiting and selecting one
of the hiding places. Call concluded that his ape subjectsknew
that they could be wrong and that Bthe looking response ap-
pears to be a function of at least three factors: the cost of
looking inside the tube, the value of the reward and the state
of the information^(p. 699).
The domestic dog (Canis familiaris) represents an interest-
ing model to study animal cognition as during the long do-
mestication process dogs have evolved special skills to func-
tion effectively in the human environment, such as reading
human social and communicative skills (Marshall-Pescini &
Kaminski, 2014), in which they even outperform great apes
(i.e., Bräuer et al., 2006; Hare et al., 2002). However, the
Learn Behav
literature has no consensus on metacognition in dogs, i.e.,
whether they have knowledge of their own cognitive states
(Bräuer, Call, & Tomasello, 2004).
McMahon, Macpherson, and Roberts (2010) applied an
information-seeking paradigm, where subjectsneeded to fetch
a reward without immediately available information. To fetch
the hidden reward, dogs had to seek additional information.
More precisely, the experimental set-up was comprised of four
boxes, all of which were completely black, except for one box,
which had a white side. In an extensive training, dogs learned
that the reward was always hidden under the box with the
white side. In the experimental manipulation, the boxes were
rotated (45°,90°, and 135°) and, thus, the one white side
gradually rotated out of the dogsview. Their findings show
that the dogsaccuracy progressively declined. The authors
concluded that if dogs could use additional information, as
stated in the information-seeking paradigm, they should have
walked around the boxes in order to choose the correct one. In
a follow-up experiment, the authors again applied an
information-seeking paradigm, but this time in a human-
oriented context, to examine whether dogs would seek further
information. The reward could be hidden underneath one of
three boxes. Before being able to select one of the boxes, dogs
had to choose one of two human experimenters, where one
was the informant (i.e., person who would point to a location)
and the other was the non-informant (i.e., person who would
not provide any information by turning his or her back to the
dog). Dogs chose the informant significantly more often than
the non-informant, which suggests that dogs seek additional
information in an information-seeking task whenthe informa-
tion source is a human (McMahon, Macpherson, & Roberts,
2010).
Similarly, Bräuer, Call, and Tomasello (2004)investigated
whether dogs are sensitive to the information they themselves
have acquired. In an object-choice task, dogs were presented
with two identical wooden boxes, of which only one
contained a baited reward. On one side of each box was a
transparent window of glass with holes through which dogs
could seek extra information about whether the food was
placed in that box, such as by looking or smelling through
the window. On the other side of that box there was a lever,
and dogs were trained to select one of the boxes by pressing
this lever with their paw. In the Seen condition, the location of
the reward was shown to the dogs and therefore the dogs had
information about the location of the food. In the Unseen
condition they were prevented from seeing the baiting proce-
dure by two occluding barriers. Beforeselecting, the dogs had
the opportunity to seek extra information regarding the loca-
tion of the hidden reward, which would be especially useful in
the Unseen condition. The results showed that the dogs select-
ed the correct box in the Seen condition, but performed only at
chance level when they were prevented from seeing the re-
wards location. Most importantly, dogs rarely showed
checking behavior before selecting one of the boxes and they
did not check more often, as assumed, in the Unseen condition
compared to the Seen condition. The authors concluded that
their findings might indicate that dogs do not have access to
their own perceptual and knowledge states (Bräuer, Call, &
Tomasello, 2004).
However, both of these studies about metacognitive abili-
ties in dogs had some constraints. First, training was involved
(i.e., pressing the lever in Bräuer et al., 2004, and learning that
the food is in the box with the white side in McMahon et al.,
2010), and second, dogs were rewarded with food. It is pos-
sible that dogs would showa more flexible searching behavior
when they searched for their favorite toy a precise object
they Bpersonally^know and that they fetch and that does not
Bdisappear^as they consume it. (Note that in dog studies
about object permanence and memory, toys are often used as
a reward; see Collier-Baker et al., 2004; Fiset et al., 2003;
Miller et al., 2009;Mülleretal.,2014.)
More importantly, training could have led to an automatic
response in the dogs. Thus, dogs chose a box because they had
learned to do so and could not inhibit this response despite
their lack of information about the contents of the boxes
(Bräuer et al., 2004). Therefore, in the current study we inves-
tigated metacognition in dogs using a new set-up in which
dogs did not have to learn new behaviors in order to check
or to make their decision. As it is clearly adaptive to differen-
tiate between certain and uncertain situations (see above,
Griffin, 2004), and as dogs show special social cognitive skills
(Marshall-Pescini & Kaminski, 2014), we hypothesized that
dogs would show flexible metacognitive skills comparable
to those of apes and human children when tested in an
appropriate set-up.
On the basis of Calls experimental set-up and procedure,
we conducted three consecutive experiments in which dogs
had to find a reward that was hidden behind one of two V-
shaped fences in order to test whether dogs were sensitive to
the information that they themselves have or have not ac-
quired and whether they seek extra information in situations
of uncertainty. We manipulated the type (Experiment 1)and
quality of reward (Experiment 2), as well as the time delay
(Experiment 3) between baiting and choosing to analyze if the
dogssearching behavior was affected.
Dogs were presented with a Seen and an Unseen condition.
They could make their decision by walking around the V-
shaped fence, and they could check before choosing through
the corner of the V to see or smell whether the reward was
there. Based on the study of Call (2010), we predicted that the
dogs would check more frequently before choosing when they
had not seen where the reward was baited (Unseen condition)
than in cases when they had. We further predicted that dogs
would show more flexibility when searching for a toy (be-
ing a concrete object they often search for) than when
searching for food pieces (Experiment 1); that dogs would
Learn Behav
be more likely to check when high-quality food was hidden
as opposed to low-quality food (Experiment 2); and that
for higher time delays between baiting and choosing, dogs
would check more or have a reduced accuracy in finding
the reward (Experiment 3).
Experiment 1: Does the type of reward impact
dogsaccuracy in an information-seeking
task?
In the first experiment, we wanted to investigate if the
witnessing of baiting (Seen and Unseen conditions) and the
type of reward (toy or food) had an impact on dogsaccuracy
to find the baited reward. Therefore, we tested subjects in the
Seen and Unseen conditions, and half of the dogs searched for
food as a reward whereas the other half searched for their
favorite toy. We predicted that if dogs did not know what they
had seen, they would seek extra information. For the type of
reward, we expected the dogs to show more flexibility when
searching for a toy (which they often do in their daily life) than
searching for food pieces, as the favorite toy is a concrete
object that the subjects know.
Methods
Subjects
In total, 48 dogs (22 males and 26 females) of various breeds
and ages (range 1.511 years, mean 4.6 years) participated
successfully in the experiment. All subjects lived as pets with
their owners and received the normal obedience training typ-
ical for domestic dogs. The dog owners were not present dur-
ing the test and they were informed about the precise research
question as well as about the specifics of their dogstasks in
the study onlyafter the completion of the test, in order to avoid
potential training (by the owners).
The owners decided voluntarily to participate in this study,
andiftheywereinterestedtheywereprovidedwiththevideo
material of the performance of their dog after the test was
completed. All of the dogs were naïve to the information-
seeking task and did not have any prior knowledge of the
experiment. They were all healthy individuals with no known
sight or hearing impairments and no known history of aggres-
sion towards humans. Another precondition for this experi-
ment was that dogs had to be interested in food or toys in order
to participate in this study. For the toy condition, owners were
asked to bring their dogsfavorite toy to the testing sessions.
In total, 24 dogs were rewarded with food and 24 other dogs
were rewarded by playing with their favorite toy. Females
were not tested during estrous.
Materials
The test took place in a quiet room (8.5 m × 4 m) at Alte
Messe in Leipzig, Germany. The experimental set-up (Fig.
1) was comprised of a two-part apparatus. Each side consisted
of two V-shaped wooden fence structures (1.20 m × 1.00 m)
that were connected with a flexible hinge at the upper end to
form a V-shape with a 45° angle. At the lower end of each V-
construction was a gap of approximately 2 cm in width,
through which the first experimenter (E1) placed the reward
as bait on a small plate. Subjects could check whether the
reward was actually hidden there or not and make their deci-
sion based on this information by walking around the fences.
The distance between the corners of both barriers was 1.55 m.
A centerline indicated both the exact middle of the room as
well as the exact middle of the apparatus. Another marking,
1.60 m away from the corners, indicated the exact position
(i.e., the nearest point) at which the dog had to wait at the
beginning of each trial. E1 sat in the middle between the two
fences and was responsible for baiting the reward, and the
second experimenter (E2) was located next to the centerline
at the starting position to hold the dog, both facing E1. The
dogs had to choose one side and move around the V-shaped
fences, which was only possible by walking around the outer
sides. Two additional barriers prevented the subjects from
passing E1 and going around the inside to fetch the reward.
The dogs were rewarded with either food or their favorite toy.
In the Unseen condition, a curtain was installed to prevent
dogs from witnessing the baiting. All trials, including the pre-
test, were video-recorded by one camera that was installed
directly across from the apparatus.
Procedure and design
All experiments in this study consisted of three consecutive
units: familiarization, pretest, and experimental phase. First,
each dog received a familiarization to become familiar with
the testing room and to understand how to properly find the
reward at the corner of the fence. Accordingly, this was
followed by a pretest, which had to be successfully passed in
order to make sure that all participating subjects understood
the experimental set-up. Only dogs that passed the pretest took
part in the experiment.
We varied whether subjects received food or their favorite
toy as a reward. Therefore, one fundamental assumption was
that dogs that participated in this study had either a high de-
gree of interest in food or their favorite toy, and were further-
more motivated to fetch the reward after being released. This
was confirmed in the pretest and during the familiarization
sessions. As dogs have trouble obtaining a reward that is
placed at the inner corner of a V-shaped fence, even if the
fence is transparent (Pongracz, Miklosi, Vida, & Csanyi,
2005), we gave subjects the opportunity to investigate the
Learn Behav
testing room with the two V-shaped fences before the actual
experimental phase (Pongrácz, Vida, Banhegyi, & Miklósi,
2008). Before the final experimental phase began, three dogs
were tested in a pilot study. None of these subjects was in-
cluded in this study. In the following section, we will explain
the experimental procedure in more detail.
Familiarization
Before the actual test, we introduced the dogs to the testing
room to ensure that they understood the apparatus and were
able to find the reward without checking. The familiarization
was conducted successively, meaning that the subjects be-
came familiar with the task step-by-step. During the familiar-
ization sessions many breaks were given. Depending on the
condition, the dog was either rewarded with food or his or her
favorite toy. Similar to the tests, familiarization was always
conducted by the same experimenter E1, who baited the re-
ward, and an arbitrary second experimenter E2, who held the
dog at the starting position. E1 used a certain command to
motivate the dogs to find the baited reward (e.g., German
BOk,^BSuch!^(BLook!^), BWo ist es?^(BWhere is it?^)).
When the dog approached the reward he or she was rewarded
either by eating the food or by playing with the toy with E1.
At first, E1 led the way and showed the dogs where the
treat was placed by walking around the fence and hiding the
reward behind the corner. E1 used nonverbal cues, such as
pointing, showing, and eye gaze to further assist the dog in
finding the treat. After the baiting was finished, the dog was
released to search for the reward. In subsequent trials E1
placed the reward through the gap. The procedure was
repeated until the subject approached the reward by going
around the fence without trying to get the reward through
the gap.
The speed of familiarization always depended on the dogs
individual learning progress and motivation to find the re-
ward. Dogs were given a break from familiarization either
when they performed the action successfully, or when their
willingness, motivation, or attention was significantly
decreased.
Pretest
In order to pass the pretest, the dogs had to be able to walk
around the fence where the reward was placed, without
checking. The subjects passed the pretest when they found
the reward in four consecutive trials or four out of six trials
without checking through the gap. In the pretest, we did not
apply any manipulation and therefore subjects witnessed the
baiting procedure completely and had no delay between
baiting and choosing. Only subjects that passed the pretest
could take part in the actual experimental phase in the second
and third sessions. However, if dogs showed no interest in
participating, if they did not learn to find the food behind the
fence within 120 min, or showed no interest in the reward,
they were excluded from this study and marked as dropouts.
For this reason, we had to exclude six dogs from the study.
Experimental phase
After becoming familiar with the testing room and passing the
pretest, the subjects were tested in two consecutive sessions.
The general procedure in the experimental trials was the same
Fig. 1 Basic set-up for Experiments 1, 2, and 3
Learn Behav
for all dogs: Two experimenters, E1 and E2, tested all subjects
individually. One experimenter (E1) had to be the same person
for all trials, as in the pretest. The second experimenter (E2),
however, could be any person. At the beginning of each trial
the dog was held by E2 at the starting position while E1 knelt
between the two fences. E1 then held up the reward to show it
to the dog while calling his or her name to get the dogs
attention. The baiting process differed according to two
conditions:
1) In the Seen condition, E1 baited the reward while
allowing the dog to see the baiting process. E1 leaned
over one fence and put the reward through the gap onto
the plate behind of the fence. E1 then returned to the
middle of the fences, placing her arms parallel to her
body.
2) In the Unseen condition, E2 closed the curtain so that the
dog could not see the baitingprocess. E1 touched first the
left and then the right gap of the two fences while placing
the reward through one of them. After that, E1 again
touched both gaps simultaneously in order to make sure
that the subject could not hear where the reward was
baited. Then she went back into the middle of the two
fences, placed her arms parallel to her body, and told E2
to open the curtain.
After the baiting process was complete, E2 released the dog
and E1 called his or her name and encouraged him or her to
find the reward. In both conditions, E1 did not move and
avoided giving any cues to the dog. She waited until the dog
had made his or her choice by walking around one fence. If the
dog chose the correct fence he or she was allowed to eat the
food or to fetch the toy, and E1 played with him or her by
throwing the toy. If the dog chose the wrong fence, i.e., where
the reward was not hidden, E1 took him or her by the collar
and led him or her behind the correct fence. E1 showed the
reward to the dog but the dog was not allowed to eat it or play
with it. After the dogs had eaten the food or played with the
toy, or the reward was shown to them (when they were
wrong), the trial was over and a new one began.
The reward was placed behind one of the barriers in the
Seen condition only whenthe dog looked and paid attention to
E1. The dogs attention was essential for the continuation of
the experiment as the dog needed to witness the wholebaiting
process. After placing the reward, E2 leaned back to the mid-
dle and placed his or her arms parallel to his or her body
without looking at the dog. It was important that both E1
and E2 did not give any accidental cues (e.g., gaze, pointing,
non-verbal cues) and, thus, they looked down at the floor
while waiting.
Half of the dogs were tested with the food reward and half
of the dogs were tested with the toy reward. They were pre-
sented in two sessions on 2 days, so that each dog received the
Seen condition 12 times per day and the Unseen condition 12
times per day. Within a session, there was a break after half of
the trails. The order of the conditions and the location of the
food were randomized, with the stipulation that a condition
occurred no more than two trials in a row, and that the food
was not hidden on the same side in more than two consecutive
trials. Each dog received 24 trials of each of the two condi-
tions, totaling 48 trials (see Online Supplementary Materials
for details).
Data scoring and analysis
All trials were analyzed from the videotapes. We scored the
following three variables for each trial: success (correct
choice), checking, and latency. For success we scored whether
the dogs selected the correct fence, having at least the front
paws and shoulder behind the outer side of the fence where the
reward was baited. For checking behavior we coded whether
and where the dogs checked before choosing by approaching
the gap, having the mouth less than 10 cm from the gap while
hesitating for at least a half a second. Finally, we scored the
latency to select a fence in the trials when subjects did not
check. Therefore, we recorded the time from E1s first call
of the dogs name until his or her front paws and shoulder
had crossed the outer part of the V-shaped fence.
To assess inter-observer reliability, one independent ob-
server scored a randomly selected sample of 20% of the trials
where the dogs were rewarded with food and the trials where
the dogs were rewarded with a toy. Reliability was excellent
for correct selection (food: Cohens Kappa=0.98, N=240; toy:
Cohens Kappa=0.98, N=239), for checking behavior (food:
Cohens Kappa=0.95, N=240; toy: Cohens Kappa=0.77,
N=239), and for the latency to select (food: Pearson
Correlation r=0.80, N=182; toy: Pearson Correlation r=0.77,
N=153).
For the analysis, we used repeated measures 2 × 2
ANOVAS with the within-subject-factor condition (Seen vs.
Unseen) and the between-subject-factor reward (food vs. toy).
To test for learning over trials, we used repeated measures 2 ×
2 × 2 ANOVAS with the within-subject-factors condition
(Seen vs. Unseen) and session (first vs. second session) and
the between-subject-factor reward (food vs. toy). For compar-
isons against chance within one condition, one-sample t-tests
were used, as indicated.
Results
Success
The dogs selected the correct fence in 94% of the trials in the
Seen condition and in 57% of the trials in the Unseen condi-
tio,n and were above chance in both conditions (Seen:
t(47)=31.09, p<0.001; Unseen: t(47)=3.90, p<0.001, one-
Learn Behav
sample t-tests). They performed better in the Seen than in the
Unseen condition (F(1.46)=282.04, p<0.001), and they
showed increased accuracy when they were rewarded with
the toy (F(1.46)=5.95, p=0.019), but there was no interaction
between Condition × Reward (F(1.46)=1.77, p=0.190).
Checking
Figure 2presents the mean percentage of trials in which the
dogs checked for the different rewards in the two conditions.
The dogs checked more frequently in the Unseen condition
than in the Seen condition (F(1.46)=35.69, p<0.001), and they
tended to check more when they were rewarded with the toy
(F(1.46)=3.91, p=0.054). There was no interaction of
Condition × Reward (F(1.46)=0.28, p=0.601).
Checking and success
If subjects checked in the Seen condition, they then selected
the correct fence above chance in 95% of the cases
(t(37)=20.32, p<0.001, one-sample t-test). Similarly, if they
checked in the Unseen condition they were correct above
chance in 68% of the cases (t(44)=4.41, p<0.001, one-
sample t-test). Thus, the dogssuccess rate was higher when
they checked in the Seen condition than in the Unseen condi-
tion (F(1.35)=40.47, p<0.001). There was no effect of reward
(F(1.35)=2.65, p=0.112) and there was no interaction between
Condition × Reward (F(1.35)=1.47, p=0.234). Figure 3illus-
trates the checking behavior of the two groups of dogs in the
Unseen condition.
Latency
Food-rewarded dogs selected a fence faster than toy-rewarded
dogs in the cases when they did not check (F(1.45)=12.20,
p=0.001), but there was no effect for condition (F(1.45)=2.18,
p=0.147) and no interaction effect (F(1.45)=0.06, p=0.816).
Learning
Regarding success, we found no learning over trials in this
experiment. Subjects did not select the correct fence more
ofteninthesecondsessioncomparedtothefirstsession
(F(1.46)=0.008, p=0.930). However, there was a significant
interaction effect (Condition × session × reward:
F(1.46)=4.429, p=0.041). In contrast, subjects checked more
in the first session than in the second session. There was a
significant effect for session (F(1.46)=8.099, p=0.007), but
no interaction effect.
Individual performance
Individuals were above chance when they selected the correct
fence in 18 (75%) out of 24 trials or more (binomial test p=0.5,
N=24, P=0.02). Forty-five dogs were above chance in the
Seen condition (21 food rewarded and all 24 toy rewarded
dogs). Eight dogs were above chance in the Unseen condition
(two food-rewarded and six toy-rewarded dogs). Two food-
rewarded dogs never checked whereas all toy-rewarded dogs
checked at least twice.
Discussion
The dogs checked more often before selecting the correct
fence when they did not see where the reward was hidden.
They showed a flexible checking behavior, indicating that
dogs may have access to their own visual perception.
Similar to primates (Call, 2005;Call&Carpenter,2001;
Hampton et al., 2004; Marsh & MacDonald, 2012; Perdue,
Evans, & Beran, 2018), they sought extra information
when they did not know the rewards location. When dogs
did not witness the baiting, they were able to adapt their
behavior by gathering additional information that might
have led to success; by checking, they could select the
correct fence where the reward was hidden. Moreover, they
were able to revise their choice when they began their
inspection at the wrong fence.
Overall, the dogs showed similar checking and searching
patterns to primates. However, there were three differences
compared to primates.
First, dogs in general checked less than the apes tested
by Call (2010) and Call and Carpenter (2001), and were
therefore less likely to be successful in the Unseen condi-
tion (they only performed slightly above chance level).
Second, having begun checking the contents of the con-
tainers before choosing, the apes continued to do so
throughout the remaining trials (Call & Carpenter, 2001).
For the dogs, we did not find such an effect. Dogs either
Fig. 2 Mean percent of trials in which the dogs checked for the different
rewards in the conditions (+/-SE) in Experiment 1
Learn Behav
checked very often or rarely. Moreover, there was a de-
crease in checking behavior between the first session and
the second session. Thus, although the dogs checked less
in the second session, they did not learn the most effective
strategy over trials, i.e., checking more when they had not
seen and less when they had seen where the reward was
hidden. This means that their flexible checking behavior
was not learned during the experiment.
The third difference was that the dogs in contrast to the
apes were not always accurate when they checked. This
might be due to the set-up, as looking and smelling through
the narrow gap might lead to less accuracythan lookinginside
a tube. But still dogs as a group were able select the correct
fence above chance level in the Unseen condition.
Can we conclude from these results that dogs have access
to what they have seen? Studies using the information-
seeking paradigm have been criticized because subjects
may simply engage in a search for information routinely
without any metacognitive involvement. According to this
hypothesis, individuals engage in a variety of exploratory
responses until they detect the reward (Hampton et al.,
2004; Kornell et al., 2007).
However, as Call (2010) pointed out, this is unlikely for
two reasons: the tested primates selected the correct tube in
about 20% of the trials after only having looked inside the
empty tube (Call & Carpenter, 2001;Call,2005; Marsh &
MacDonald, 2012; Perdue, Evans, & Beran, 2018). Dogs in
the current study were also able to make this inference by
exclusion (as was also shown in other studies, see, e.g., Aust
et al., 2008; Erdohegyi et al., 2007; Wallis et al., 2016), al-
though less than the primates, in about 5% of the cases. This
means that subjects did not need to smell or see the reward to
select the correct alternative. The second reason why it is
unlikely that subjects simply engaged in a search for informa-
tion routinely is that the dogs, like the primates, also checked
when they had seen where the reward was hidden (in more
than 10% of the trials). However, it is unlikely that they had
forgotten the location of the reward because the delays were
very short and subjects were correct in nearly 100% of the
trials even when they did not check.
Our results are in contrast to previous findings of Bräuer
et al. (2004) and McMahon et al. (2010), Experiment 1.The
dogs in these studies were apparently influenced by the fact
that they were trained before the test in how to select the
food toy
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
1st check-empty-
wrong
1st check-empty-
correct direct
1st check-empty-
correct with 2nd
check
1st check-baited-
wrong
1st check-baited-
correct
Fig. 3 Percentage of dogsperformance regarding checking behavior
followed by their decision for one side depicted for food and toy in the
Unseen condition in Experiment 1. Depicted are all five possibilities of
dogschecking behavior with food and toy rewards, i.e., which fence they
attempted first when they check and which side they selected.By chance
dogs can first check the baited side, then they can either go to the correct
side (first check-baited-correct) or wrongly to the fence where the
reward is not hidden (first check-baited-wrong). When dogs check the
wrong side ontheir first attempt, they can either then select the wrong side
(first check-empty-wrong), or they can do a second check at the baited
side and select the correct fence (first check-empty-correct with second
check), or they can choose the baited side without further checking (first
check-empty-correct direct)
Learn Behav
correct box. In the current study the dogs also had some pre-
vious experience with the apparatus but they did not have to
learn to press a lever or a cue to locate the reward. Thus, with
the current paradigm dogs could search naturally for the re-
ward and it was shown that they are able to distinguish be-
tween a situation in which they had and had not seen where
the reward was hidden. Our results are supported by
Experiments 2 and 3 of McMahon et al. (2010), in which dogs
had a choice between an informant and a non-informant. The
dogs preferred to approach the informative human who then
pointed to the location of the reward. This again suggests that
dogs are seeking extra information when they do not know
where the reward is hidden.
Interestingly, dogs selected the correct fence more often
when they were rewarded with the toy, and they then also
tended to check more often. Thus, it is possible that dogs
search in a more flexible way when they are rewarded with
a toy. Dogs also showed flexible searching strategies when
they searched for a toy in a number of other studies
(Erdohegyi et al., 2007;Fiset,2009; Fiset et al., 2000,
2003,2006). Why were dogs more successful with the
toy? One possibility is that they were able to perceive the
toy better when they were checking through the gap, as it is
bigger. The second possibility is that the dogs rewarded
with food were less motivated to search for their reward
than the dogs that searched for the toy. That is very unlike-
ly because the dogs actually approached the food reward
even faster than the toy reward. It could, however, be the
case that the dogs were too motivated to get the food re-
ward,sothatitwasmoredifficulttobepatientenoughto
check before choosing. In other words, dogs may have
been more impulsive, and therefore less likely to show
metacognitive abilities, when the reward was food. A third
possibility is that the dogs perceived the two rewards in
different modalities. It is not clear how the dogs perceived
the reward behind the gap, whether they saw or whether
they smelled it. It is possible that the dogs used smell to
checkforthefoodandvisiontocheckforthetoy,andthat
the visual modality makes them more flexible (see also
Szetei et al., 2003). In an information-seeking experiment
with capuchin monkeys conducted by Vining and Marsh
(2015), subjects were either shown where the food was
hidden, they could infer its location, or they were not given
information about the location of the food. Monkeys also
had the opportunity to search for extra information, and
similar to our dogs they used this opportunity especially
in the Unseen condition but less in the Seen condition. But
when the monkeys potentially could infer the rewardslo-
cation, they were more likely to search for further informa-
tion. The authors conclude that capuchins only
metacognitively control their information seeking in situa-
tions in which information is presented in the visual do-
main (Vining & Marsh, 2015).
The fourth possibility lies in the nature of the rewards. The
favorite toy is a concrete object that the subjects know. Thus, it
is a focused search, as subjects know exactly what they are
looking for. In contrast, searching for food is more diffuse, as
there could potentially be more pieces around (although sub-
jects probably perceived that the test is about one piece).
Moreover, the dogs in their daily life probably have much
more experience with searching for a toy, and especially their
favorite toy, than searching for food.
Experiment 2: Do subjects check more when
a high-quality reward is involved?
In Experiment 1, we demonstrated that dogs seek out extra
information when they have not seen where a reward was
hidden and that they were more accurate when their favorite
toy was hidden. Following Call (2010), we were interested in
the question whether the location of a high-quality reward was
better remembered than the location of a low-quality reward.
We predicted that dogs would check more often when a high-
quality reward was baited as opposed to a low-quality reward.
In this experiment, new subjects that were unfamiliar with the
task were presented with two types of reward in Seen and
Unseen trials.
Methods
Subjects
We tested 24 dogs that did not take part in the previous exper-
iment but were chosen based onthe same selection criteria. As
food was given as a reward, it was crucial for the experiment
that dogs were motivated by food. There were 12 females and
12 males ranging from 1 to 6 years of age. As in Experiment 1,
all subjects lived as normal family dogs and were individually
tested at the Alte Messe in Leipzig, Germany.
Materials
The same experimental set-up was used, including the ap-
paratus with the aforementioned two V-shaped wooden
fence structures as in Experiment 1. Again, a gap of ap-
proximately 2 cm in width was used to hide the reward. For
high-quality food, dogs were given meat sausages
(BHundewürstchen^), while dry dog food served as low-
quality food.
Procedure and design
The basic procedure was the same as in Experiment 1:While
E1 kneeled between the two barriers facing the dog, E2 held
him or her at the starting position. The two experimenters
Learn Behav
tested all subjects individually, whereby one experimenter
(E1) placed a piece of food behind one of the fences, as was
the procedure in Experiment 1.
After the familiarization and the pretest, a classical food
preference test was conducted before each session to ensure
that dogs had a preference for one type of food. We assumed
that dogs would prefer meat sausages as high-quality food as
opposed to dry dog food, which was seen as low-quality food.
The preference test was conducted in another part of the room,
where dogs were presented with a wooden table. To ensure
that dogs really preferred the meat sausages (the high-quality
food) over the low-quality reward, E1 sat behind the table
across from the dog and fed him or her with a piece of high-
quality and a piece of low-quality food. E1 moved towards the
dog, holding a piece of food in each hand close to the dogs
nose, and then placed simultaneously a piece of each type of
food at the end of the board. Dogs were included if they
showed a clear preference for the high-quality reward (sau-
sages), i.e., chose the high-quality reward above chance level
in the four food preference tests. Overall, dogs chose the pre-
ferred food in 89% of trials (t(23)=16.31, p<0.001, one-
sample t-test).
Only if the pretest and food preference test were success-
fully passed, were the Seen and Unseen conditions, which
were similar to Experiment 1, introduced. It was also varied
whether high-quality food (sausage) or low-qualify food (dry
food) was hidden, resulting in four conditions: Seen-high /
Seen-low / Unseen-high / Unseen-low. Dogs were tested in
four sessions (two sessions per day). Each session consisted of
16 trials: the food preference test has four trials and each of the
four main conditions contained three trials. All trials were
presented randomly with the requirement that a condition oc-
curred in no more than two trials in a row. Moreover, the food
was not hidden on the same side in more than two consecutive
trials.
1
Data scoring and analysis
All trials were videotaped and scored in the same way as in
Experiment 1. Thus, we scored success (correct choice),
checking, and latency. The inter-rater reliability, which was
based on 20% of the trials, was very good for correct selection
(Cohens Kappa=0.988, N=24), for checking behavior
(Cohens Kappa=0.792, N=24), and for the latency to select
(Pearson Correlation r=0.936, N=24). For the main analysis
we used a 2 × 2 ANOVA with the within-subject factor con-
dition (Seen vs. Unseen) and the between-subject factor re-
ward (high- vs. low-quality reward).
Results
Success
The dogs selected the correct fence in 94% of the trials in the
Seen condition and in 52% of the trials in the Unseen condi-
tion, and were above chance in the Seen condition
(t(23)=22.20, p<0.001), but not in the Unseen condition
(t(23)=1.12, p=0.274, one-sample t-tests). They were more
accurate in the Seen than in the Unseen condition
(F(1.23)=239.74, p<0.001), but there was no effect for the
type of food (F(1.23)=0.64, p=0.429), and no interaction ef-
fect (F(1.23)=0.89, p=0.354).
Checking
Figure 4presents the mean percentage of trials in which the
dogs checked for the two different food rewards in the two
conditions. The dogs checked more frequently in the Unseen
condition than in the Seen condition (F(1.23)=8.32, p=0.008),
but there was no effect for the type of food (F(1.23)=0.74,
p=0.400) and no interaction effect (F(1.23)=1.00, p=0.328).
Checking and success
If subjects checked in the Seen condition they then selected
the correct fence above chance in 98% of the cases
(t(18)=23.82, p<0.001, one-sample t-test). However, if they
checked in the Unseen condition they were correct only in
60% of the cases, which was not above chance (t(21)=1.44,
p=0.165, one-sample t-test). The dogssuccess rate was higher
when they checked in the Seen condition than in the Unseen
condition (F(1.9)=18.76, p=0.002). These results suggest that
dogs in the Seen condition simply might have re-assured
themselves that the food was still there, but did not use
checking successfully in the Unseen condition. Moreover,
there was no effect of type of food (F(1.9)=1.01, p=0.342)
1
Similar to Call (2011), we implemented 12 trials of a control condition at the
end of each test to make sure that the dogs remembered where the food was
hidden. None of the tested dogs made a mistake.
0%
10%
20%
30%
40%
50%
unseen seen
High
mean percent of trials checking
Low
Fig. 4 The mean percentage of trials in which dogs checked for the two
different food rewards in the two conditions in Experiment 2
Learn Behav
and there was no interaction between Condition × Type of
food (F(1.9)=0.91, p=0.365).
Latency
On average it took subjects 2.8 s to select a fence in the cases
when they did not check before choosing. Dogs selected a
baited fence faster when the food was preferred than when it
was not preferred (F(1.22)=6.59, p=0.018), but there was no
effect for condition (F(1.22)=0.69, p=0.414) and no interac-
tion effect (F(1.22)=0.20, p=0.659).
Learning
There was no learning over trials. Subjects did not select the
correct fence more often in the second session compared to the
first session: although there was a significant effect for condi-
tion (F(1.23)=239,60, p<0.001), there was no interaction ef-
fect (Condition × Session F(1.23)=2.38, p=0.137) and for ses-
sion (F(1.23)=0.03, p=0.863). Similarly, subjects did not
check more in the first session compared to the second session
(F(1.23)=0.27, p=0.608), and there was no interaction effect
(Condition × Session F(1.23)=1.28, p=0.270), but again there
was an effect for condition (F(1.23)=8.30, p=0.008).
Individual performance
Again, individuals were above chance when they selected the
correct fence in 18 (75%) out of 24 trials or more (binomial
test p=0.5, N=24, P=0.02). Whereas 23 dogs were above
chance in the Seen condition, only one dog was above chance
in the Unseen condition. Two dogs never checked at all, 22
dogs checked at least once in the Unseen condition and 19
dogs checked at least once in the Seen condition.
Discussion
As in Experiment 1, dogs checked more frequently in the
Unseen condition than in the Seen condition, but there was
no effect for type of reward. Dogs remembered the locations
of both food types equally well, and did not remember the
location of a high-quality reward better. We predicted that dogs
would check more often when a high-quality reward was baited
as opposed to a low-quality reward, independent of whether
subjects had or had not witnessed the baiting. However, that
wasnotthecaseasdogsshowedthesamepatternsfor
checking, no matter whether the food was preferred or not.
One could argue that dogs did not perceive or forgot which
type of food was hidden. However, dogs selected the baited
fence faster when a high-quality reward was hidden compared
to a low-quality reward, indicating that they did indeed know
which food was hidden. This increased selection of preferred
food could indicate some evidence for the response
competition hypothesis (Hampton, Zivin, & Murray, 2004).
It predicts that if a highervalue reward isavailable,the subject
will be more motivated to go for it, i.e., in our case to go faster.
However, as we did not find decreased checking for the high-
value reward, the evidence remains weak.
In sum, in contrast to apes and humans, dogschecking
response was independent of the value ofthe reward, although
they were aware of the type of food that was hidden.
Experiment 3: Does forgetting predict
checking?
In this experiment, we raised the question whether time delay
had an impact on dogsaccuracy and checking responses. The
delay between baiting the reward and selecting one of the
fences was manipulated to foster forgetting and examine
whether checking would increase accordingly. Thus, we
adapted the previously used information-seeking paradigm
and varied the time delay (5, 20, 60, 120 s) between baiting
the fences and letting dogs choose one side (similar to Call,
2010). Longer time delays are associated with a higher degree
of difficulty to locate a baited reward (Call, 2010). Because
forgetting would predict an increase in checking, we proposed
that longer time delays lead to greater forgetting and, thus,
foster checking.
Methods
Subjects
The selection criteria for the subjects were the same as in the
previous two experiments, i.e., dogs had to be interested in
food and to be able to pass the pretest. All subjects were
normal family dogs that lived as pets with their owners in
Jena and surroundings. In total, 25 privately owned dogs (11
males and 14 females; mean age = 5.21 years) of various
breeds and ages (range 112 years) participated for the first
time in this kind of experiment. All 25 dogs were rewarded
with food either Frolic or, in case of food allergies, equally
preferred food.
Materials
All tests were conducted in a test room (7.20 m × 5.50 m) at
the Dog Lab of the Max Planck Institute for the Science of
Human History in Jena from April to August 2017. The ex-
perimental set-up was exactly the same as in the previous
experiments. Additionally, we used two thick blue mats that
were placed in front of the gaps and served to block visual
access to the fencescontents (see below). All trials, including
the pretest, were video-recorded by one camera that was
installed directly across from the apparatus.
Learn Behav
Procedure and design
In this experiment, the general procedure was the same as in
the Seen condition of Experiments 1 and 2, but we varied the
time delay between baiting and the dogs release. We used
time delays of 5, 20, 60, and 120 s. E1 measured the exact
time delays with a stopwatch, starting right after the reward
was baited and E1 was in the initial position.
Similar to Call (2010), we also wanted to implement a
BBlocked^condition in which dogs were prevented from
checking by placing two thick blue mats in front of the gaps.
However, as dogs did not show any difference in their behav-
ior between the conditions BBlocked^and Unblocked^for
success, checking, or latency, we concluded that we could
not prevent them from checking as they used their nose to
check. Thus, as this manipulation did not work, we treat the
BBlocked^and BUnblocked^trials as one condition.
In total, each subject received two administered 24-trial
blocks (one block for Seen and one block for Unseen trials),
resulting in a total of 48 trials. The order of the time delays (5,
20, 60, and 120 s) was randomized for all dogs within eight
trials and repeated in the exact same order afterwards. Food
was placed an equal number of times on each side with the
only restriction that the reward was hidden not more than
twice in a row in the same place in a session.
Data scoring and analysis
All trials were analyzed and scored from the video material in
the same way as in Experiments 1 and 2. Thus, we used the
three measures checking, success, and latency.
In order to assess inter-rater reliability, a second observer
unfamiliar with the task scored a randomly selected sample of
20% of the trials, which equaled a total of five dogs. Subjects
were chosen randomly. For all measures, the inter-rater reli-
ability was excellent and similar to Experiments 1 and 2
(Correct choice: kappa = 1.0, Checking: kappa = 0.80,
Latency: Pearson Correlation r=0.78, N=25). For the main
analysis we used a 1 × 4 ANOVA with the within-subject
factor time delay.
Results
Success
On average, dogs selected the correct fence in 93% of the
trials.
2
For a time delay of 5 s, they chose the correct side in
94% of trials, for 20 s in 95% of trials, and for both 60 and
120 s in 91% of trials (see Fig. 5). We found a significant
effect for time delay (F(3.72)=3.21, p=0.028). A paired-
sample t-test revealed that dogs were significantly more accu-
rate in 5 s compared to 60 s (t(24)=2.681, p=0.013), 20 s
compared to 60 s (t(24)=2.071, p=0.049, and 20 s compared
to 120 s (t(24)=2.089, p=0.047).
Checking
On average, dogs checked in 28% of trials with a time
delay of 5 s, in 27% of 20 s, in 31% of 60 s, and in 35%
of 120 s (see Fig. 5). We had assumed that dogs would
check more when the task is more difficult, e.g., when
the time delay between hiding and the possibility of
searching for the food is longer. However, the statistical
analysis revealed that there were no significant differences
between the four time delays. Dogs did not check more
often depending on time delay (F(3.72)=2.086, p=0.11).
Latency
Subjects took on average 3.1 s to select a fence in the cases
when they did not check. There was no effect for latency
(F(3.69)=2.038, p=0.12), thus, subjects did not take signifi-
cantly longer to retrieve the reward as a function of delay
when they did not check.
Learning
In this experiment, there was no learning over trials. Subjects
did not select the correct fence more often in the last session
compared to the first session: Although there was a significant
effect for delay (F(3.72)=3.21, p=0.028), there was no inter-
action effect (delay × session F(3.72)=1.21, p=0.31) and for
session (F(1.24)=1.81, p=0.19). Similarly, subjects did not
check less in the first session as opposed to the second session
(F(1.23)=0.104, p=0.75), there was no interaction effect (de-
lay × session F(3.69)=0.79, p=0.5), and no effect for delay
(F(3.69)=1.97, p=0.13).
Discussion
In this experiment, we investigated metacognition in dogs by
assessing the impact of time delay in an information-seeking
task. We found that the dogsoverall retrieval accuracy was
significantly higher for shorter timedelays, i.e., dogs were less
accurate when the delay was longer. However, in contrast to
the apes tested by Call (2010), dogs did not check more often
in situations in which the task was more difficult. Thus, dogs
did not search for extra information when they were uncertain,
which might suggest that they did not have access to their own
knowledge in that situation.
Similar to Call (2010), we also wanted to implement a
BBlocked^condition in which dogs were prevented from
2
As mentioned above we did not find any significant difference between
blocked and unblocked trials, and therefore analyzed the data regardless of
this manipulation.
Learn Behav
checking by using an occlude in front of the gap. However,
that manipulation did not work, as dogs did not show any
difference in their behavior between the BBlocked^and
BUnblocked^trials. Thus, dogs checked and were equally
successful in BBlocked^trials, meaning that they were able
to check through the occlude. This indicates that dogs used
their olfactory sense to check whether the reward was present
or absent, which is not so surprising as dogs very much rely on
their nose when they search for a reward (Gazit & Terkel,
2003;Miklosi,2007; see also Bräuer & Belger, 2018).
General discussion
Similar to apes, monkeys, and 2.5-year-old children (Call &
Carpenter, 2001; Hampton et al., 2004), dogs tend to actively
seek extra information when they have not seen where a re-
ward is hidden. Although subjects checked more often before
selecting the correct fence when they did not see where the
reward was hidden (Experiments 1 and 2), their searching
behavior was not affected by their preference for a type of
food (Experiment 2). Manipulating the time delay between
baiting and choosing slightly affected dogsperformance:
subjects were significantly less accurate, but they did not
check more often for higher time delays (Experiment 3). In
contrast to previous studies (Bräuer et al., 2004)wewereable
to demonstrate that dogs showed some aspects of information-
seeking behavior related to metacognition, but less flexibly
than apes.
The main objective was to examine whether dogs were
sensitive to the information that they themselves have or have
not seen and whether they seek extra information in situations
of uncertainty. As shown in Experiments 1 and 2, dogs
checked more when they did not witness the baiting proce-
dure. This suggests that they expected the bait was hidden
behind one of the fences and, additionally, they grasped that
they did not have enough information about where exactly it
was hidden. Therefore, dogs must have adapted their
searching behavior to increase their chance of success. We
showed that dogs checked more in the Unseen condition, al-
though they were not as successful as in the Seen condition
when they checked. We were able to replicate this effect in
two experiments with two independent cohorts of dogs.
However, we found one major difference between
Experiment 1and Experiment 2regarding the performance
of the two cohorts of dogs (probably due to the fact that a
toy reward was easy to follow, see discussion of Experiment
1and below). While checking in the Unseen condition in
Experiment 1indeed helped dogs to increase their accuracy
above chance level, increased checking behavior in
Experiment 2did not lead to a higher accuracy. In other words,
although dogs checked more often in the situation of uncer-
tainty (i.e., when they did not witness the baiting process) in
Experiment 2, they did not find the food more often than what
was expected by chance. Thus, although this did not lead
necessarily to increased success, dogs looked for extra infor-
mation. They sometimes checked, but not until they were
certain where the food actually was. This might indicate that
dogs sometimes have a problem inhibiting the approach to the
reward,evenwhentheyperceivethattheyneedtogatherextra
information. Apes in the study of Call (2010) did not have that
problem. They could get the information with a glance into the
tube. However, overall, dogs were potentially able to gather
enough information through the gap in order to get enough
information to find the reward (as proven in Experiment 1).
5s 20s 60s 120s
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Successful trials Checking
mean percent of successful trials
Fig. 5 The mean percent of trials in which dogs retrieved the reward successfully in Experiment 3. Dogs showed higher accuracy in trials with shorter
time delays compared to trials with longer delays between baiting and choosing
Learn Behav
All dogs that passed the pretest indicated that they under-
stood the experimental set-up and they also knew that a re-
ward was hidden behind one of the fences. Similar to the tip-
of-the-tongue phenomenon I know that I know something
but cannot retrieve the information checking in the Seen
condition could be seen as some kind of verification process
to maximize the chance of reward. The fact that dogs checked
more when they had no knowledge of the rewardslocation
(Unseen condition) could suggest that dogs show
metacognitive abilities, as they meet one of the assumptions
of knowing about knowing (Beran, Brandl, Perner, & Proust,
2012; Fleming, Dolan, & Frith, 2012; Hofer & Pintrich, 1997;
Metcalfe & Shimamura, 1994; Nelson & Narens, 1994).
From our results we can furthermore conclude that dogs
not only showed increased accuracy but also checked more
when the reward was their favorite toy as opposed to food.
One possible reason for this is that the dogsmotivation was
higher when a toy was at stake (see also above). Another
possibility is that they smelled the food and therefore checked
less often. However, dogs were faster in approaching food
than approaching the toy behind the fence (when not
checking), suggesting that their motivation for food was
higher. Hence, one might argue that dogs had an inhibition
problem for food, which is furthermore confirmed by the food
quality condition: dogsretrieved the preferred food faster than
the less preferred food, although there were no effects for
success as well as for checking. Thus, we speculate that the
latency to approach the food might be correlated with motiva-
tion. The higher the motivation, the less dogs are able to in-
hibit a direct approach without checking. Consequently, dogs
greater performance with a toy reward might not only be ex-
plained by their experience with searching for a toy but also by
the fact that they are not as over-motivated as with food.
Indeed, in an inhibition task with food, dogs were shown to
commit a number of seek errors, simply induced by ostensive-
communicative cues (Topál et al., 2009).
As pointed out by Hampton (2009), several studies on non-
human animal metacognition showed that difficult trials in
memory or perception tests were avoided while the searching
behavior could be adapted by gathering more information to
maximize the reward. The dogs overall performance may be
the result of response competition theory (Hampton, Zivin, &
Murray, 2004), as an alternative explanation. Knowing the
location of the food may have predisposed the dogs to select
a side while excluding all other options, such as searching for
the reward. According to this interpretation, dogs had two
competing options in our experimental design: retrieving food
or searching for further information. In the Seen trials the drive
to retrieve the reward was dominant, and so the dogs went
directly to the location where the reward was hidden. In
Unseen trials, however, the dogs did not knowwhere the food
was located and therefore the drive to search for information
was more dominant (Hampton, 2009). Thus, one could even
argue that searching for the food is the default behavior of
foraging dogs, and this default behavior is inhibited by know-
ing where food is.
While Call (2010) defined the looking responses (i.e.,
bending down to look into the tubes) as crucial features for
seeking additional information for the apes, we introduced
checking through the gap of the V-shaped fence as an equiv-
alent measure. Dogs did not show any differences in their
performance between situations in which the gap was and
was not blocked by an occlude. That means that they were
able to successfully check through the occlude, indicating that
they mainly used their olfactory sense to check whether the
reward was present or absent. This is not as surprising as dogs
very much rely on their nose when they search for a reward
(see Bräuer & Belger, 2018), and their olfactory perception is
proven to be excellent (Vonk & Leete, 2017). Thus, it is likely
that dogs and apes used different senses for checking. Indeed,
other studies have also shown that apes and dogs use different
strategies to deal with the same task. For example, Bräuer and
Call (2011) investigated object individuation in dogs and apes
by implementing a classical violation-of-expectation para-
digm. Their findings revealed that while apes showed in-
creased begging and looking behaviors, dogs showed in-
creased smelling when their expectation was violated
(Bräuer & Call, 2011). Moreover, other studies have shown
that dogs sniff more with increasing difficulty of the task, be it
when searching for a toy (Bräuer & Belger, 2018)orinobject
permanence tasks (Gagnon & Dore, 1992), thus gathering
information from other sensory modalities when one was not
sufficient. Future studies investigating metacognition in dogs
should therefore consider that dogs will mainly use their sense
of smell when searching for extra information in situations of
uncertainty.
So far, our results have only been interpreted in the light of
humanlike metacognitive abilities while other alternative non-
metacognitive explanations could also apply. According to the
non-metacognitive anxiety model by Carruthers (2008), the
subjects react to their anxiety produced by their knowledge
states and not to their knowledge states, which are opaque to
the individual. This alternative explanation could also offer an
alternative explanation for the passport effect (Call &
Carpenter, 2001; Call, 2010). Subsequently, this would imply
that not receiving the high-quality reward generates a higher
state of anxiety as opposed to not receiving the low-quality
reward. Therefore, dogs may be more likely to seek informa-
tion even though they already know where the reward is
baited, since the costs of failing to locate the high-quality
reward would be higher. The same anxiety model can be ap-
plied to our Seen and Unseen condition in which dogs
checked more often when they had not seen where the reward
was hidden: while in the Unseen condition more anxiety
should result in an increase in checking, less anxiety entails
less checking in the Seen condition. According to the response
Learn Behav
competition hypothesis (Hampton, Zivin, & Murray, 2004),
which potentially explains behavior without evoking a
metacognitive decision, checking in the Seen condition
should, contrary to our results, be reduced for high-quality
rewards, because the strength of the motivation to reach the
food would be much higher. This, however, does not match
with our results, which show that dogs checked more in the
Unseen condition.
In contrast to Calls(2010) study withapes, we did not find
evidence that the dogssearching for extra information
depended on the value of the food reward (i.e., food quality)
and the time delay between hiding and searching. As for the
time delays, we found that although dogsaccuracy was better
for shorter delays, they did not adapt their searching strategy
to compensate their lack of knowledge by checking. In con-
trast to the apes, dogs checked in fewer trials, and more im-
portantly they did not check more for longer time delays. One
could argue that this was due to a ceiling effect, as dogs overall
selected the correct fence in 93% of trials, and the pressure for
seeking extra information was low. However, apes showed a
similar accuracy (see Call, 2010, Fig. 3) but showed increased
looking for longer delays. It might, however, be that dogs
would show increased checking when the pressure is higher,
i.e., when their accuracy gets much lower as the delays are
longer. However, from the current data we can conclude that
dogs do not have the flexibility that is described in the pass-
port effect, and thus their search for extra information does not
depend on the value of the reward or on the time delay be-
tween hiding and searching.
In sum, we tested in three experiments whether dogs
know that they could be wrong. Our hypotheses that dogs
show flexible metacognitive skills were not fully con-
firmed and our results were only partly consistent with
Calls(2010) results. Dogs checked significantly more in
the Seen than in the Unseen condition, indicating that they
may have metacognitive abilities to some extent. Checking
was voluntarily used to reduce the probability of being
wrong and to maximize the possible reward. However,
dogssearching behavior for extra information did not depend
on the value of the food reward or the time delay between
hiding and seeking, which according to Call (2010) would
be clear evidence that they knew that they could be wrong.
Dogs are able toadapt their searching behavior by looking for
extra information in a flexible way, indicating that they have
access to what theyhave seen. However, further work is need-
ed to determine which specific monitoring processes related to
metacognition are involved.
Acknowledgements Open access funding provided by Max Planck
Society. We thank Katrin Schumann, Ines Neuhof, Marie Nitzschner,
and Nina Oettel for helping with data collection, and Franziska Becker
as well as Loren Billings and Susanne Michaelis for coding. We also
thank Tanya Behne, Josep Call, and Juliane Kaminski for useful com-
ments and Anne Gibson for proofreading of the manuscript.
Open Access This article is distributed under the terms of the Creative
Commons Attribution 4.0 International License (http://
creativecommons.org/licenses/by/4.0/), which permits unrestricted use,
distribution, and reproduction in any medium, provided you give
appropriate credit to the original author(s) and the source, provide a link
to the Creative Commons license, and indicate if changes were made.
PublishersNote Springer Nature remains neutral with regard to jurisdic-
tional claims in published maps and institutional affiliations.
References
Aust, U., Range, F.,Steurer, M., & Huber, L. (2008). Inferential reasoning
by exclusion in pigeons, dogs, and humans. Animal Cognition,
11(4),587597. doi:https://doi.org/10.1007/s10071-008-0149-0
Beran, M. J., Brandl, J., Perner, J., & Proust, J. (2012). Foundations of
Metacognition.Oxford: Oxford University Press.
Bräuer, J. & Belger, J. (2018). A ball is not a Kong: Odor representation
and search behavior in domestic dogs (Canis familiaris) of different
education. Journal of Comparative Psychology 132(2), 189199.
doi:https://doi.org/10.1037/com0000115.
Bräuer, J., & Call, J. (2011). The Magic Cup: Great Apes and Domestic
Dogs (Canis familiaris) Individuate Objects According to Their
Properties. Journal of Comparative Psychology, 125(3), 353361.
Bräuer, J., Call, J., & Tomasello, M. (2004). Visual perspective taking in
dogs (Canis familiaris) in the presence of barriers. Applied Animal
Behaviour Sciences, 88, 299317.
Bräuer, J., Kaminski, J., Riedel, J., Call, J., & Tomasello, M. (2006).
Making inferences about the location of hidden food: Social dog,
causal ape, Journal of Comparative Psychology, 120(1), 3847.
Browne, D. (2004). Do dolphins know their own minds? Biology and
Philosophy, 19(4), 633653.
Call, J. (2005). The Self and Other: A Missing Link in Comparative
Social Cognition. In H. S. Terrace & J. Metcalfe (Eds.), The missing
link in cognition: Origins of self-reflective consciousness (pp. 321
341). New York, NY, US: Oxford University Press.
Call, J. (2010). Do Apes Know that They Could Be Wrong? Animal
Cognition, 13(5),689700.
Call, J. (2012). Seeking information in non-human animals: weaving a
metacognitive web. In M. J. Beran, J. Brandl, J. Perner & J. Proust
(Eds.), Foundations of Metacognition (pp. 6275). Oxford: Oxford
University Press.
Call, J., & Carpenter, M. (2001). DoApes and Children Know What They
Have Seen? Animal Cognition, 4,207220.
Carruthers, P. (2008). Meta-cognition in Animals: A Skeptical Look.
Mind & Language, 23,5889. doi:https://doi.org/10.1111/j.1468-
0017.2007.00329.
Collier-Baker, E., Davis, J. M., & Suddendorf, T. (2004). Do Dogs (Canis
familiaris) Understand Invisible Displacement? Journal of
Comparative Psychology, 118(4), 421433.
Crystal, J. D., & Foote, A. L. (2009). Metacognition in animals.
Comparative Cognition & Behavior Reviews, 4,116.
Crystal, J. D., & Foote, A. L. (2011). Evaluating information-seeking
approaches to metacognition. Current Zoology, 57(4), 531542.
Erdohegyi, Á., Topál, J., Virányi, Z., & Miklósi, Á. (2007). Dog-Logic:
Inferential Reasoning in a Two-Way Choice Task and Its Restricted
Use. Animal Behaviour 74(4),725737.
Fiset, S. (2009). Evidence for Averaging of Distance from Landmarks in
the Domestic Dog. Behavioural Processes 81(3), 429438.
Fiset, S., Beaulieu, C., & Landry, F. (2003). Duration of Dogs' (Canis
familiaris) Working Memory in Search for Disappearing Objects.
Animal Cognition, 6(1),110.
Learn Behav
Fiset, S., Gagnon, S., Beaulieu, C. (2000). Spatial Encoding of Hidden
Objects in Dogs (Canis familiaris). Journal of Comparative
Psychology, 114, 315324.
Fiset, S., Landry, F., Ouellette, M. (2006). Egocentric Search for
Disappearing Objects in Domestic Dogs: Evidence for a
Geometric Hypothesis of Direction. Animal Cognition, 9,112.
Flavell, J. H. (1979). Metacognition and cognitive monitoring: A new
area of cognitive-development inquiry. American Psychologist,
34(10),906911.
Fleming, S. M., Dolan, R. J., & Frith, C. (2012). Metacognition: compu-
tation, biology and function, Philosophical Transactions of the
Royal Society B: Biological Sciences, 367,12801286.
Foote, A. L. & Crystal, J. D. (2007). Metacognition in the Rat. Current
Biology, 17(6),551555.
Gagnon, S., & Dore, F. Y. (1992). Search behavior in various breeds of
adult dogs (Canis familiaris): Object permanence and olfactory cues.
Journal of Comparative Psychology, 106(1), 5868.
Gazit, I., & Terkel, J. (2003). Domination of olfaction over vision in
explosives detection by dogs. Applied Animal Behaviour Science,
82(1), 6573.
Griffin, D. R. (2004). Significant Uncertainty Is Common in Nature.
Behavioral and Brain Sciences, 26,346.
Hampton, R. R. (2001). Rhesus monkeys know when they remember.
Proceedings of the National Academy of Sciences, 98(9),5359
5362.
Hampton, R. R. (2009). Multiple demonstrations of metacognition in
nonhumans: Converging evidence or multiple mechanisms?
Comparative Cognition & Behavior Reviews, 4, 1728.
Hampton, R. R., Zivin, A., & Murray, E. A. (2004). Rhesus Monkeys
(Macaca mulatta) Discriminate Between Knowing and Not
Knowing and Collect Information as Needed Before Acting.
Animal Cognition, 7,239246.
Hare, B., Brown, M., Williamson, C., & Tomasello, M. (2002). The
domestication of social cognition in dogs. Science, 298(5598),
16341636.
Hertzog, C. & Hultsch, D. F. (2000). Metacognition in adulthood and old
age. In F. I. M. Craik & T. A. Salthouse (Eds.), The handbook of
aging and cognition (pp. 417466). Mahwah: Lawrence Erlbaum
Associates.
Hofer, B. & Pintrich, P. (1997). The development of epistemological
theories: Beliefs about knowledge and knowing and their relation
to learning. Review of Educational Research,67(1), 88140.
Kornell, N. (2009). Metacognition in humans and animals. Current
Directions Psychological Science, 18,1115.
Kornell, N. (2014). Where is the "Meta" in Animal Metacognition?.
Journal of Comparative Psychology, 128(2), 143149.
Kornell,N.,Son,L.K.,&Terrace,H.S.(2007).Transferof
metacognitive skills and hint seeking in monkeys. Psychological
Science, 18,6471.
Marsh, H. I., & MacDonald, S. E. (2012). Orangutans (Pongo abelii)
Bplay the odds^: Information-seeking strategies in relation to cost,
risk, and benefit. Journal of Comparative Psychology,126(3), 263
278.
Marshall-Pescini, S., & Kaminski, J. (2014). The social dog: history and
evolution. In J. Kaminski, & S. Marshall-Pescini (Eds.), The social
dog: cognition and behavior (pp. 334). San Diego: Academic Press
Inc.
McMahon, S., Macpherson, K., & Roberts W. A. (2010). Dogs choose a
human informant: Metacognition in canines. Behavioural
Processes, 85,293298.
Metcalfe J. & Shimamura A. (1994). Metacognition: Knowing About
Knowing. MIT Press: Cambridge.
Miklosi, A. (2007). Dog Behaviour, Evolution, and Cognition (1st).
Oxford: Oxford University Press.
Miller H. C., Rayburn-Reeves R., & Zentall T. R. (2009). What Do Dogs
know about Hidden Objects? Behavioural processes, 81(3), 439
2446. doi:https://doi.org/10.1016/j.beproc.2009.03.018.
Müller C. A., Riemer S., Range F., & Huber L. (2014). The use of a
displacement device negatively affects the performance of dogs
(Canis familiaris) in visible object displacement tasks. Journal of
Comparative Psychology, 128(3), 240250. doi:https://doi.org/10.
1037/a0036032.
Nelson, T. O., & Narens, L. (1994). Why investigate metacognition? In J.
Metcalfe & A. P. Shimamura (Eds.), Metacognition: Knowing about
knowing (pp. 125). Cambridge: MIT Press.
Perdue, B. M., Evans, T. A., & Beranm M. J. (2018).Chimpanzees show
some evidence of selectively acquiring information by using tools,
making inferences, and evaluating possible outcomes. PLoS ONE,
13(4). doi:https://doi.org/10.1371/journal.pone.0193229.
Pongrácz, P., Miklosi, Á., Vida, V., & Csanyi, V. (2005). The pet dogs
ability for learning from a human demonstrator in a detour task is
independent from the breed and age. Applied Animal Behaviour
Science, 90,309323.
Pongrácz, P., Vida, V., Banhegyi, P., & Miklósi, Á. (2008). How does
dominance rank status affect individual and social learning perfor-
mance in the dog (Canis familiaris)?. Animal Cognition, 11,7582.
Rosati, A. G. & Santos, L. R. (2016). Spontaneous Metacognition in
Rhesus Monkeys. Psychological Science, 27(9), 11811191.
Smith, J. D. (2009). The Study of Animal Metacognition. Trends in
Cognitive Sciences, 13,389396.
Smith, J. D., Beran, M. J., Couchman, J. J., & Coutinho, M. V. C. (2008).
The comparative study of metacognition: Sharper paradigms, safer
inferences, Psychonomic Bulletin & Review, 15, 679691.
Smith, J. D., Beran, M. J., Redford, J., & Washburn, D. (2006).
Dissociating uncertainty responses and reinforcement signals in
the comparative study of uncertainty monitoring. Journal of
Experimental Psychology General, 135(2), 28297.
Smith, J. D., Shields, W., & Washburn D. (2003). The comparative psy-
chology of uncertainty monitoring and metacognition. Behavioral
and Brain Sciences, 26, 317373.
Smith, J. D., Zakrzewski, A. C., & Church, B. A. (2016). Formal models
in animal-metacognition research: the problem of interpreting ani-
malsbehavior. Psychonomic Bulletin Review, 23(5), 13411353.
Suda-King, C. (2008). Do Orangutans (Pongo pygmaeus) Know When
They Do Not Remember? Animal Cognition, 11, 2142.
Szetei, V., Miklósi, Á., Topál, J., Csányi, V. (2003). When Dogs Seem to
Lose Their Nose: An Investigation on the Use of Visual and
Olfactory Cues in Communicative Context Between Dog and
Owner. Applied Animal Behaviour Science, 83, 141152.
Topál, J., Gergely, G., Erdohegyi, A., Csibra, G., & Miklósi, A. (2009).
Differential sensitivity to human communication in dogs, wolves,
and human infants, Science, 325(5945), 126972.
Vining, A. Q. & Marsh, H. L. (2015). Information seeking in capuchins
(Cebus apella): a rudimentary form of metacognition? Animal cog-
nition, 18(3), 667681.
Vonk, J. & Leete, J. A. (2017). Carnivore concepts: Categorization in
carnivores "bears" further study. International Journal of
Comparative Psychology, 30,122.
Wallis, L. J., Virányi, Z., Müller, C. A., Serisier, S., Huber, L., & Range,
F. (2016). Aging effects on discrimination learning, logical reason-
ing and memory in pet dogs. AGE, 38(6). doi:https://doi.org/10.
1007/s11357-015-9866-x
Zohar, A. (1999). Teachersmetacognitive knowledge and the instruction
of higher order thinking, Teaching and Teacher Education, 15(4),
413429.
Learn Behav

Supplementary resources

This research hasn't been cited in any other publications.
  • Article
    Full-text available
    Metacognition refers to thinking about one’s thinking or knowing what one knows. Research suggests that this ability is not unique to humans and may be shared with nonhuman animals. In particular, great apes have shown behaviors on a variety of tasks that are suggestive of metacognitive ability. Here we combine a metacognitive task, the information-seeking task, with tool use and variable forms of initial information provided to chimpanzees to explore how informational states impact behavioral responses in these apes. Three chimpanzees were presented with an apparatus that contained five locations where food could be hidden. If they pointed to the correct location, they received the reward, but otherwise they did not. We first replicated several existing findings using this method, and then tested novel hypotheses. The chimpanzees were given different types of information across the experiments. Sometimes, they were shown the location of the food reward. Other times, they were shown only one empty location, which was not useful information. The chimpanzees also could use a tool to search any of those locations before making a selection. Chimpanzees typically used the tool to search out the location of the reward when they could not already know where it was, but they did not use the tool when they already had been given that information. One chimpanzee made inferences about the location of hidden food, even when that food was never shown in that location. The final experiment involved hiding foods of differing preference values, and then presenting the chimpanzees with different initial knowledge states (i.e., where the best food was located, where the less-preferred food was located, or where no food was located). All chimpanzees used the tool when they needed to use it to find the best possible item on that trial, but responded by choosing a location immediately when they did not need the tool. This finding highlights that their behavior was not the result of a simple rule following such as pointing to where any food had been seen.
  • Article
    There has been a growing interest in the cognitive skills of domestic dogs, but most current knowledge about dogs’ understanding of their environment is limited to the visual or auditory modality. Although it is well known that dogs have an excellent olfactory sense and that they rely on olfaction heavily when exploring the environment or recognizing individuals, it remains unclear whether dogs perceive odors as representing specific objects. In the current study, we examined this aspect of dogs’ perception of the world. Dogs were presented with a violation-of-expectation paradigm in which they could track the odor trail of one target (Target A), but at the end of the trail, they found another target (Target B). We explored (a) what dogs expect when they smell the trail of an object, (b) how they search for an object, and (c) how their educational background influences their ability to find a hidden object, by comparing family dogs and working dogs that had passed exams for police or rescue dogs. We found that all subjects showed a flexible searching behavior, with the working dogs being more effective but the family dogs learning to be effective over trials. In the first trial, dogs showed measurable signs of “surprise” (i.e., further searching for Target A) when they found Target B, which did not correspond to the odor of Target A from the trail. We conclude that dogs represent what they smell and search flexibly, which is independent from their educational background.
  • Article
    Full-text available
    Metacognition is the ability to think about thinking. Although monitoring and controlling one’s knowledge is a key feature of human cognition, its evolutionary origins are debated. In the current study, we examined whether rhesus monkeys (Macaca mulatta; N = 120) could make metacognitive inferences in a one-shot decision. Each monkey experienced one of four conditions, observing a human appearing to hide a food reward in an apparatus consisting of either one or two tubes. The monkeys tended to search the correct location when they observed this baiting event, but engaged in information seeking—by peering into a center location where they could check both potential hiding spots—if their view had been occluded and information seeking was possible. The monkeys only occasionally approached the center when information seeking was not possible. These results show that monkeys spontaneously use information about their own knowledge states to solve naturalistic foraging problems, and thus provide the first evidence that nonhumans exhibit information-seeking responses in situations with which they have no prior experience.
  • Book
    Metacognition has been described as 'thinking about thinking', and metacognition research has become a rapidly growing field of interdisciplinary research in the cognitive sciences. There have been major changes in this field recently, stimulated by the controversial issues of metacognition in nonhuman animals and in early infancy. Consequently the question what defines a metacognitive process has become a matter of debate: How should one distinguish between simple minds that are not yet capable of any metacognitive processing, and minds with a more advanced architecture that exhibit such a capacity? This is the basic question that the present volume addresses from three different perspectives: from an evolutionary point of view the book asks whether there is sufficient evidence that some non-human primates or other animals monitor their mental states and thereby exhibit a form of metacognition. From a developmental perspective we ask when children start to monitor, evaluate and control their own minds. And from a philosophical point of view the main issue is how to draw the line between cognitive and metacognitive processes, and how to integrate the different functions in which metacognition is involved into a single coherent picture of the mind. The foundations of metacognition - whatever they will turn out to be - have to be as complex as this pattern of connections we discover in its effects. This is the view we promote with this volume.
  • Article
    Full-text available
    In laboratory dogs, aging leads to a decline in various cognitive domains such as learning, memory and behavioural flexibility. However, much less is known about aging in pet dogs, i.e. dogs that are exposed to different home environments by their caregivers. We used tasks on a touchscreen apparatus to detect differences in various cognitive functions across pet Border Collies aged from 5 months to 13 years. Ninety-five dogs were divided into five age groups and tested in four tasks: (1) underwater photo versus drawing discrimination, (2) clip art picture discrimination, (3) inferential reasoning by exclusion and (4) a memory test with a retention interval of 6 months. The tasks were designed to test three cognitive abilities: visual discrimination learning, logical reasoning and memory. The total number of sessions to reach criterion and the number of correction trials needed in the two discrimination tasks were compared across age groups. The results showed that both measures increased linearly with age, with dogs aged over 13 years displaying slower learning and reduced flexibility in comparison to younger dogs. Inferential reasoning ability increased with age, but less than 10 % of dogs showed patterns of choice consistent with inference by exclusion. No age effect was found in the long-term memory test. In conclusion, the discrimination learning tests used are suitable to detect cognitive aging in pet dogs, which can serve as a basis for comparison to help diagnose cognition-related problems and as a tool to assist with the development of treatments to delay cognitive decline. Electronic supplementary material The online version of this article (doi:10.1007/s11357-015-9866-x) contains supplementary material, which is available to authorized users.
  • Article
    Over the last decade, studies on metacognition have been playing an increasingly prominent role in the field of animal cognition. Although a growing number of studies have documented responses consistent with metacognition, currently there is some debate about their proper interpretation. This chapter reviews the evidence that has accumulated in the last decade in the so-called information-seeking paradigm, which involves confronting subjects with two or more containers where food can be hidden. Researchers have manipulated several variables including the visibility of the baiting, the food type, and the time since the baiting took place. To get the food, subjects have to select the baited container but before they do so, they can, if they wish, look inside the container to verify its contents. Although the initial results could be explained as a result of random search, response competition, or perceiving anxiety rather than monitoring memories, recent findings have in turn challenged each of these alternative explanations. Information seeking in the great apes can be characterized as targeted (i.e. individuals donot search randomly), integrated (i.e. individuals can incorporate multiple types of information into their decision, including information derived by inference), and facultative (i.e. subjects can increase or decrease their searches depending on the information that they possess and the cost of searching and/or choosing wrongly). These findings, together with those from other metacognition paradigms, suggest that the great apes have some access to the causes of their uncertainty, and they can flexibly deploy means to remedy this situation.
  • Article
    Full-text available
    The recent explosion of studies on dogs' social behaviour and cognitive abilities are impressive, opening a new field of studies on a species that has economic, social, and emotional significance to humans across the globe. The origin of domestic dogs has been firmly established to be from an ancestor common to wolves, but the 'where, when, and how' of domestication, as well as the effects of this event on the dogs' mind and behaviour have engendered lively debates in journals and at conferences. In this chapter, we aim to introduce the reader of this book to some of the more salient and some of the more neglected aspects in the field. Hence, in the first part of this chapter (Section 1.1), we set dogs within the framework of their canine family, presenting some of the intriguing features that appear to set canids apart from other mammal families and that may have set the ground on which the wolf-human encounter took place. We also highlight areas where more research is needed because so little has been carried out to compare different canid species from a behavioural and cognitive perspective. In the second part (Section 1.2), we focus more on the dog-human story, summarising the archaeological evidence and genetic data helping us to draw the picture of the early history of men and dogs and presenting a brief overview of the different hypotheses put forward as regards the effects of domestication on dogs' social behaviour and cognition. Finally, in this section, we also outline some of the key issues that need to be addressed to assess the competing hypotheses and move the field of canine cognition forward. We conclude (in Section 1.3) by suggesting that dogs' sociality and their potentially 'special' socio-cognitive skills likely emerge both from the specific characteristics of their canid ancestry and the unique event of having encountered and started living alongside humans. We further present an overview of the chapters in this book, highlighting how contributions cover studies looking at both dogs' social behaviour and cognitive skills directed at both conspecifics and humans, because both are equally necessary for a well-rounded understanding of our four-legged companion.
  • Article
    Ongoing research explores whether animals have precursors to metacognition-that is, the capacity to monitor mental states or cognitive processes. Comparative psychologists have tested apes, monkeys, rats, pigeons, and a dolphin using perceptual, memory, foraging, and information-seeking paradigms. The consensus is that some species have a functional analog to human metacognition. Recently, though, associative modelers have used formal-mathematical models hoping to describe animals' "metacognitive" performances in associative-behaviorist ways. We evaluate these attempts to reify formal models as proof of particular explanations of animal cognition. These attempts misunderstand the content and proper application of models. They embody mistakes of scientific reasoning. They blur fundamental distinctions in understanding animal cognition. They impede theoretical development. In contrast, an energetic empirical enterprise is achieving strong success in describing the psychology underlying animals' metacognitive performances. We argue that this careful empirical work is the clear path to useful theoretical development. The issues raised here about formal modeling-in the domain of animal metacognition-potentially extend to biobehavioral research more broadly.
  • Article
    The main aim of this book is to provide a basis for a complete dog behavioural biology based on concepts derived from contemporary ethology. Thus, dog behaviour is viewed from both functional (evolution and ecology) and mechanistic and developmental points of view. The study of dogs is placed in a comparative context which involves comparison with their ancestors (wolves), as well as with humans with which dogs share their present environment. Instead of advocating a single theory which would explain the emergence of dogs during the last 20,000 years of human evolution, this book gives an overview of present knowledge which has been collected by scientists from various fields. It aims to find novel ways to increase our understanding of this complex evolutionary process by combining different methods originating from different scientific disciplines. This is facilitated by describing complementing knowledge provided by various field of science, including zooarchaeology, cognitive and comparative ethology, human-animal interaction, behaviour genetics, behavioural physiology and development, and behavioural ecology. This interdisciplinary approach to the study of dogs deepens our biological understanding of dog behaviour, but also utilizes this knowledge to reveal secrets to behavioural evolution in general, even with special reference to the human species.
  • Article
    Metacognition has been divided into information monitoring and control processes. Monitoring involves knowing that you know or do not know some information without taking corrective action. Control involves taking corrective action based on the knowledge that you know or do not know some information. In comparative metacognition, considerable attention has been paid toward critically assessing putative evidence for information monitoring in non-human animals. However, less attention has been paid toward critically evaluating evidence for control processes in animals. We briefly review a critique of information- monitoring in animals. Next, we apply these concepts to a number of studies that focus on information seeking in animals. The main type of evidence for control processes in animals come from tube tipping experiments. Before having the opportunity to search for the bait in these experiments, the subject sometimes observes opaque tubes being baited but is sometimes prevented from seeing the baiting. The observations that the subjects look more if baiting was not seen and are more accurate if baiting was seen have been taken as evidence for metacognition in information-seeking experiments. We propose simple alternative hypotheses that are sufficient to explain putative evidence for information seeking in animals without positing metacognition. The alternative explanation focuses on two relatively simple principles: First, an animal has a default "look before you go" response which supersedes random searches in space. Second, spatially guided behavior follows a default rule of "go where something good is." These principles can explain the results of tube tipping experiments without proposing metacognition.