Access to this full-text is provided by Springer Nature.
Content available from Scientific Reports
This content is subject to copyright. Terms and conditions apply.
1
Vol.:(0123456789)
Scientic Reports | (2022) 12:22529 | https://doi.org/10.1038/s41598-022-26991-5
www.nature.com/scientificreports
Breed dierences in social
cognition, inhibitory control,
and spatial problem‑solving
ability in the domestic dog (Canis
familiaris)
Saara Junttila
1*, Anna Valros
1, Katariina Mäki
2, Heli Väätäjä
3, Elisa Reunanen
4 &
Katriina Tiira
5,6
The extraordinary genetic and behavioural diversity of dog breeds provides a unique opportunity
for investigating the heritability of cognitive traits, such as problem‑solving ability, social cognition,
inhibitory control, and memory. Previous studies have mainly investigated cognitive dierences
between breed groups, and information on individual dog breeds is scarce. As a result, ndings are
often contradictory and inconsistent. The aim of this study was to provide more clarity on between‑
breed dierences of cognitive traits in dogs. We examined the performance of 13 dog breeds (N = 1002
dogs) in a standardized test battery. Signicant breed dierences were found for understanding of
human communicative gestures, following a human’s misleading gesture, spatial problem‑solving
ability in a V‑detour task, inhibitory control in a cylinder test, and persistence and human‑directed
behaviour during an unsolvable task. Breeds also diered signicantly in their behaviour towards an
unfamiliar person, activity level, and exploration of a novel environment. No signicant dierences
were identied in tasks measuring memory or logical reasoning. Breed dierences thus emerged
mainly in tasks measuring social cognition, problem‑solving, and inhibitory control. Our results
suggest that these traits may have come under diversifying articial selection in dierent breeds.
These results provide a deeper understanding on breed‑specic traits in dogs.
Cognitive abilities (such as learning, memory, inhibitory control, problem-solving, and social cognition) are
important traits in almost all aspects of an animal’s life, from nding food to cooperating with conspecics.
Despite this, the heritability of cognitive traits in non-human animals is still a largely unknown topic1. e
domestic dog (Canis familiaris), with its extraordinary genetic and phenotypic diversity, provides a unique oppor-
tunity for advancing our understanding of this subject. Behavioural variation between dog breeds is substantial2,
and there is evidence that many of these breed dierences are attributable to genetic factors3 (but see Morrill
etal.4).
Cognitive dierences between breeds have been investigated to some extent, but the results seem to be largely
contradictory and inconsistent. For example, studies have demonstrated dierences in breeds’ understanding of
human referential cues, such as pointing or gazing5–8, and genetic relatedness between breeds seems to account
for a substantial portion of variation in this trait9. Several other studies, however, have not been able to replicate
these results10–13. Similarly, studies have failed to nd any signicant dierences between breeds for logical rea-
soning ability13, inhibitory control13, or memory12,13, although Gnanadesikan etal.9 demonstrated that genetic
relatedness among breeds accounted for a signicant proportion of variation in these traits.
e diculty in interpreting previous results lies partly in the fact that most studies have grouped breeds
together, either based on original function or genetic relatedness. What further complicates interpretation
is that across studies, groups are oen composed of dierent breeds, and the criteria for categorizing breeds
OPEN
1Department of Production Animal Medicine, University of Helsinki, 00014 Helsinki, Finland. 2International
Partnership for Dogs, Helsinki, Finland. 3Digital Solutions, Lapland University of Applied Sciences, Jokiväylä 11C,
96300 Rovaniemi, Finland. 4Department of Finnish and Finno-Ugric Languages, University of Turku, 20014 Turku,
Finland. 5smartDOG Ltd, 05800 Hyvinkää, Finland. 6Department of Equine and Small Animal Medicine, University
of Helsinki, 00014 Helsinki, Finland. *email: saara.junttila@helsinki.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
2
Vol:.(1234567890)
Scientic Reports | (2022) 12:22529 | https://doi.org/10.1038/s41598-022-26991-5
www.nature.com/scientificreports/
dier greatly. Some authors have questioned the use of breed group classications based on the breeds’ original
purpose14–17, mainly because selection pressures may have changed drastically. For instance, Svartberg17 found no
associations between behavioural proles of breeds and their original purpose; instead, breed-typical behaviour
was correlated with the current use of the breed. Even grouping breeds based on their genetic relatedness is not
without its faults. For example, both Svartberg17 and Turcsán etal.16 used cluster analysis to group breeds based
on behavioural similarity, and found that these clusters corresponded poorly with the genetic categorization of
breeds. ese results suggest that when breeds are divided into groups, important dierences between breeds
may be missed.
Only a handful of studies have compared cognitive dierences between individual breeds instead of breed
groups7,8,12,14,18–20. Unfortunately, most of these have involved restricted sample sizes and a limited number
of breeds. Moreover, very little empirical research has targeted non-social cognitive traits such as memory,
inhibitory control, spatial problem-solving, and logical reasoning. e aim of our study was to provide a more
complete picture of cognitive dierences between dog breeds. We explored breed dierences not only in dogs’
socio-cognitive abilities, but also in several cognitive traits not involving a social aspect. In addition, we com-
pared breeds regarding their exploration in a novel space, greeting an unfamiliar person, and activity level, since
these may also be linked to cognitive performance21–24. Our large sample size allowed us to look at dierences
between individual breeds rather than having to resort to breed group classications. Our aim was to provide
a more comprehensive understanding of how breeds vary cognitively and behaviourally, which improves our
ability to predict how individual dogs are likely to behave. Our results may also further our understanding of
the heritability and genetics of cognition, both in humans and in other animals.
Methods
Subjects. A total of 2,352 adult dogs participated in a smartDOG™ test battery25 between March 2016 and
February 2022. Participating dogs were required to be interested in working for food, and to not be overly
aggressive to people. We limited the analysis to dogs between the ages of 1 and 8years, since cognitive traits
may not have fully developed in younger dogs26,27, while older dogs may experience cognitive decline28. We
included only dog breeds with a minimum of 40 individuals tested per breed. is resulted in a nal sample size
of 1,002 dogs representing 13 breeds, including one category consisting of mixed breeds. Any individual which
had parents belonging to dierent breeds was classied as mixed breed, except for the Labradoodle. Details of
participant dogs’ ages, sexes, and breeds are presented in Table1.
Participant dogs took part in a test battery involving multiple tests, but not all dogs had results for every test
included in this study (see Supplementary TableS1 for number of participant dogs in each test section). Most
of the results for each test came from the same dogs, with some results missing, mostly because 17% of the dogs
took part in a shortened version of the test battery (see Supplementary Materials S1 for details).
Most participant dogs were privately owned pet dogs. We did not have information on the training history of
most dogs, apart from 31 police dogs (see Suppl. TableS2). A large proportion of the pet dogs were actively used
in various dog sports (e.g., agility, scent work, obedience, etc.), based on discussions the testers have had with
participating owners. Most dogs (including the police dogs) lived inside the house with their owners.
Cognitive test battery. Dogs included in the study were participants in a commercial cognitive test bat-
tery (smartDOG™)25, which was developed by one of the authors (KT) based on previous scientic publications.
Tests were performed by eight trained female smartDOG licence testers (including KT) at testing sites across
Finland. We included 10 tests, seven of which measured cognitive traits and three of which measured behaviour.
Descriptions of included tests are outlined in Table2.
All tests involved solving various problems with food rewards. e owners were asked to bring the dog’s
favourite treats, which were then used as rewards. In some cases, a toy was also used, if the dog was more
Table 1. Number of dogs within each breed and their sexes and median ages.
Breed n Males (%) Females (%) Median age in years ± IQR
Australian Kelpie 41 20 (48.8) 21 (51.2) 3.0 ± 2.8
Australian Shepherd 49 23 (46.9) 26 (53.1) 2.9 ± 2.6
Belgian Shepherd Malinois 49 34 (69.4) 15 (30.6) 3.6 ± 2.6
Border Collie 106 48 (45.3) 58 (54.7) 2.6 ± 2.3
English Cocker Spaniel 60 29 (48.3) 31 (51.7) 2.7 ± 2.7
Finnish Lapphund 59 28 (47.5) 31 (52.5) 2.8 ± 3.4
German Shepherd 82 34 (41.5) 48 (58.5) 2.6 ± 2.9
Golden Retriever 74 35 (47.3) 39 (52.7) 2.8 ± 2.5
Hovawart 50 23 (46.0) 27 (54.0) 2.7 ± 2.9
Labrador Retriever 163 69 (42.3) 94 (57.7) 2.6 ± 2.3
Mixed Breed 149 54 (36.2) 95 (63.8) 3.3 ± 2.8
Shetland Sheepdog 48 19 (40.4) 28 (59.6) 3.3 ± 2.9
Spanish Water Dog 72 22 (30.6) 50 (69.4) 2.8 ± 2.5
Tot a l 1 002 438 (43.8) 563 (56.2) 2.9 ± 2.7
Content courtesy of Springer Nature, terms of use apply. Rights reserved
3
Vol.:(0123456789)
Scientic Reports | (2022) 12:22529 | https://doi.org/10.1038/s41598-022-26991-5
www.nature.com/scientificreports/
motivated by toys than food. e owner was advised not to feed their dog prior to the test to ensure food moti-
vation during the test. Fresh drinking water was available throughout the test. Testing took place indoors, and
the minimum size of the testing room was approximately 30 m2. Most oen only the tester, owner, and dog were
present during testing, but occasionally family members were present as observers. e dog was o the lead
throughout the test battery. A short break was included in the COGNITION test battery, during which the owner
was asked to walk the dog outside for a maximum of two minutes.
e order of tests was the same for all dogs to ensure consistency across subjects. Each test included several
trials. At the beginning of each trial, the tester always drew the dog’s attention, if necessary, by showing them a
treat and saying the dog’s name. A (Finnish) word, such as “ok” or similar, was used to indicate to the owner the
beginning of each trial. e owner was advised to release the dog upon hearing this word during the test battery.
Aer each trial, the owner called or moved the dog back to the starting position using the collar and held the
dog in place before releasing the dog for the next trial. e owner was advised to remain quiet and neutral when
the dog was focusing on a task, but they were allowed to praise the dog when the dog ate a treat. Only during the
V-detour was the owner allowed to encourage the dog to continue if the dog continuously stared at the owner.
Between test sections, the dog was free to walk around for about 1–2min.
A short description of each test is provided below, and more detailed information can be found in the Sup-
plementary Information S1. An example of each cognitive test can also be found in the supplementary videos S1.
Greeting. When the owner and dog rst entered the test room, the dog’s response to the tester (an unknown
person) was rated. e tester faced the dog while talking in a friendly voice and allowed the dog to approach
herself. If the dog was not fearful or aggressive, the tester bent down and attempted to stroke the dog. e tester
then continued stroking the dog as long as the dog was willing. is test lasted between 1–2min in total.
e tester rated the dog’s response according to a scale ranging from 1 to 7. For analysis purposes, these
scores were condensed into four groups: ‘fearful’ (score 1–3), ‘indierent’ (score 4–5), ‘friendly’ (score 6), and
‘overexcited’ (score 7). e largest category (‘friendly’) was used as the reference category.
Activity level. Aer the tester had greeted the dog, she attached a FitBark (FitBark Inc., Kansas City, MO,
USA) activity monitor to the dog’s collar or harness. FitBark generates ‘BarkPoints’ (from here on referred to as
‘activity points’), which represent a proxy measure for the average activity level of the dog during the test battery.
e monitor was kept on throughout testing and taken o when the test battery ended. Only dogs which had
taken part in the COGNITION test battery were analysed.
Exploration. Aer the FitBark had been attached, the dog was released and allowed to freely explore the test
room for approximately ve minutes. e tester rated the dog’s behaviour on a scale of 1–5. ese scores were
condensed into four groups: ‘low investigation’ (score 1–2), ‘moderate investigation’ (score 3), active investiga-
tion, walking (score 4), and very active investigation, running (score 5). e aim of this test was to measure the
extent to which the dogs were willing to investigate a novel environment, which is thought to be an indication of
Table 2. Description of each test included in the study, in the order in which they were presented to the dog
during the COGNITION test battery.
Tes t Trait(s) measured Short description
Greeting Greeting behaviour towards an unfamiliar person e dog’s behaviour upon rst greeting the tester was rated on a scale of 1–7
Activity level Activity level during the test e dog was supplied with a FitBark57 accelerometer for the duration of the
test battery, which provided an average activity level for the dog
Exploration Dog’s exploration of a novel environment e dog’s behaviour during its rst few minutes in the testing environment
was rated on a scale of 1–5
Cylinder test Inhibitory control, impulsivity Food is placed inside a transparent cylinder, and the dog has to inhibit their
response to reach directly for the food, and instead go around the barrier to
reach the food
Gestures Social cognition, dog’s ability to understand human communicative cues Object choice test: the human provides a gesture towards the bowl which
contains food. Gestures included: dynamic distal pointing, momentary distal
pointing, dynamic foot pointing, dynamic cross-forward pointing, and gaze
V-detour Inhibitory control, spatial problem-solving ability e dog is required to reach a food reward by detouring around a V-shaped
fence
Unsolvable task Social cognition, help-seeking behaviour, persistence, problem-solving
strategy
e dog is faced with a problem where food is visible but out of reach. e
dog’s response was measured as (a) human-directed behaviour, (b) independ-
ent behaviour, attempting to solve the task, and (c) abandoning the task;
engaging in behaviours not directed at a human or the apparatus
Logical reasoning Logical reasoning, ability to make inferences based on exclusion e dog can see that one of two bowls is empty and has to infer that the treat
is hidden under the other bowl
Memory vs gesture Dog’s tendency to choose based on human gesture vs visual information,
social cognition
e dog is required to choose between two bowls. e dog can see a human
placing a treat in one bowl, and the human gestures towards the other (empty)
bowl. e dog’s choice is then observed
Memory Spatial short-term memory e dog was required to remember the location of a food treat which was
hidden under one of three bowls, for an increasing duration of time (from 1
to 2.5min)
Content courtesy of Springer Nature, terms of use apply. Rights reserved
4
Vol:.(1234567890)
Scientic Reports | (2022) 12:22529 | https://doi.org/10.1038/s41598-022-26991-5
www.nature.com/scientificreports/
curiosity, boldness, and activity level, whereas remaining by the owner’s side was thought to be a possible indica-
tion of fear, anxiety, or neophobia.
The cylinder test. e cylinder test has been used extensively in animal cognition research to study impul-
sivity and inhibitory control29, more specically the motor inhibitory response. Inhibitory control is a core exec-
utive function, which involves suppression of a prepotent but inecient behaviour in favour of a more benecial
response. In this test, the dog is required to inhibit reaching directly for a visible food reward and instead detour
around the transparent barrier to reach the reward.
e owner and dog were positioned 2–2.5m away from the cylinder. During the training trials, the cylinder
was opaque, and the dog was taught to access a food reward from either of the open sides. e experimenter
stood directly behind the cylinder and placed a food reward inside while the dog was watching. e dog was then
released and allowed to eat the treat (Supplementary Video S1). Aer the dog fullled the learning criteria (4 out
of 5 trials without touching the outside of the cylinder), the test phase began. During the test phase the cylinder
was transparent, and the dog was required to inhibit reaching for the now visible food directly, and to instead
go around to the side of the cylinder to access the reward (Supplementary Video S2). Each trial during which
the dog touched the outside of the cylinder was marked as an incorrect trial. If the dog ate the food without rst
touching the outside of the cylinder, this was marked as a correct trial. Percentage of correct responses (out of a
total of 10 trials) was used as the response variable.
Human gestures. Dogs’ understanding of pointing and other human gestures is oen used as a measure of
social cognition30,31. e test battery included ve dierent gestures. Before the test phase, the dog was familiar-
ized with the test set-up over four training trials, during which the dog was simply taught that food is available
in either of the two bowls. (See Supplementary Video S3 for the training phase procedure.).
During the test phase, dogs took part in six trials for each gesture (30 trials in total). (See Supplementary
Video S4 for the test procedure for each of the ve gestures.) e percentage of correct responses (out of 30
trials) was calculated from all the gesture tests combined. e order of rewarded bowls was the same for each
dog, starting with the le side. Every other trial was rewarded to the right and every other to the le. In order
to ensure dogs were not learning this pattern, the percentage of correct responses from the nal 6 trials (gaze)
were compared to the percentage of correct responses from the rst 6 trials (dynamic distal pointing) using a
two-tailed paired t-test.
e trial always started with the dog and owner facing the tester, standing 2–2.5m away. e tester showed a
piece of food to the dog and placed it inside one of the bowls which she held in her hands. e bowls were placed
on the oor in front of the tester, 95cm apart from each other. Making sure the dog was watching, the tester
provided the gesture (these are described in more detail below). Aer this, the dog was released and allowed to
make a choice.
e procedure for each gesture was the same as described above. Each dog received the gestures in the same
order (6 trials each): (1) dynamic distal pointing: the tester pointed at the correct bowl with an extended arm
and index nger aer which the dog was released, and the tester kept her arm in the same position while the dog
made their choice, (2) momentary distal pointing: the tester pointed at the correct bowl with an extended arm
and index nger for a duration of 2s, aer which the tester lowered her arm and the dog was released while the
tester’s arms were ush at her sides, (3) dynamic proximal foot pointing: the tester placed the tip of her foot on
the ground directly behind the correct bowl, and the dog was released while the tester remained in this position,
(4) dynamic cross-forward pointing: THE tester used her contralateral arm to point at the correct bowl while
rotating her shoulders in the same direction, sustaining this position while the dog made their choice, (5) gaze:
the tester alternated her gaze between the dog and the bowl three times, and the dog was released while the tester
maintained her gaze on the correct bowl.
V‑detour. e V-detour has been used in canine cognitive research to investigate spatial problem-solving
ability29. e dog has to detour around a transparent V-shaped fence to access a food reward which is placed on
the other side (Supplementary Video S5). Since the dog is required to move away from the visible treat to access
it, the task is also oen considered to measure inhibitory control.
e V-shaped fence was made out of compost fence panels, which were attached at an angle of approximately
70°. e owner and dog waited about 40cm away from the intersecting angle of the V-shaped fence. e tester
showed the dog several treats (or a toy) and placed them inside the fence while standing outside the fence. e
owner released the dog while the dog was looking at the food. e number of seconds taken to solve the task
was measured using a stopwatch. If the dog was not able to solve the task within 3min, the trial was terminated.
Unsolvable task. e unsolvable task has been used in canine cognitive research to assess persistence,
problem-solving behaviour, human-directed communication, and social cognition32. In our version of this task,
the dog was presented with four solvable trials, aer which the task became impossible to solve (Supplementary
Video S6).
e test involved a plastic or wooden box with a transparent lid, which had small holes to allow the dog to
smell the food inside. During the four training trials, the dog was taught to access a treat placed inside the box by
moving the plastic lid o. e diculty of the trials increased gradually. Once the dog was successful in opening
the lid, the test trial was begun. With the dog watching, the tester placed several treats inside the box. She then
secured the lid in place so that it could not be opened, aer which the owner released the dog. Both owner and
experimenter remained quiet and still, looking only at the box during the subsequent 2-min period. e tester
measured the time the dog spent on each behaviour: (a) independent problem-solving: attempting to solve the
Content courtesy of Springer Nature, terms of use apply. Rights reserved
5
Vol.:(0123456789)
Scientic Reports | (2022) 12:22529 | https://doi.org/10.1038/s41598-022-26991-5
www.nature.com/scientificreports/
task independently, (b) human-directed behaviour: initiating social contact with either the tester or the owner,
or (c) abandoning the task: not directing their behaviour toward the task or a person.
ree variables were used to measure the dogs’ behaviour during the unsolvable task. (1) Complete inde-
pendence (comparing dogs which did not spend any time on human-directed behaviour to those which spent
any amount of time on human-directed behaviour), (2) Percentage of time spent on human-directed behaviour,
and (3) Abandoning task (comparing dogs which abandoned the task to those which never abandoned the task).
Logical reasoning. is test aimed to measure the dog’s ability to make inferences based on exclusion33.
e dog could see that one of two bowls was empty, and it had to infer that the treat was hidden under the other
bowl. e tester sat on a chair or on the oor, about 1m away from the dog. Two opaque bowls were placed
upside down in front of the tester at arm’s length, one on the le and one on the right side. In each trial, a piece
of food was placed under one of the bowls. e order of baiting was the same for each dog; rst the le-hand
bowl was baited, aer which every other trial was baited to the right and every other trial to the le. e dog
rst took part in a training phase consisting of four trials, the aim of which was to familiarize the dog with the
set-up and learn that there is always food hidden under one of the bowls. (See Supplementary Video S7 for the
procedure of the training phase.).
When the dog correctly performed all four training trials, the test phase was initiated. is consisted of six
trials. e tester held a writing pad as a visual barrier in front of the le-hand bowl and placed a treat under the
bowl. She then placed the writing pad in front of the right-hand bowl and sham-baited the bowl. e writing
pad was then removed, and the tester lied the empty bowl up about 30cm above the oor, keeping it there for
about 1–2s while the dog was watching. At the same time, the tester held her other hand on top of the baited
bowl. e bowl was then placed back down, the tester placed her hands on her lap, and the dog was released. If
the dog approached the correct bowl, they were allowed to eat the treat. e trial was then repeated another ve
times, every other trial rewarded on the right and every other on the le. (See Supplementary Video S8 for the
procedure of the test phase.).
Two variables were used to measure logical reasoning of dogs: (a) percentage of correct responses and (b)
understanding of the task. For percentage of correct responses, dogs were divided into three groups: (1) 0–50%
of trials correct, (2) 51–82% of trials correct, and (3) 83–100% of trials correct. e tester also evaluated whether
the dog had understood the task and was successfully making inferences based on exclusion. e dog was con-
sidered to have understood the task if at least 4 out of 6 trials were correct or the nal 2–3 trials were correct.
Memory vs gesture. Previous studies have shown that dogs are more likely to choose an empty bowl out of
two choices if a human points towards it, even when they have seen that the other bowl has food in it34–37. Similar
to the gesture tests, this test aims to measure social cognition. e procedure for this test was also similar to the
gesture test, but instead of gesturing towards the baited bowl, the tester gestured towards the empty bowl. e
tester used the gesture which the dog had been most successful with—this was oen the dynamic distal point.
e bowls remained on the oor throughout the test, and the treat was placed in one of the bowls while the dog
was watching. e test consisted of two trials, with the rst trial always baited to the le and the second trial
baited to the right. (See Supplementary Video S9 for the procedure.) Dogs were divided into two groups: those
which chose the baited container on both trials (relied on their memory), and those which chose the empty
container on 1–2 trials (relied on the human’s gesture).
Memory. e aim of this test was to measure the duration of the dogs’ short-term memory38. ree identical
opaque bowls were placed upside down on the oor in a straight line, about 1m apart from each other. A piece
of food (or a toy) was placed under one of the bowls in each trial. e owner sat on a chair 3m away from the
middle bowl with the dog in front of her. During 7 training trials the dog learned to nd a food reward from
under one of the three bowls. Only a very short delay between hiding the food and releasing the dog was in place
during this phase.
Once the dog had passed the training phase, the four test trials began. e tester placed a treat under one
bowl while the dog was watching, aer which a visual barrier was placed in front of the dog. e tester waited
an increasing duration of time behind the owner and the dog (1st trial: 1min, 2nd trial: 1.5min, 3rd trial: 2min,
4th trial: 2.5min). Aer the waiting period, the tester removed the barrier, and the dog was released and allowed
to make a choice. e order of baiting the bowls was the same for each dog: trial 1: middle, trial 2: le, trial 3:
right, and trial 4: middle. (See Supplementary Video S11 for the procedure of the test phase.) Number of correct
trials (out of a total of 4) was calculated for each dog. Since a very small number of dogs had a score of 0, these
dogs were combined into a group with dogs that had a score of 1.
Data analysis. All statistical analyses were performed using IBM SPSS Statistics Version 28. An alpha level
of 0.05 was used for all statistical tests. Multiple and logistic regression analyses were used with the enter method
to analyse dierences between breeds for each variable. For ordinal variables, cumulative odds ordinal logistic
regression with proportional odds was used. In each model, we included age and sex of the dogs as control vari-
ables, since previous research suggests these may aect various measures of cognition and behaviour39–42. Each
model therefore included the predictors breed, age, and sex (apart from success in the V-detour, which only
included breed as a predictor). e Labrador Retriever, one of the most popular breeds worldwide, was used
as the reference breed, since it had the largest number of individuals tested out of all included pedigree breeds.
For the greeting variable, the data failed the assumption of proportional odds according to the full likelihood
ratio test, and therefore a multinomial logistic regression was conducted instead of ordinal logistic regression. For
success in the V-detour task, Fisher’s Exact test using the Monte Carlo method was run with 10,000 simulations,
Content courtesy of Springer Nature, terms of use apply. Rights reserved
6
Vol:.(1234567890)
Scientic Reports | (2022) 12:22529 | https://doi.org/10.1038/s41598-022-26991-5
www.nature.com/scientificreports/
since an insucient number of individuals failed to solve the task for a binary logistic regression to be reliable. For
latency (s) to solve the V-detour task, the dependent variable was log-transformed to normalize its distribution.
Since the cylinder test variable ‘percentage of correct responses’ was negatively skewed, a reect and square root
transformation was applied (i.e., each data point was subtracted from the maximum value plus 1, and a square
root transformation was then applied to these scores). Due to the transformation, the variable was inverted, and
therefore named ‘percentage of incorrect responses’ to aid with interpretability. For all transformed variables, the
original values are included in the gures, whereas the transformed variables are reported in the tables and text.
Ethical statement. We conrm that the procedures comply with national and EU legislation. Research was
performed in accordance with the Declaration of Helsinki. e study was approved by the University of Helsinki
Viikki Campus Research Ethics Committee (Statement 12/2021, accepted on 18/05/2021). Before participating
in the cognitive test battery, each dog owner gave informed written consent for using their dogs’ test results in
research. Reporting of results follows the recommendations of the ARRIVE guidelines. Informed consent was
given by each subject for publication of identifying images/videos in an online-access publication.
Results
Greeting. A total of 934 dogs had results for the greeting test. Out of these, 12.8% were fearful when meeting
the unknown person, 28.2% were indierent, 41.1% responded in a friendly manner, and 11.1% of dogs had an
overexcited greeting.
e multinomial regression model signicantly predicted greeting scores over and above the intercept-
only model (χ2(42) = 174.1, p < 0.001). e model explained 18.8% (Nagelkerke R2) of the variance in greeting
behaviour. Sex did not have a signicant eect on the prediction of greeting (χ2(3) = 2.65, p = 0.45), but age did
(χ2(3) = 20.77, p < 0.001). An increase in age was associated with a decrease in the odds of having an overexcited
greeting compared to a friendly greeting (which was used as the reference group) (χ2(1) = 9.94, p = 0.002).
Breed had a signicant eect on the prediction of greeting score (χ2(36) = 152.86, p < 0.001). See Fig.1a for the
proportions of greeting scores for each breed, and Supplementary TableS3 for parameter estimates for each
greeting score.
Activity level. A total of 759 dogs had activity level results available. Dogs received a mean of 29.4 FitBark
activity points, ranging from 9.9 to 93.4 activity points. e multiple regression model signicantly predicted
activity levels of dogs (F (14) = 7.91, p < 0.001, adj. R2 = 0.12). ere were 10 outliers but none of these were inu-
ential according to Cook’s Distance. Age and sex did not signicantly predict activity level in dogs aged 1–8years
(age β = −0.04, p = 0.23, sex β = −0.03, p = 0.42). Six breeds diered signicantly from the Labrador Retriever
(which was used as the reference breed for all analyses). See Fig.1b for mean activity points for each breed, and
Supplementary TableS4 for parameter estimates.
Exploration. A total of 820 dogs had results for exploration in a novel environment. Out of these, 8.3%
showed low investigation, 15.8% showed moderate investigation, 39.7% showed active investigation (walking),
and 36.3% showed very active investigation (running).
e ordinal regression model signicantly predicted exploration scores over and above the intercept-only
model (χ2(14) = 61.98, p < 0.001). e model explained 8% (Nagelkerke R2) of variance in exploratory behav-
iour. Age and sex did not have a signicant eect on the prediction of exploration in dogs aged 1–8years (age
χ2(1) = 1.9, p = 0.17; sex χ2(1) = 0.05, p = 0.82), but breed did (χ2(12) = 57.68, p < 0.001). See Fig.1c for percentage
of each exploration score for each breed, and Supplementary TableS5 for parameter estimates.
Cylinder test. A total of 992 dogs took part in the cylinder test. e median success rate of dogs was 80%
(IQR = 30%). e multiple regression model signicantly predicted the percentage of incorrect trials in the cylin-
der test (F (14, 961) = 6.96, p < 0.001, adj. R2 = 0.08). No outliers were detected. Age was a signicant predictor of
performance in the cylinder test, β = 0.13, p < 0.001, with increasing age associated with an increasing percentage
of incorrect trials (in dogs aged 1–8-years old). Sex was also a signicant predictor of cylinder test performance,
β = −0.09, p = 0.004, with females making fewer mistakes than males. Five breeds had signicantly lower per-
centages of incorrect trials compared to the Labrador Retriever. See Fig.2 for the mean percentage of correct
trials for each breed, and Supplementary TableS6 for parameter estimates.
Human gestures. A total of 831 dogs took part in each of the ve gesture tests. e mean percentage of
correct trials for these dogs was 79% (SD = 12.9%). A two-tailed paired samples t-test showed that there was no
signicant change from the rst 6 object choice trials to the last 6 trials, with a mean decrease of 0.84% in cor-
rect responses (95% CI [−2.65, 0.96], t(831) = −0.92, p = 0.36, d = -0.03). erefore, it is unlikely that dogs were
learning the pattern for baiting the bowls.
e multiple regression model signicantly predicted the percentage of correct trials for the gesture tests
(F (14, 804) = 3.41, p < 0.001, adj. R2 = 0.04). ere were two outliers, but these were not inuential according
to Cook’s Distance. Age and sex did not signicantly predict performance in dogs aged 1–8years (age β = 0.03,
p = 0.45; sex β = −0.06, p = 0.12). ree breeds had signicantly lower scores than the Labrador Retriever. See
Fig.3 for the mean percentage of correct trials for each breed, and Supplementary TableS7 for parameter
estimates.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
7
Vol.:(0123456789)
Scientic Reports | (2022) 12:22529 | https://doi.org/10.1038/s41598-022-26991-5
www.nature.com/scientificreports/
Figure1. Breed dierences of greeting unfamiliar person, activity level, and exploration in novel environment. Signicant
P-values (Bonferroni-corrected) are indicated with asterisks: ***p ≤ 0.001, **p ≤ 0.01, *p ≤ 0.05. e Labrador Retriever was
used as the reference breed. (a) Percentage of dogs within each breed receiving each of the four greeting scores (n = 934).
Breeds have been ordered based on odds ratios, from most fearful (highest odds ratio for score 1–3) on the le to least fearful
(lowest odds ratio for score 1–3) on the right. P-values have been indicated for fearful response using asterisks. (b) Mean
activity level scores (measured in FitBark activity points) for each breed (N = 759). Error bars represent 95% condence
intervals. Breeds are ordered based on B-values, with lowest activity levels on the le and highest activity levels on the right.
(c) Percentage of dogs within each breed with each exploration score (N = 820). Breeds are ordered based on B-values, with
lowest exploration on the le and highest exploration on the right. Score 1–2: low investigation (stays close to owner). Score 3:
moderate investigation. Score 4: active investigation (walking). Score 5: very active investigation (running).
Content courtesy of Springer Nature, terms of use apply. Rights reserved
8
Vol:.(1234567890)
Scientic Reports | (2022) 12:22529 | https://doi.org/10.1038/s41598-022-26991-5
www.nature.com/scientificreports/
Figure2. Mean percentage of correct trials for each breed in the cylinder test are presented, using the
untransformed, original data (n = 992). Error bars represent 95% condence intervals. Breeds are ordered
based on B-values, with lowest success (low inhibitory control) on the le and highest success (high inhibitory
control) on the right. e Labrador Retriever was used as the reference breed. Signicant P-values (Bonferroni-
corrected) are indicated with asterisks: ***p ≤ 0.001, **p ≤ 0.01, *p ≤ 0.05.
Figure3. Mean percentage of correct trials in the gesture tests within each breed (n = 831). Error bars represent
95% condence intervals. Breeds have been ordered based on B-values, with lowest success on the le and
highest success on the right. e Labrador Retriever was used as the reference breed. Signicant P-values
(Bonferroni-corrected) are indicated with asterisks: ***p ≤ 0.001, **p ≤ 0.01, *p ≤ 0.05.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
9
Vol.:(0123456789)
Scientic Reports | (2022) 12:22529 | https://doi.org/10.1038/s41598-022-26991-5
www.nature.com/scientificreports/
V‑detour. A total of 993 dogs took part in the V-detour test. Out of these, 88.5% successfully solved the
V-detour within 3min. ere was no signicant dierence in the proportions of dogs within each breed suc-
ceeding compared to those failing (p = 0.13, 99% CI [0.12, 0.14]). See Fig.4a for the proportion of successful
dogs in each breed.
e median time for solving the V-detour task (when unsuccessful dogs were removed from the analysis)
was 12s (IQR = 29s). e multiple regression model signicantly predicted the time taken to solve the task
(n = 863, F (14, 848) = 2.52, p = 0.002, adj. R2 = 0.02). No outliers were detected. Age and sex did not signicantly
Figure4. Performance of dogs during the V-detour task. e Labrador Retriever was used as the reference
breed. Signicant P-values (Bonferroni-corrected) are indicated with asterisks: ***p ≤ 0.001, **p ≤ 0.01, *p ≤ 0.05.
(a) Percentage of dogs within each breed succeeding vs failing in the V-detour task (n = 993). Breeds have been
ordered based on B-values, with lowest succeeding breeds on the le and highest succeeding breeds on the right.
(b) Mean latency (s) to solve the V-detour for each breed are presented using the untransformed, original data
(n = 863). Error bars represent 95% condence intervals. Breeds are ordered based on B-values, with lowest
succeeding breeds (took a long time to solve the task) on the le and highest succeeding breeds (solved the task
quickly) on the right.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
10
Vol:.(1234567890)
Scientic Reports | (2022) 12:22529 | https://doi.org/10.1038/s41598-022-26991-5
www.nature.com/scientificreports/
predict performance in the V-detour in dogs aged 1–8years (age β = 0.05, p = 0.14; sex β = −0.05, p = 0.12). e
Border Collie had a signicantly lower predicted mean log score compared to the Labrador Retriever (β = −0.14,
p = 0.01). See Fig.4b for mean time taken to solve the task for each breed, and Supplementary TableS8 for
parameter estimates.
Unsolvable task. A total of 969 dogs took part in the unsolvable task. Only 57 (5.7%) of these were com-
pletely independent during the unsolvable task (i.e., they spent 0% of their time on human-directed behav-
iour). When comparing these dogs to those which spent over 0% of their time on human-directed behav-
iour, the binomial logistic regression model was statistically signicant (χ2(14) = 34.28, p = 0.002). e model
explained 9.7% (Nagelkerke R2) of the variance. Age was not a signicant predictor of independence in dogs
aged 1–8years (χ2(1) = 0.59, p = 0.44), but females had lower odds of being completely independent compared
to males (χ2(1) = 5.22, p = 0.02). Breed was also a signicant predictor of independence (χ2(12) = 23.13, p = 0.03),
but no breed was signicantly dierent from the Labrador Retriever. See Fig.5a for the percentage of completely
independent dogs within each breed, and Supplementary TableS9 for parameter estimates.
When completely independent dogs had been removed from the analysis, the median amount of time spent
on human-directed behaviour was 54.2% (65s out of a total of 120s) (IQR = 40%). e multiple regression model
signicantly predicted the percentage of time spent on human-directed behaviour (n = 912, F (14, 885) = 5.7,
p < 0.001, adj. R2 = 0.07). No outliers were detected. Age signicantly predicted time spent on human-directed
behaviour (β = 0.09, p = 0.009), with older dogs spending a larger proportion of their time on human-directed
behaviour. Sex did not signicantly predict time spent on human-directed behaviour (β = 0.002, p = 0.95). e
Golden Retriever and Australian Kelpie had signicantly higher scores than the Labrador Retriever. See Fig.5b
for the percentage of time spent on human-directed behaviour for each breed, and Supplementary TableS10
for parameter estimates.
A total of 37.2% of dogs spent over 0% of their time abandoning the task. When comparing these dogs to those
which persisted with the task (0% of time spent on abandoning the task), the binomial logistic regression model
was statistically signicant (χ2(14) = 40.89, p < 0.001). e model explained 5.7% (Nagelkerke R2) of the variance.
Age was a signicant predictor (χ2(1) = 16.34, p < 0.001), with older dogs being less likely to abandon the task.
Breed was a signicant predictor of abandoning the task, χ2(12) = 24.16, p = 0.02, but sex was not (χ2(1) = 0.32,
p = 0.57). No breed signicantly diered from the Labrador Retriever. See Fig.5c for the percentage of dogs within
each breed abandoning the task, and Supplementary TableS11 for parameter estimates.
Logical reasoning. A total of 826 dogs took part in the logical reasoning task. Out of these, 33.4% were
rated as having understood the logical reasoning task. When comparing dogs which understood the task to
dogs which failed to understand the task, the binomial logistic regression model was not statistically signicant
(χ2(14) = 20.95, p = 0.1). See Supplementary Fig.S5a for the percentage of dogs within each breed which under-
stood the task.
For percentage of correct trials, 60.9% of dogs were correct on 0–50% of trials, 22.6% of dogs were correct
on 51–82% of trials, and 16.5% of dogs were correct on 83–100% of trials. e ordinal regression model did
not signicantly predict success in the logical reasoning task above the intercept-only model, χ2(14) = 20.03,
p = 0.13. See Supplementary Fig.S5b for the percentage of dogs within each breed with dierent rates of success.
Gesture vs memory. A total of 823 dogs took part in the gesture vs memory test. Out of these, 40.5% fol-
lowed the human gesture to the baited container in at least one trial. e binomial logistic regression model
for trusting the human gesture was statistically signicant (χ2(14) = 29.23, p = 0.01). e model explained 4.8%
(Nagelkerke R2) of the variance in the gesture vs memory test. Age and sex were not signicant predictors in
dogs aged 1–8years, but breed was a signicant predictor of trusting the gesture, χ2(12) = 25.42, p = 0.01. See
Fig.6 for percentage of dogs within each breed which trusted the human gesture vs their own memory, and Sup-
plementary TableS12 for parameter estimates.
Memory. A total of 822 dogs took part in the memory test. Out of these, 14.1% were correct on 0–1 trials,
30.4% were correct on two out of four trials, 35.3% were correct on three trials, and 20.2% were correct on all
four trials. e ordinal logistic regression model did not signicantly predict the number of correct trials in the
spatial memory task above the intercept-only model (χ2(14) = 19.26, p = 0.16). See Supplementary Fig.S6 for the
percentage of dogs within each breed with dierent success rates.
Discussion
We found signicant dierences between individual dog breeds for ve of the seven cognitive tests included in
the test battery. Breed dierences were found for measures of social cognition, persistence, inhibitory control,
and spatial problem-solving ability. Dierences were also evident for activity level, greeting of an unfamiliar
person, and exploration of a novel environment. In contrast, no breed dierences were found for short-term
memory or logical reasoning.
Both inhibitory control and social cognition are likely to be especially important traits during articial selec-
tion of dog breeds, both historically and in the present day. For example, inhibitory control may be a valued
trait in herding dogs, which are required to inhibit their predatory responses. e Border Collie and Australian
Shepherd were among the highest-scoring breeds in the cylinder test, indicating high inhibitory control. In
contrast, the Malinois and German Shepherd were some of the lowest-scoring breeds. ese breeds are oen
used in working roles requiring high responsiveness, which is oen associated with low inhibitory control and
high impulsivity43,44. Human-directed behaviour and socio-cognitive abilities may be highly valued in pet dogs
Content courtesy of Springer Nature, terms of use apply. Rights reserved
11
Vol.:(0123456789)
Scientic Reports | (2022) 12:22529 | https://doi.org/10.1038/s41598-022-26991-5
www.nature.com/scientificreports/
Figure5. Performance of breeds in the unsolvable task (n = 969). e Labrador Retriever was used as the reference breed.
Signicant P-values (Bonferroni-corrected) are indicated with asterisks: ***p ≤ 0.001, **p ≤ 0.01, *p ≤ 0.05. (a) Percentage
of dogs within each breed belonging to the human-directed compared to the completely independent groups during the
Unsolvable Task. Breeds have been ordered based on odds ratios, with most independent breeds (most likely to spend 0%
of their time on human-directed behaviour) on the le and least independent breeds (least likely to spend 0% of their time
on human-directed behaviour) on the right. (b) Mean percentage of time (out of a total of 120s) spent on human-directed
behaviour during the unsolvable task for each breed. Error bars represent 95% condence intervals. Breeds are ordered based
on B-values, with lowest amount of time spent on human-directed behaviour on the le and highest amount of time on the
right. (c) Percentage of dogs within each breed abandoning the task (over 0% of time spent on abandoning task) vs persisting
(0% of time spent on abandoning task) with the unsolvable task. Breeds are ordered based on odds ratios, with the most
persistent breeds (least likely to abandon task) on the le and the least persistent breeds (most likely to abandon task) on the
right.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
12
Vol:.(1234567890)
Scientic Reports | (2022) 12:22529 | https://doi.org/10.1038/s41598-022-26991-5
www.nature.com/scientificreports/
and breeds required to work closely with people, such as herding dogs and retrievers45,46. In line with this, the
Kelpie, Golden Retriever, Australian Shepherd, and Border Collie spent the largest proportion of their time on
human-directed behaviour during the unsolvable task. In contrast, the ability to work independently may be
important for various working dogs, such as detection dogs47,48. In our study, the two breeds which were most
likely to be completely independent during the unsolvable task (spending 0% of their time on human-directed
behaviour) were the German Shepherd and Malinois.
In conclusion, many of our results seem to reect the breeds’ original or current function, but several breed
dierences could not be easily explained by breed function alone. For example, many of the breeds in our study
belonged to the herding group, but there was oen large variation in their results in several cognitive tests. e
Finnish Lapphund received the lowest score in the gesture tests, whereas the Kelpie and Malinois had the highest
scores for this test, despite all three being herding breeds. Similarly, during the unsolvable task, the Australian
Shepherd was the least likely breed to abandon the task, whereas the Kelpie was among the breeds most likely to
abandon the task. It seems evident that breeds can vary behaviourally from each other even within their breed
groups, since dierent traits may have been (both intentionally and unintentionally) selected for in dierent
breeds, despite the breed group they belong to.
Our results replicate the ndings of various studies which have investigated dierences between individual
breeds in the V-detour task19, understanding of human gestures7,8, and the unsolvable task8,18. In contrast, most
studies which have failed to nd breed dierences have compared breed groups10,11,13,19,49,50. It is possible that
potential dierences between breeds may have been missed in these studies, since (as shown by our current
study) behavioural variation within breed groups can be substantial. For example, the Golden Retriever diered
signicantly from the Labrador Retriever in the unsolvable task and the gesture test, even though both belong
to the retriever group. Our ndings therefore highlight the importance of investigating behavioural dierences
between individual breeds rather than only relying on breed group categorizations.
We found no signicant breed dierences for logical reasoning or memory, and these results seem to be in
line with previous studies12,13. However, it is important to note that even though we did not nd dierences
between these specic breeds, this does not mean they do not exist between other breeds. For example, we did
not include ancient breeds, hounds, or terriers in our analysis. erefore, more research on individual breed
dierences on these cognitive traits is warranted.
It is important to note that the population used in this study is not representative of the entire dog population,
since ndings may dier across countries and cultures. In addition, since we used a commercial test battery, only
certain types of owners and dogs were inevitably self-selected. e breeds included in our study were mostly
breeds used in dog sports, and most owners were active in various dog sports or competitions. Even though this
limits the generalization of our results, it is noteworthy that signicant breed dierences emerged despite the
similarity of participant dogs and their training histories.
Figure6. Percentage of dogs within each breed which trusted the human gesture compared to those which
trusted their own memory (n = 823). Breeds have been ordered based on odds ratios, with breeds which are
most likely to trust their own memory on the le and breeds which are most likely to trust the human’s gesture
on the right. Signicant P-values (Bonferroni-corrected) are indicated with asterisks: ***p ≤ 0.001, **p ≤ 0.01,
*p ≤ 0.05. e Labrador Retriever was used as the reference breed.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
13
Vol.:(0123456789)
Scientic Reports | (2022) 12:22529 | https://doi.org/10.1038/s41598-022-26991-5
www.nature.com/scientificreports/
ere isa possibility that the dierences seen in our study were not based on genetic dierences between
breeds but rather due to variation in life experiences or training, since these have also been found to inuence
behaviour in cognitive tests8,13,49,51,52. Unfortunately, we were not able to control for the possible eects of train-
ing, environment, life experiences, or background of the dogs, since this information was not available to us.
erefore, the extent of their possible eects on breed dierences in our study is not known; however, this is
something we will investigate next. Previous studies suggest that environmental eects are unlikely to be the
only explanation for breed dierences. Genetic relatedness between dog breeds has been shown to account for
a substantial portion of variation in understanding human gestures and inhibitory control9. In addition, when
training history of participating dogs has been controlled for, signicant breed dierences have still been found
in the unsolvable task18,27,53, problem-solving tasks20,49, inhibitory control9, and point-following ability9,54. How-
ever, this topic warrants further investigation to determine the extent of heritable dierences between breeds as
opposed to the eect of life experiences.
It could be argued that the large proportion of police dogs within the Malinois breed (45%) could have biased
breed dierences in our study. However, our sample also consisted of other working dogs as well as several pet
dogs which participated extensively in training and competitions, but we did not have information about these
individuals. erefore, it would not have made much sense to exclude police dogs. We nevertheless analysed
the cylinder test, V-detour, and unsolvable task (which were the only tests which police dogs took part in) with
police dogs excluded. As a result, breed dierences in the proportions of completely independent dogs in the
unsolvable task became non-signicant, and the proportion of completely independent Malinois dropped from
14.6 to 7.7%. For all other variables breed dierences were still signicant and the scores of Malinois changed
only slightly (results not shown). erefore, it seems that the inclusion of police dogs only aected complete
independence during the unsolvable task but did not inuence other results to a great extent.
Another factor which could aect breed dierences is the brain size and skull shape of the breeds13,54–56.
However, the breeds included in our study all had similar skull shapes, and there were no extremes in body size.
e Shetland Sheepdog and Cocker Spaniel were the only breeds which markedly diered in size from the others,
but they did not seem to consistently dier from the larger breeds in their results.
In conclusion, we found signicant breed dierences for various behavioural and cognitive traits in dogs.
is is one of the few studies investigating individual breed dierences in dogs, especially in non-social cogni-
tive traits which are rarely studied in this context. Our results provide a more complete picture of breed-typical
behaviour in dogs.
Data availability
e datasets analysed during the current study are not publicly available due to privacy restrictions but are avail-
able from the corresponding author on reasonable request.
Received: 20 September 2022; Accepted: 22 December 2022
References
1. Croston, R., Branch, C. L., Kozlovsky, D. Y., Dukas, R. & Pravosudov, V. V. Heritability and the evolution of cognitive traits. Behav.
Ecol. 26, 1447–1459. https:// doi. org/ 10. 1093/ beheco/ arv088 (2015).
2. Mehrkam, L. R. & Wynne, C. D. L. Behavioral dierences among breeds of domestic dogs (Canis lupus familiaris): Current status
of the science. Appl. Anim. Behav. Sci. 155, 12–27. https:// doi. org/ 10. 1016/j. appla nim. 2014. 03. 005 (2014).
3. MacLean, E. L., Snyder-Mackler, N., vonHoldt, B. M. & Serpell, J. A. Highly heritable and functionally relevant breed dierences
in dog behaviour. Proc. Biol. Sci. 286, 20190716. https:// doi. org/ 10. 1098/ rspb. 2019. 0716 (2019).
4. Morrill, K. et al. Ancestry-inclusive dog genomics challenges popular breed stereotypes. Science 376, eabk0639. https:// doi. org/
10. 1126/ scien ce. abk06 39 (2022).
5. Gacsi, M., Kara, E., Belenyi, B., Topal, J. & Miklosi, A. e eect of development and individual dierences in pointing comprehen-
sion of dogs. Anim. Cogn. 12, 471–479. https:// doi. org/ 10. 1007/ s10071- 008- 0208-6 (2009).
6. Wobber, V., Hare, B., Koler-Matznick, J., Wrangham, R. & Tomasello, M. Breed dierences in domestic dogs’ (Canis familiaris)
comprehension of human communicative signals. Interact. Stud. Soc. Behav. Commun. Biol. Artif. Syst. 10, 206–224. https:// doi.
org/ 10. 1075/ is. 10.2. 06wob (2009).
7. Udell, M. A. R., Ewald, M., Dorey, N. R. & Wynne, C. D. L. Exploring breed dierences in dogs (Canis familiaris): does exaggera-
tion or inhibition of predatory response predict performance on human-guided tasks?. Anim. Behav. 89, 99–105. https:// doi. org/
10. 1016/j. anbeh av. 2013. 12. 012 (2014).
8. Sundman, A.-S. et al. Understanding of human referential gestures is not correlated to human-directed social behaviour in Labrador
retrievers and German shepherd dogs. Appl. Anim. Behav. Sci. 201, 46–53. https:// doi. org/ 10. 1016/j. a ppla nim. 2017. 12. 017 (2018).
9. Gnanadesikan, G. E., Hare, B., Snyder-Mackler, N. & MacLean, E. L. Estimating the heritability of cognitive traits across dog
breeds reveals highly heritable inhibitory control and communication factors. Anim. Cogn. 23, 953–964. https:// doi. org/ 10. 1007/
s10071- 020- 01400-4 (2020).
10. Riedel, J., Schumann, K., Kaminski, J., Call, J. & Tomasello, M. e early ontogeny of human–dog communication. Anim. Behav.
75, 1003–1014. https:// doi. org/ 10. 1016/j. anbeh av. 2007. 08. 010 (2008).
11. Dorey, N. R., Udell, M. A. & Wynne, C. D. Breed dierences in dogs sensitivity to human points: A meta-analysis. Behav. Processes
81, 409–415. https:// doi. org/ 10. 1016/j. beproc. 2009. 03. 011 (2009).
12. Szabo, D., Mills, D. S., Range, F., Viranyi, Z. & Miklosi, A. Is a local sample internationally representative? Reproducibility of four
cognitive tests in family dogs across testing sites and breeds. Anim. Cogn. 20, 1019–1033. https:// doi. org/ 10. 1007/ s10071- 017-
1133-3 (2017).
13. Horschler, D. J. et al. Absolute brain size predicts dog breed dierences in executive function. Anim. Cogn. 22, 187–198. https://
doi. org/ 10. 1007/ s10071- 018- 01234-1 (2019).
14. C larke, T., Mills, D. & Cooper, J. Exploring the utility of traditional breed group classication as an explanation of problem-solving
behavior of the domestic dog (Canis familiaris). J. Vet. Behav. 33, 103–107. https:// doi. org/ 10. 1016/j. jveb. 2019. 06. 004 (2019).
15. Bradshaw, J., Goodwin, D., Lea, A. & Whitehead, S. A survey of the behavioural characteristics of pure-bred dogs in the United
Kingdom. Vet. Rec. 138, 465–468 (1996).
Content courtesy of Springer Nature, terms of use apply. Rights reserved
14
Vol:.(1234567890)
Scientic Reports | (2022) 12:22529 | https://doi.org/10.1038/s41598-022-26991-5
www.nature.com/scientificreports/
16. Turcsán, B., Kubinyi, E. & Miklósi, Á. Trainability and boldness traits dier between dog breed clusters based on conventional breed
categories and genetic relatedness. Appl. Anim. Behav. Sci. 132, 61–70. https:// doi. org/ 10. 1016/j. appla nim. 2011. 03. 006 (2011).
17. Svartberg, K. Breed-typical behaviour in dogs—Historical remnants or recent constructs?. Appl. Anim. Behav. Sci. 96, 293–313.
https:// doi. org/ 10. 1016/j. appla nim. 2005. 06. 014 (2006).
18. Maglieri, V., Prato-Previde, E., Tommasi, E. & Palagi, E. Wolf-like or dog-like? A comparison of gazing behaviour across three dog
breeds tested in their familiar environments. R Soc. Open Sci. 6, 190946. https:// doi. org/ 10. 1098/ rsos. 190946 (2019).
19. Pongrácz, P., Miklósi, Á., Vida, V. & Csányi, V. e pet dogs ability for learning from a human demonstrator in a detour task is
independent from the breed and age. Appl. Anim. Behav. Sci. 90, 309–323. https:// doi. org/ 10. 1016/j. appla nim. 2004. 08. 004 (2005).
20. Shimabukuro, C., Putrino, N., Helbling, J., Tognetti, S. & Bentosela, M. Individual dierences in social and non-social behaviors
in domestic dogs (Canis familiaris) during the acquisition, extinction and reacquisition of a problem solving task. Behav. Processes
113, 179–186. https:// doi. org/ 10. 1016/j. beproc. 2015. 02. 002 (2015).
21. Pongrácz, P., Rieger, G. & Vékony, K. Grumpy dogs are smart learners—the association between dog-owner relationship and dogs’
performance in a social learning task. Animals 11, 961 (2021).
22. Overall, K. L., Dunham, A. E., Scheifele, P. & Sonstrom Malowski, K. Fear of noises aects canine problem solving behavior and
locomotion in standardized cognitive tests. Appl. Anim. Behav. Sci. 221. https:// doi. org/ 10. 1016/j. appla nim. 2019. 104863 (2019).
23. Passalacqua, C., Marshall-Pescini, S., Merola, I., Palestrini, C. & Previde, E. P. Dierent problem-solving strategies in dogs diag-
nosed with anxiety-related disorders and control dogs in an unsolvable task paradigm. Appl. Anim. Behav. Sci. 147, 139–148.
https:// doi. org/ 10. 1016/j. appla nim. 2013. 05. 003 (2013).
24. Bray, E. E., Sammel, M. D., Seyfarth, R. M., Serpell, J. A. & Cheney, D. L. Temperament and problem solving in a population of
adolescent guide dogs. Anim. Cogn. 20, 923–939. https:// doi. org/ 10. 1007/ s10071- 017- 1112-8 (2017).
25. Tiira, K. smartDOG, http:// smart dog. / engli sh (2016).
26. Lazarowski, L., Krichbaum, S., Waggoner, L. P. & Katz, J. S. e development of problem-solving abilities in a population of can-
didate detection dogs (Canis familiaris). Anim. Cogn. 23, 755–768. https:// doi. org/ 10. 1007/ s10071- 020- 01387-y (2020).
27. Passalacqua, C. et al. Human-directed gazing behaviour in puppies and adult dogs, Canis lupus familiaris. Anim. Behav. 82,
1043–1050. https:// doi. org/ 10. 1016/j. anbeh av. 2011. 07. 039 (2011).
28. Chapagain, D., Range, F., Huber, L. & Viranyi, Z. Cognitive aging in dogs. Gerontology 64, 165–171. https:// doi. org/ 10. 1159/ 00048
1621 (2018).
29. Kabadayi, C., Bobrowicz, K. & Osvath, M. e detour paradigm in animal cognition. Anim. Cogn. 21, 21–35 (2018).
30. Elgier, A. M., Jakovcevic, A., Mustaca, A. E. & Bentosela, M. Pointing following in dogs: Are simple or complex cognitive mecha-
nisms involved?. Anim. Cogn. 15, 1111–1119 (2012).
31. Kaminski, J. & Nitzschner, M. Do dogs get the point? A review of dog–human communication ability. Learn. Motiv. 44, 294–302
(2013).
32. Mendes, J. W. W., Resende, B. & Savalli, C. A review of the unsolvable task in dog communication and cognition: comparing dif-
ferent methodologies. Anim. Cogn. 24, 907–922 (2021).
33. Erdőhegyi, Á., Topál, J., Virányi, Z. & Miklósi, Á. Dog-logic: Inferential reasoning in a two-way choice task and its restricted use.
Anim. Behav. 74, 725–737. https:// doi. org/ 10. 1016/j. anbeh av. 2007. 03. 004 (2007).
34. Szetei, V., Miklósi, Á., Topál, J. & Csányi, V. When dogs seem to lose their nose: An investigation on the use of visual and olfactory
cues in communicative context between dog and owner. Appl. Anim. Behav. Sci. 83, 141–152. https:// doi. org/ 10. 1016/ s0168-
1591(03) 00114-x (2003).
35. Petter, M., Musolino, E., Roberts, W. A. & Cole, M. Can dogs (Canis familiaris) detect human deception?. Behav. Processes 82,
109–118. https:// doi. org/ 10. 1016/j. beproc. 2009. 07. 002 (2009).
36. Dwyer, C. & Cole, M. R. Domesticated dogs (Canis familiaris) tend to follow repeated deceptive human cues even when food is
visible. Learn Behav. 46, 442–448. https:// doi. org/ 10. 3758/ s13420- 018- 0356-8 (2018).
37. Bhattacharjee, D. & Bhadra, A. Adjustment in the point-following behaviour of free-ranging dogs—roles of social petting and
informative-deceptive nature of cues. Anim. Cogn. https:// doi. org/ 10. 1007/ s10071- 021- 01573-6 (2021).
38. Fiset, S., Beaulieu, C. & Landry, F. Duration of dogs’ (Canis familiaris) working memory in search for disappearing objects. Anim.
Cogn. 6, 1–10. https:// doi. org/ 10. 1007/ s10071- 002- 0157-4 (2003).
39. Hori, Y., Kishi, H., Inoue-Murayama, M. & Fujita, K. Dopamine receptor D4 gene (DRD4) is associated with gazing toward humans
in domestic dogs (Canis familiaris). Open J. Anim. Sci. 03, 54–58. https:// doi. org/ 10. 4236/ ojas. 2013. 31008 (2013).
40. Persson, M. E., Roth, L. S., Johnsson, M., Wright, D. & Jensen, P. Human-directed social behaviour in dogs shows signicant
heritability. Genes Brain Behav. 14, 337–344. https:// doi. org/ 10. 1111/ gbb. 12194 (2015).
41. Scandurra, A., Alterisio, A., Di Cosmo, A. & D’Aniello, B. Behavioral and perceptual dierences between sexes in dogs: An overview.
Animals (Basel) 8. https:// doi. org/ 10. 3390/ ani80 90151 (2018).
42. Watowich, M. M. et al. Age inuences domestic dog cognitive performance independent of average breed lifespan. Anim. Cogn.
23, 795–805. https:// doi. org/ 10. 1007/ s10071- 020- 01385-0 (2020).
43. Brady, K., Cracknell, N., Zulch, H. & Mills, D. S. Factors associated with long-term success in working police dogs. Appl. Anim.
Behav. Sci. 207, 67–72. https:// doi. org/ 10. 1016/j. appla nim. 2018. 07. 003 (2018).
44. Piotti, P., Satchell, L. P. & Lockhart, T. S. Impulsivity and behaviour problems in dogs: A reinforcement sensitivity theory perspec-
tive. Behav. Proc. 151, 104–110 (2018).
45. Van Poucke, E., Hoglin, A., Jensen, P. & Roth, L. S. V. Breed group dierences in the unsolvable problem task: herding dogs prefer
their owner, while solitary hunting dogs seek stranger proximity. Anim. Cogn. https:// doi. org/ 10. 1007/ s10071- 021- 01582-5 (2021).
46. Bognar, Z., Szabo, D., Dees, A. & Kubinyi, E. Shorter headed dogs, visually cooperative breeds, younger and playful dogs form eye
contact faster with an unfamiliar human. Sci. Rep. 11, 9293. https:// doi. org/ 10. 1038/ s41598- 021- 88702-w (2021).
47. Lazarowski, L. et al. Comparing pet and detection dogs (Canis familiaris) on two aspects of social cognition. Learn. Behav. 48,
432–443. https:// doi. org/ 10. 3758/ s13420- 020- 00431-8 (2020).
48. Lazarowski, L., Rogers, B., Waggoner, L. P. & Katz, J. S. When the nose knows: ontogenetic changes in detection dogs’ (Canis
familiaris) responsiveness to social and olfactory cues. Anim. Behav. 153, 61–68. https:// doi. org/ 10. 1016/j. anbeh av. 2019. 05. 002
(2019).
49. Marshall-Pescini, S., Frazzi, C. & Valsecchi, P. e eect of training and breed group on problem-solving behaviours in dogs. Anim.
Cogn. 19, 571–579. https:// doi. org/ 10. 1007/ s10071- 016- 0960-y (2016).
50. Hirschi, A., Mazzini, A. & Riemer, S. Disentangling help-seeking and giving up: dierential human-directed gazing by dogs in a
modied unsolvable task paradigm. Anim. Cogn. https:// doi. org/ 10. 1007/ s10071- 021- 01595-0 (2022).
51. McKinley, J. & Sambrook, T. D. Use of human-given cues by domestic dogs (Canis familiaris) and horses (Equus caballus). Anim.
Cogn. 3, 13–22 (2000).
52. Hoglin, A. et al. Long-term stress in dogs is related to the human-dog relationship and personality traits. Sci. Rep. 11, 8612. https://
doi. org/ 10. 1038/ s41598- 021- 88201-y (2021).
53. Konno, A., Romero, T., Inoue-Murayama, M., Saito, A. & Hasegawa, T. Dog breed dierences in visual communication with
humans. PLOS ONE 11, e0164760. https:// doi. org/ 10. 1371/ journ al. pone. 01647 60 (2016).
54. Gacsi, M., McGreevy, P., Kara, E. & Miklosi, A. Eects of selection for cooperation and attention in dogs. Behav. Brain Funct. 5,
31. https:// doi. org/ 10. 1186/ 1744- 9081-5- 31 (2009).
Content courtesy of Springer Nature, terms of use apply. Rights reserved
15
Vol.:(0123456789)
Scientic Reports | (2022) 12:22529 | https://doi.org/10.1038/s41598-022-26991-5
www.nature.com/scientificreports/
55. Helton, W. S. & Helton, N. D. Physical size matters in the domestic dog’s (Canis lupus familiaris) ability to use human pointing
cues. Behav. Processes 85, 77–79. https:// doi. org/ 10. 1016/j. beproc. 2010. 05. 008 (2010).
56. Hecht, E. E. et al. Signicant neuroanatomical variation among domestic dog breeds. J. Neurosci. 39, 7748–7758. https:// doi. org/
10. 1523/ JNEUR OSCI. 0303- 19. 2019 (2019).
57. FitBark, https:// tba rk. com.
Acknowledgements
We thank all the owners and their dogs who participated in this research, as well as the smartDOG licence testers:
Katja Kontu, Liisa Tikka, Sanni Somppi, Kati Vierimaa, and Ulla-Marja Ruistola. We would also like to thank the
University of Helsinki and the Finnish Foundation of Veterinary Research for funding this study.Open access
funded by the Helsinki University Library.
Author contributions
K.T., H.V., and E.R. collected the data. A.V., K.M., and K.T. supervised the project. S.J. designed and performed
the analysis. S.J. wrote the manuscript. All authors discussed and reviewed the manuscript and contributed to
the nal manuscript.
Competing interests
KT is the founder and owner of smartDOG Ltd, and HV and ER are licence testers of the company. KM, SJ, and
AV have no aliation with smartDOG. SJ’s work has been funded by the the Finnish Foundation of Veterinary
Research. ere are no other conicts of interest.
Additional information
Supplementary Information e online version contains supplementary material available at https:// doi. org/
10. 1038/ s41598- 022- 26991-5.
Correspondence and requests for materials should be addressed to S.J.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional aliations.
Open Access is article is licensed under a Creative Commons Attribution 4.0 International
License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the
Creative Commons licence, and indicate if changes were made. e images or other third party material in this
article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons licence and your intended use is not
permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.
© e Author(s) 2022, corrected publication 2023
Content courtesy of Springer Nature, terms of use apply. Rights reserved
1.
2.
3.
4.
5.
6.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:
use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
use bots or other automated methods to access the content or redirect messages
override any security feature or exclusionary protocol; or
share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at
onlineservice@springernature.com