Point Topography and Within-Session Learning Are Important Predictors of Pet Dogs’ (Canis lupus familiaris) Performance on Human Guided Tasks

Abstract

Pet domestic dogs (Canis lupus familiaris) are generally considered successful on object choice tasks, reliably following human points to a target. However, defining the specific topography of the point types utilized and assessing the potential for dogs to generalize their responses across similar point types has received little attention. In Experiment 1, we assessed pet dogs’ performance on an object choice task utilizing nine different point types that varied across the dimensions of movement, duration, and distance. These dimensions reliably predicted the performance of pet dogs on this task. In Experiment 2, pet dogs presented with nine different point types in the order of increasing difficulty performed better on more difficult point types than both naive dogs and dogs experiencing the nine points in the order of decreasing difficulty. In Experiment 3, we manipulated the attentional state of the experimenter (as in perspective taking studies) and found that human orientation was not a strong predictor of performance on pointing tasks. The results of this study indicate that dogs do not reliably follow all point types without additional training or experience. Furthermore, dogs appear to continuously learn about the dimensions of human points, adjusting their behavior accordingly, even over the course of experimental testing. These findings bring claims of pet dogs’ spontaneous success on pointing tasks into question. The ability to learn about, and respond flexibly to, human gestures may benefit pet dogs living in human homes more than a spontaneous responsiveness to specific gesture types.

Full text

Revista Argentina de Ciencias del Comportamiento, Julio 2013, Vol. 5, N°2, 3-20
ISSN 1852-4206
www.psyche.unc.edu.ar/racc
Revista Argentina de
Ciencias del Comportamiento
(RACC)
3
Point Topography and Within-Session Learning Are Important Predictors of Pet
Dogs’ (Canis lupus familiaris) Performance on Human Guided Tasks
Udell, Monique A.R.
*
a
; Hall, Nathaniel J.
b
; Morrison, James
b
; Dorey, Nicole R.
b
& Wynne, Clive D.L.
b
a
Department of Animal and Rangeland Sciences, Oregon State University. United States.
b
Department of Psychology, University of Florida. United States.
Psicología Comparada y Cognición Animal
Abstract
Resumen
Pet domestic dogs (Canis lupus familiaris) are generally considered
successful on object choice tasks, reliably following human points
to a target. However, defining the specific topography of the point
types utilized and assessing the potential for dogs to generalize their
responses across similar point types has received little attention. In
Experiment 1, we assessed pet dogs’ performance on an object
choice task utilizing nine different point types that varied across the
dimensions of movement, duration, and distance. These dimensions
reliably predicted the performance of pet dogs on this task. In
Experiment 2, pet dogs presented with nine different point types in
the order of increasing difficulty performed better on more difficult
point types than both naive dogs and dogs experiencing the nine
points in the order of decreasing difficulty. In Experiment 3, we
manipulated the attentional state of the experimenter (as in
perspective taking studies) and found that human orientation was
not a strong predictor of performance on pointing tasks. The results
of this study indicate that dogs do not reliably follow all point types
without additional training or experience. Furthermore, dogs appear
to continuously learn about the dimensions of human points,
adjusting their behavior accordingly, even over the course of
experimental testing. These findings bring claims of pet dogs’
spontaneous success on pointing tasks into question. The ability to
learn about, and respond flexibly to, human gestures may benefit pet
dogs living in human homes more than a spontaneous
responsiveness to specific gesture types.
Los perros domésticos son generalmente considerados exitosos en la tarea de
elección de objeto, siguiendo fiablemente señales humanas hacia el lugar
correcto. Sin embargo, tanto el definir la topografía precisa de las señales así
como el evaluar la capacidad de los perros para generalizar sus respuestas a
través de claves similares, ha recibido poca atención. En el Experimento 1,
evaluamos el rendimiento de los perros en la tarea de elección de objeto,
utilizando nueve diferentes tipos de señalamientos que variaron a través de tres
dimensiones: movimiento, duración, y distancia. Estas dimensiones fueron
predictores confiables del desempeño de los perros en esta tarea. En el
Experimento 2, los perros a los cuales se les presentaron las nueve formas de
señalamiento en un orden de dificultad creciente, tuvieron un mejor rendimiento
en las claves complejas que los perros que no fueron expuestos a ninguna clave, o
aquellos a los que se les presentaron las mismas señales en orden dificultad
decreciente. En el Experimento 3, variamos el estado de atención del investigador
(como en los estudios de toma de perspectiva) y encontramos que la orientación
del cuerpo de la persona no fue un buen predictor del desempeño de los perros en
respuesta al señalamiento. Los resultados de esta investigación indican que los
perros no siguen todos los tipos de señalamientos sin tener entrenamiento
adicional o experiencia. Más aun, los perros parecen aprender continuamente
acerca de estas dimensiones de movimiento, duración, y distancia, ajustando su
comportamiento de acuerdo a ello, aun durante la prueba experimental. Estos
hallazgos cuestionan las afirmaciones de que los perros sean espontáneamente
exitosos en las pruebas de señalamiento. La habilidad de aprender acerca de los
gestos humanos y responder flexiblemente a ellos, puede beneficiar a los perros
que viven en hogares humanos aún más que la capacidad espontánea de
responder a un tipo de gesto específico.
Key Words:
Canis Lupus Familiaris; Dog; Domestication; Learning;
Generalization; Communication; Pointing.
Palabras Claves:
Canis Lupus Familiaris; Perro; Domesticación; Aprendizaje; Generalización;
Comunicación; Señalar.
Received March 13, 2013, Received the review on May 7, 2013,
accepted on May 16, 2013.
*
Send Correspondence to: Udell, M.A.R.
E-mail: moniqueudell@gmail.com
1. Introduction
Over a decade of research has established that
many pet domestic dogs, Canis lupus familiaris, can
reliably follow a variety of human points to a target for
food reward (for a review see Udell, Dorey & Wynne,
2010a). In fact, pet dogs’ reputation for success in
human-guided tasks has made them a model species for

Udell et al. / RACC, 2013, Vol. 5, N°2, 3-20
4
investigating the origins of human socio-cognitive
behavior, especially with respect to point following and
sensitivity to attentional state (Miklósi, Topál & Csányi,
2004, 2007). Some have asserted that dogs are a good
model species for evolutionary reasons, arguing that
domestication or convergent evolution over the last
14,000 years (Nobis, 1979) can explain dogs’ human-
oriented behaviors (Hare, Brown, Williamson &
Tomasello, 2002). Others have proposed that in
addition to the dogs evolutionary history, ontogeny is
also critical for the development of dogs’ social
behavior, including human-oriented social behavior
(Bentosela, Barrera, Jakovcevic, Elgier, & Mustaca,
2008; Dorey, Udell & Wynne, 2010; Udell & Wynne
2010; Wynne, Udell & Lord, 2008), and that both
evolutionary and lifetime considerations should be
taken into account when interpreting dogs’ response to
human behavior (see the Two Stage Hypothesis, Udell
et al., 2010a). Indeed, a wide range of studies have
demonstrated that pet dogs show improvement on point
following tasks with age and experience (Dorey et al.,
2010; Miklosi et al., 1998; Wynne et al., 2008). Dogs
also readily learn about the relationship between human
actions and availability of reinforcement for acting in
accordance with them. Pet and shelter dogs can learn to
follow novel or challenging human gestures to a target
with repeated exposure- often in less than 15 trials
(Udell, Dorey, & Wynne, 2010b; Udell, Giglio, &
Wynne, 2008) and can learn to move towards a target
opposite of the one pointed to when that is the
reinforced response (Elgier, Jakovcevic, Mustaca, &
Bentosela, 2009). Dogs can also learn to increasingly
gaze at a human who provides treats and stop gazing
when reinforcement is no longer available (Bentosela et
al., 2008), and gain knowledge about novel occluders
that predict human attention or inattention, as well as
the relative likelihood of reinforcement for behaviors
such as begging, with experience (Udell, Dorey, &
Wynne, 2011). The domestic dogs’ proclivity for
learning about human behavior (Udell et al., 2011a), as
well as their ability to flexibly adapt to different
environments and relationships with humans worldwide
(Coppinger, & Coppinger, 2001), may be an important
factor in their success as a species, as well as their
success in human homes (Udell & Wynne, 2008). This
may also be a contributing factor to the growing
number of working roles dogs are now found in: from
search and rescue, to guide dogs for the blind, therapy
dogs, sniffer dogs, herding and livestock guarding dogs,
hunting dogs, competitive athletes and the list goes on.
While much attention has been given to the
possible evolutionary (Hare et al., 2002; Miklósi, Topál
& Csányi, 2004, 2007) and lifetime (Dorey et al., 2010;
Udell & Wynne 2010; Wynne, Udell & Lord, 2008)
origins of these behaviors in recent years, less attention
has been given to unprogrammed learning that could be
occurring during the course of experimental testing. It is
also unclear whether there are physical elements (or
stimulus properties) of human points that might
increase or decrease the salience of these stimuli in the
context of a choice task. Therefore the purpose of the
current study is not to further investigate the origins of
pet dogs responsiveness to human pointing. Instead this
study has three goals: (1) To provide a systematic
comparison of different forms of the basic human
pointing gesture by manipulating stimulus properties
along the dimensions of movement, duration and
distance (2) To investigate how experience and
generalization during the course of experimental testing
influences object choice task performance and (3) To
determine whether human attentional state acts as a
reliable independent predictor of dogs’ success on a
pointing task.
Experiment 1: What Is a Human Point?
Miklósi and Soproni (2006) compiled 24 studies
where non-human animals were required to utilize a
point in an object choice task. Based on the description
of stimuli used in these studies, the authors broke the
basic pointing gesture into three temporal categories
(static, dynamic, or momentary). Each of these
categories could be broken down further into five
spatial designations (at target/ touching, proximal,
distal, cross body, or asymmetric) and then divided
again into three attentional state categories (no gazing,
gazing at target, gazing at subject, gaze alternation). As
a result, over 60 different point-type topographies were
possible given the dimensions introduced by different
experimenters (Udell et al., 2010a), and this is no longer
a comprehensive list, as many additional point types
have been used since that time.
What’s more, individuals and populations of dogs
do not appear to respond to the 60+ variations of the
human point currently found in the literature as a
unified stimulus. Variability is regularly found across
‘point’ types with different topographies as well as
between individuals or populations of dogs
experiencing the same point type (e.g. Gácsi, Kara,
Belenyi, Topál, & Miklósi, 2009; Lakatos, Soproni,
Dóka, & Miklósi, 2009; Udell, Giglio, et al., 2008;

Udell et al. / RACC, 2013, Vol. 5, N°2, 3-20
5
Udell, Spencer, Dorey, & Wynne, 2012). For example,
a series of experiments demonstrated that roughly 93%
of dogs living in a shelter initially failed to follow a
momentary distal point, where the human arm and hand
was more than 50 cm from the target at full extension
and returned to a neutral position before the dog was
released to make a choice (Udell, Dorey & Wynne,
2008, 2010b). However these same dogs, as well as
shelter dogs in other studies, have been found to follow
simpler forms of human pointing, such as a dynamic
proximal point, which comes within 10 cm of the target
and is left in place until a choice has been made (Hare
et al., 2010; Udell et al., 2010b).
While many studies have suggested that pet dogs as
a group are more proficient at following momentary
distal points to a target (Gácsi et al. 2009; Udell, Dorey
et al., 2008), many individual pet dogs also initially fail
to follow this point type. Instead pet dogs that do
perform above chance on this gesture often do so with
perfect, or near perfect accuracy (bringing up the
population average to above chance levels), allowing
many of those that fail the test to go undetected. In fact,
most studies looking at pet dogs’ responsiveness to a
wide range of human gestures find different levels of
success across gesture types and between individuals,
independent of whether the average performance of the
subjects is above or below chance levels (Udell et al.,
2010a).
In other words after two decades of declarations
that domestic dogs follow points, we have yet to answer
a simple but important question: What is a point?
The large amount of variability in pet dog
performance across point types suggests that there may
not be a single answer. However, it may be possible to
identify common features of point types that the
majority of pet dogs follow (and also common features
of point types dogs often struggle with). If so it might
be feasible to identify point types that are more
prototypical than others (and also point types that are
less so). This could aid in future experimental designs
and interpretations of data that may be especially
relevant to cross-lab and cross-species comparisons.
While prior studies and meta-analyses have looked
at differences in pet dog performance on object choice
tasks in the presence of different gesture types (e.g.
Dorey et al., 2009; Miklósi, et al., 1998; Miklósi &
Soproni, 2006; Soproni, Miklósi, Topál & Csányi,
2001, 2002; Udell, Giglio et al., 2008), a systematic
experimental manipulation of stimulus dimensions
making up the basic human point with the extended arm
and hand has not yet been achieved to our knowledge.
Therefore, in our first experiment we look at a
continuum of related but distinct point types, common
to the literature, to assess how successful
experimentally naive pet dogs might be on point types
that systematically vary along dimensions of
movement, duration (role of memory), and distance
from the target (table 1).
Table 1. Point type conditions identified by combinations of relevant stimulus dimensions. All nine point type conditions were
utilized in Experiments 1 and 2. Black cells indicate point types also tested in Experiment 3 where attention was removed
during stimulus presentation.
2. Materials and Methods
2.1. Subjects
Seventy-two pet dogs (41 male, 30 female)
reported in good health comprised the study. Dogs
ranged from six months to eleven years of age (M = 2.7

Udell et al. / RACC, 2013, Vol. 5, N°2, 3-20
6
years, SD = 3) and represented a wide range of breeds
and mixes. While developmental factors have been
implicated in point following performance (Dorey et al.,
2010; Wynne et al., 2008), no age-based decrement in
performance has been reported for dogs over the age of
four months (Dorey et al., 2010). Therefore all recruited
subjects were not only over this age, but had been
residing in their current home for at least 4 months. All
dogs were naive to experimental pointing tasks at the
time of testing and were tested indoors by an unfamiliar
experimenter.
To prevent generalization across point types, each
subject only experienced ten trials of a single point
type, or in other words participated in only one
condition of the nine possible point type conditions
tested (see table 2 for descriptions of each point type).
Therefore each condition required eight experimentally
naive dogs. Dogs were randomly assigned to a
condition before testing began.
Table 2. Point type definitions.
Point Type
Definition
Static touch
The experimenter touches the target container with one finger while the dog’s view of the
testing area is blocked. The dog is then allowed into the testing area while the experimenter
maintains his touching position until the dog makes its choice.
Dynamic tap
The experimenter extends his arm toward the target container while the dog watches and
continually taps the container with one finger until the dog makes its choice.
Momentary tap
The experimenter extends his arm toward the target container while the dog watches and
taps four times on the top of the container with one finger. The experimenter then returns to
a neutral position and the dog is released to make its choice.
Static proximal point
The experimenter begins pointing towards the target container, with his finger 10 cm from
the container, while the dog’s view of the testing area is blocked. The dog is then allowed
into the testing area and the experimenter maintains his pointing position until the dog
makes its choice.
Dynamic proximal point
The experimenter extends his arm toward the target container while the dog watches and
maintains a point with his finger 10 cm from the container until the dog makes its choice.
Momentary proximal point
The experimenter extends his arm toward the target container while the dog watches and
maintains a point with his finger 10 cm from the container for 4 seconds. The experimenter
then returns to a neutral position and the dog is released to make its choice.
Static distal point
The experimenter begins pointing towards the target container, with his finger 50 cm from
the container, while the dog’s view of the testing area is blocked. The dog is then allowed
into the testing area and the experimenter maintains his pointing position until the dog
makes its choice.
Dynamic distal point
The experimenter extends his arm toward the target container while the dog watches and
maintains a point with their finger 50 cm from the container until the dog makes its choice.
Momentary distal point
The experimenter extends his arm toward the target container while the dog watches and
maintains a point with their finger 50 cm from the container for 4 seconds. The
experimenter then returns to a neutral position and the dog is released to make its choice
2.2. Testing materials and layout
Two empty paint cans (15 cm diameter, 22 cm tall)
with lids tightly fastened served as response objects.
During experimental testing food was not present in or
on either can until the subject made a correct response.
This was done to control for smell given off by hidden
food, which could guide the dog’s response
independent of experimental stimuli. Although sham
baiting, or smearing/false baiting both choice objects
with food prior to testing, has also been used to address
this potential confound in the past (e.g. Miklósi et al.,
1998; Riedel, Schumann, Kaminski, Call, & Tomasello,

Udell et al. / RACC, 2013, Vol. 5, N°2, 3-20
7
2008) at least one study has demonstrated that sham
baiting alone is an insufficient olfactory control for
some canine subjects (see Udell, Dorey et al., 2008).
Another study demonstrated that dogs are capable of
using olfactory cues to locate hidden food in an object
choice task (Szetei, Miklósi, Topál, & Csányi, 2003)
although dogs may sometimes continue to favor visual
human stimuli to olfactory cues.
The target cans were placed 0.5 m on either side of
the experimenter (E1) and remained there throughout
testing. At the start of each trial an assistant (E2) held
the subject 2.5 m back from the center-line of the
experimenter (see figure 1). All distances were
measured prior to testing and marked with masking tape
on the floor.
Figure 1. Testing Layout.
During testing dogs were rewarded with a preferred
type of commercially available dog treat. To ensure
food motivation and absence of fear in the experimental
setting, dogs were required to readily eat this treat from
the experimenter’s hand prior to testing to be included
in the study. The correct container or target was
determined pseudorandomly before sessions, subject to
the constraints that no one location was designated
correct more than three times in a row and each location
was correct for 50% of the trials.
2.3. Motivation Test
All testing began with a motivation test (MT) to
familiarize the dog with the response objects and ensure
that the dogs were motivated to eat food off the cans
when given freely. This consisted of the experimenter
(E1) calling the dog’s name to gain its attention. He
then placed a treat on top the designated paint can in
view of the dog. The dog was allowed to approach the
can and consume the treat. Experimental trails began
after a subject successfully completed this motivation
test four times (two MT for each can). Dogs then
immediately moved on to experimental testing.
2.4. Experimental Testing
During experimental trials the dog was held 2.5 m
back from the empty cans by the assistant, E2; the
experimenter E1 called the dog’s name to gain its
attention. The experimenter then administered the
designated stimulus (one of the nine possible point
types described in table 2) indicating the previously
determined target can. The assistant released the dog,
which was then allowed to approach one of the two
cans. A choice was recorded when the dog’s muzzle
came within 10 cm of either can or when the dog
touched the can with any part of its body. If the dog
chose the correct can first, the experimenter placed a
treat on top the correct can for the dog to consume. To
minimize any effects of delay between the subject’s
response and receipt of food, the experimenter also
marked a correct response by saying “good dog” while
placing the treat on the can. The only response
considered correct during analysis was approach of the
target –the can pointed to- during the one minute
maximum duration of a trial; if the alternative can was
approached first or any other response was made this
was considered incorrect. If during testing the dog made
three incorrect responses in a row, two additional MTs
were given, one to each can. Loss of motivation, as
indicated by failure to approach a can and take the food
during a MT, resulted in as suspension of testing. No
dog ever failed a test of motivation.
Each subject experienced a total of ten
experimental trials, only witnessing a single assigned
point type.
2.5. Control trials
A control trial followed every two experimental
trials, with an additional control trial at the end of
testing. In total each subject received six control trials.
Control trials were carried out in an identical way to
experimental trials, except that after calling the dog, the
experimenter remained in a neutral position facing the
dog (no point was given). This neutral position was
held until the subject made a choice or until one minute
had passed indicating that the trial had timed out. Just
as in experimental trials, a correct or target can was
predetermined (the correct can was pseudo-randomly
assigned so that each can was correct 50% of the time)
before testing and the experimenter was aware of which
can was the target. Just like experimental trials, subjects
were allowed to eat food from the target can after
correct choices and did not receive food if an incorrect
response was made. This was done to detect the
presence of extraneous stimuli that could be controlling

Udell et al. / RACC, 2013, Vol. 5, N°2, 3-20
8
the dog’s behavior beyond the designated point in
experimental trials (including unintentional movements
on the part of the experimenter).
Dogs did not perform above chance in the absence
of a pointing stimulus (Mean of 1.99 correct responses
out of 6; 95% CI [1.73, 2.25]), suggesting that
successful point following performance during
experimental trials was not a product of other available
stimuli within the experimental setting. In fact in the
absence of a point (as in the case of control trials) many
dogs choose neither can (this response was more
common after a dog had already experienced one or
more control trials), instead they engaged in exploratory
activities, waited at the starting point, or approached the
experimenter often sitting neutrally or begging near by.
This might suggest that dogs come to use human points
not only as a stimulus predicting the location of food,
but also a stimulus indicating the beginning of a choice
trial. It is also possible that in comparison to simple
point types, where dogs often reliably earn food > 80%
of the time, control trials may offer too little payoff (on
average 50%) to ensure a response is made on each of
these trials, suggesting that dogs may learn to
discriminate between experimental and control trials
over the course of testing. However such outcomes still
suggest that dogs are responding to the point, and not
other external environmental stimuli, during
experimental trials.
2.6. Statistical analysis
Performance analysis was based on correct
responses. An individual was considered successful on
the task if it made eight or more correct responses out
of ten trials (binomial test, p < .05). A one-sample t-test
was used to determine if a group of eight dogs followed
a point type to the target more often than would be
expected by chance. To determine if differences in
performance existed across point types a single factor
ANOVA was utilized. Performance between point types
differing in designated point dimensions - movement,
duration and distance- were then compared using
corrected t-tests.
All statistical tests were two-tailed and had alpha
set at .05 unless otherwise noted.
3. Results
3.1. Performance across point types
Each group of dogs was successful in following its
assigned point type at above chance levels (one sample
t-tests, t (7) = 6.00, p < .001) with the exception of the
static distal point group (t (7) = 2.27, p = .06) and the
momentary distal point group (t (7) = .34, p = .75).
Mean performance scores and number of individual
successes for each group can be found in figure 2.
When comparing group performances for the different
point types, a highly significant difference in the
average number of correct responses between the nine
point types arose (between-subject single-factor
ANOVA, F (11, 84) = 8.03, p < .001).
Figure 2. Mean number of correct responses and number of successful individuals across point types in Experiment 1. Point
types are abbreviated as follows: DT (Dynamic tap), DPP (Dynamic proximal point), ST (Static touch), MT (Momentary tap),
SPP (Static proximal point), MPP (Momentary proximal point), DDP (Dynamic distal point), SDP (Static distal point), MDP
(Momentary distal point). Error bars represent +/- SEM. ** indicates one sample t-test, t (7) > 6.00, p < 0.001. Individuals
were considered successful with a point type if they made eight or more correct responses out of ten (binomial test, p < 0.05).
Dashed line at chance.

Udell et al. / RACC, 2013, Vol. 5, N°2, 3-20
9
3.2. Stimulus dimensions
Our original prediction was that the source of such
differences between groups would be related to the
stimulus dimensions of movement, duration, and
distance (as measured between the end of the stimulus
and target container), therefore two additional analyses
were conducted:
1) Movement/duration could be broken into three
categories based on the point-types utilized in this
study: dynamic (movement, point in place at time of
choice), static (no movement, point in place at time of
choice), and momentary (movement, point no longer in
place at time of choice). Using corrected two-sample t-
tests (corrected alpha, .02), we found a significant
difference between pet dog performance on dynamic
points [in place at time of choice] versus momentary
points [absent at time of choice] (t (46) = 2.70, p = .01),
with dogs making more correct choices on average
when presented with dynamic points. We found no
significant difference between momentary [containing
movement] and static points [containing no movement]
(t (46) = 1.19, p = .24) nor between dynamic
[containing movement] and static points [containing no
movement] (t (46) = 1.82, p = .08). Therefore point
duration (or presence at the time of choice) seemed to
have a larger influence than movement alone. At the
individual level, more dogs were successful in static or
dynamic conditions (20/24 each) than in momentary
conditions (17/24), however this difference was not
statistically significant (two-way Fisher’s exact test, p =
.49). See figure 3A.
Figure 3. Group and individual performance by dimension. (A) Mean number of correct responses (out of 10) for each
dimension. (B) Number of dogs successful in following point types categorized under each dimension (out of 24). Error bars
represent +/- SEM. ** p < 0.001, * p < 0.05. Individuals were considered successful with a point type if they made eight or
more correct responses out of ten (binomial test, p < 0.05).
2) Distance between the end of the pointing finger
and the target could also be broken into three
categories: tap/touch (direct contact made with the
target), proximal points (10 cm from target), and distal
points (50 cm from target). Using corrected t-tests
(corrected alpha, .02) we found a significant difference
between pet dog performance when comparing distal
points with proximal points (t (46) = 4.21, p < .001) and
between distal points and tap/touch (t (46) = 4.25, p <
.001). In both cases dogs performed more accurately
when the human point came closer to (or touched) the
target. There was no difference between tap/touch and
proximal points (t (46) = 0, p = 1.00). At the individual
level, significantly more dogs were successful in
proximal conditions (22/24) compared with distal
(12/24) (two-way Fisher’s exact test, p < .01), and in
tap/touch conditions (23/24) compared to distal (two-
way Fisher’s exact test, p < .001). A significant
difference between proximal points and tap/touch was
not found. See figure 3B.
Experiment 2: Learning & Generalization
Experiment 1 suggested that both the duration
(favoring points that remained in place until a choice
was made) and distance (favoring points coming close
to or touching the target) of a human point can
significantly influence the likelihood that a dog will be
successful in following a human point to a target. Point
types lacking both long duration and proximity, such as
the momentary distal point, appear to be the most
difficult for experimentally naive dogs to respond to.
General failure to follow the static distal point may

Udell et al. / RACC, 2013, Vol. 5, N°2, 3-20
10
suggest that the absence of movement, coupled with
increased distance, could also make some gestures more
difficult to follow.
Yet prior studies have reported that pet dogs do
sometimes follow points lacking movement, made from
a distance, or presented briefly at higher levels, and can
in some cases perform well on point types that combine
these elements – including the momentary distal point.
Certainly individual dogs might have adequate
experience (possibly beyond that of the general
population) allowing them to perform well using these
more difficult point types (for example dogs with
agility training or even those who spend most of the day
with their owner might be at an advantage over pet dogs
with little to no training and those that spend much of
the day home alone). Indeed in most studies at least a
few individuals perform successfully even when more
subtle gestures are used. It is also possible that some
breeds may be more sensitive to specific stimulus
properties than others (Dorey et al., 2009). However
another important factor may be the methods used to
assess dogs’ ‘spontaneous’ responsiveness to human
points; including the number of trials or point types a
dog will experience over the course of experimental
testing. While all of these factors are of potential
importance, here we intend to focus specifically on the
latter.
Many previous studies have presented a single
group of dogs with a large number of point types over
the course of a single experiment (e.g. Soproni et al.,
2001, 2002; Udell et al., 2012). While this approach is
not inherently problematic (it can be used to assess a
dogs capacity to follow a variety of point types),
elevated success rates in studies using this methodology
may indicate that subject performance is not truly
spontaneous (even if the dog could be considered naive
at the start of the experiment), but instead influenced by
experience gained during testing itself. After all,
research has demonstrated that pet and shelter dogs can
learn to follow a novel or challenging human gesture to
a target with repeated exposure- often in less than 15
additional trials (Udell et al., 2010b; Udell, Giglio et al.,
2008).
On the other hand, it might be argued that while
repeated exposure to the same human point type
improves canid performance on an object choice task
(Udell et al., 2010b; Udell, Giglio et al., 2008; Virányi
et al., 2008), studies presenting dogs with a string of
topographically distinct human points are not subject to
the same criticisms. Whether exposure to physical
properties of one point type, sharing characteristics with
more difficult or unusual point types, might allow dogs
to generalize their response to novel gestures has
remained untested. In Experiment 2 we directly test and
measure the effect of experimental exposure on pet
dogs’ point following performance.
4. Materials and Methods
4.1. Subjects
Sixteen additional pet dogs reported in good health
comprised the study. Subjects ranged in age from nine
months to nine years (M = 4.3 years, SD = 2.5), eight
were male and eight female, and represented a range of
breeds and mixes. All subjects had been residing in
their current home for at least 4 months. All dogs were
naive to the task at the time of testing and were tested
indoors by an unfamiliar experimenter.
Each subject experienced the full series of nine
point types, as defined in table 2. Testing was broken
into three sessions; each dog experienced three point
type conditions per session. Breaks between sessions
were determined by participant availability but were
never shorter than one day and never longer than two
weeks. Half of the subjects experienced the point type
conditions in the order of increasing difficulty (easy to
difficult), as established by Experiment 1 and
additionally confirmed by independent difficulty ratings
made by eleven anonymous researchers in the field
naive to the purpose of the study (these measures were
highly correlated: Pearson’s correlation coefficient,
rating x performance, R = 0.94). The remainder of the
subjects experienced each point type in order of
decreasing rank difficulty (difficult to easy). See figure
4 for point types in order of increasing/decreasing
difficulty. Before testing began, dogs were randomly
assigned to their respective conditions with one
exception: if two dogs from the same household
participated in the study each was assigned to a
different condition to avoid potential confounds
between condition assignment and living environment.
4.2. Testing materials, layout, MT, and experimental
trials
Materials, layout, motivation tests, and
experimental trials were identical to those in
Experiment 1, with the following exceptions:
As in Experiment 1 subjects experienced MT at the
beginning of testing. Since subjects in Experiment 2
were required to complete three point-type conditions
per session (a total of 30 experimental trials, compared
to 10 in Exp 1) an additional two MT, one to each side,

Udell et al. / RACC, 2013, Vol. 5, N°2, 3-20
11
were conducted after the first and second conditions of
each session to ensure the dog was still food motivated
before proceeding to the next condition. No subject
failed a test of motivation within the course of a
session.
Each subject received a total of 90 experimental
trials over the course of testing; 10 trials per point type
condition.
4.3. Control trials
A control trial followed every ten experimental
trials, resulting in three control trials per session and
nine control trials per dog. Control trials were carried
out in the same manner as in Experiment 1. Dogs did
not perform above chance on control trials, mean of
3.44 (95% CI [2.71, 4.17]) control trials correct out of
9, suggesting that point following performance was not
influenced by other stimuli within the experimental
setting.
4.4. Statistical analysis
Performance analysis was based on correct
responses. An individual was considered successful on
the task if it made eight or more correct responses out
of ten trials (binomial test, p < .05). A one-sample t-test
was used to determine if a group of eight dogs
performed better on a point type than would be
predicted by chance.
A two-factor within subject ANOVA was used to
determine if there were significant differences in
performance across point types and between the two
subject groups (difficult to easy; easy to difficult). For
each group, we also compared the performance between
point types differing in designated point dimensions
(movement, duration and distance) using corrected t-
tests.
All statistical tests were two-tailed and had alpha
set at .05 unless otherwise noted.
Figure 4. Mean number of correct responses and number of successful individuals across point types in Experiment 2. Point
types abbreviations are the same as in figure 2. Dog subjects in the Easy to Difficult (E-D) condition experienced all point
types in order from left to right. Dog subjects in the Difficult to Easy (D-E) condition experienced all point types in order from
right to left. Error bars represent +/- SEM. ** indicates one sample t-test, t (7) > 6.00, p < 0.001; * indicates one sample t-test, t
(7) > 3.25, p < 0.05. *** Located over the momentary distal point bracket indicates a significant difference between groups (t-
test, t (7) = 4.72, p < 0.0006). Individuals were considered successful if they made eight or more correct responses out of ten
on a point type (binomial test, p < 0.05). Dashed line at chance.
5. Results
Experiment 2 was designed to determine if dogs
would learn about human point types over the course of
experimental testing. We were interested in the
possibility of stimulus generalization across point types.
Specifically, we looked for improved performance on
novel point types sharing some but not all the stimulus
properties with previously experienced point types.
Each subject received ten trials of all of the nine

Udell et al. / RACC, 2013, Vol. 5, N°2, 3-20
12
different point types (90 trials total). Eight of the
subjects experienced the point type conditions in the
order of increasing rank difficulty (easy to difficult); the
other eight experienced the point types in order of
decreasing rank difficulty (difficult to easy).
Dogs in the easy to difficult condition were
successful on each of the nine point types as a group
(one sample t-tests, t (7) > 4.50, p < .01). At the
individual level at least half the subjects performed
significantly above chance (binomial tests, p < .05) on
each point type. Dogs in the difficult to easy condition
were successful on eight of the nine point types as a
group (one sample t-tests, t (7) > 3.25, p < .01), failing
to reach above chance performance only on the
momentary distal point (one sample t-test, t (7) = 1.67,
p = .14). No dog in the difficult to easy condition was
individually successful on the momentary distal point
(binomial tests, p > .05), and fewer than half of the
subjects experiencing point types in order of decreasing
difficulty were successful on the momentary proximal
point (see figure 4).
5.1. Experience and learning
A significant difference was found between the
mean performances of dogs in the easy to difficult
condition compared to dogs in the difficult to easy
condition, with the former outperforming the latter on
the series of object choice tasks (two-factor within
subject ANOVA, F (1, 14) = 5.97, p = .03). There was
also a highly significant difference in performance
between point types (F (8, 112) = 15.3, p < .001), as
well as a significant interaction between condition and
point type (F (8, 112) = 5.66, p < .001). Because dogs
were least successful on the momentary distal point in
Experiment 1 we predicted that the effect of experience
would be most apparent for this point type, therefore we
directly compared the average performance of dogs
experiencing this point first (difficult to easy condition)
with dog who experienced this point last (easy to
difficult condition). A highly significant difference was
found between the mean performance of dogs
experiencing eight simpler point type conditions prior
to encountering the momentary distal point (mean =
7.89 correct out of 10), and those without prior
experience (mean = 4.38 correct out of 10) (t-test, t (7)
= 4.72, p < .001). At both the group and individual
level, dogs with more pointing experience performed
significantly better on the momentary distal point, even
though they had not previously encountered this
specific gesture type earlier in testing.
5.2. Stimulus dimensions
As in Experiment 1, two additional analyses were
conducted to compare the salience of our focal stimulus
dimensions (movement, duration, and distance) based
on the performance of pet dogs on the object choice
task. This was done separately for the two subject
groups because prior analyses indicated that order of
point exposure influenced performance, especially for
the most difficult point types. We wanted to determine
if each group’s overall pattern of response across
stimulus dimensions was different as well.
1. As in Experiment 1, movement/duration could
be broken into three categories based on the point types
utilized in this study: dynamic, static, and momentary.
Using corrected two-sample t-tests (corrected alpha,
.02), we found no significant difference in mean trials
correct between dynamic (9.5/10 correct), static
(8.9/10) and momentary (8.6/10) points for dogs in the
easy to difficult condition (t (46) < 2.28, p > .03). On
the other hand, dogs in the difficult to easy condition
chose the correct target significantly more often on
average when the pointing stimulus was dynamic
(9.5/10 correct) as opposed to static (8.2/10) (t (46) =
3.37, p < .01) or momentary (7.1/10) (t (46) = 4.13, p <
.001). No significant difference was found between
static and momentary points (t (46) = 1.66, p = .11).
2. Distance between the point and the target could
also be broken into three categories: tap/touch, proximal
points, and distal points. Using corrected t-tests
(corrected alpha, .02), we found no significant
differences in mean number of trials correct between
tap/touch (8.8/10 correct), proximal (9.5/10) and distal
points (8.7/10) for dogs in the easy to difficult group (t
(46) < 2.02, p > .03). However dogs in the difficult to
easy condition performed significantly better on
average with tap/touch stimuli (9.3/10 correct) (t (46) =
5.21, p < .001) and proximal points (8.8/10) (t (46) =
3.73, p < .001) when compared with distal points
(6.7/10). No significant difference was found between
tap/touch and proximal points (t (46)= 1.17, p = .24).
Therefore experiencing points in order of
increasing difficulty may have allowed dogs to
overcome decrements in performance associated with
greater pointing distance and shorter point duration (or
the need for memory, given that in momentary points
the point it removed prior to the dog making a choice)
initially identified in Experiment 1 and also seen in the
difficult to easy condition of Experiment 2. This
strongly suggests that experience acquired during the
course of experimental testing can have a significant

Udell et al. / RACC, 2013, Vol. 5, N°2, 3-20
13
impact on the performance of dogs across different
point types utilized in human-guided object choice
tasks. This effect can be influenced by the testing order
itself and in some cases could lead to performances that
appear to support spontaneous success on a novel
gesture type, but are really the by-product of learning
and generalization from earlier testing.
Experiment 3: Does Human Attentional State
Matter
Point following behavior is often considered a
measure of joint attention and has been associated with
healthy socio-cognitive development, language
formation and even theory of mind in the human
developmental literature (Carpenter, Nagell &
Tomasello, 1998; Goldin-Meadow, 2007; Tomasello,
Carpenter & Liszkowski, 2007). To some, following the
point of another individual implies a deep
understanding of communicative intent or even
knowledge of the mental states of others (Gómez, 2007;
Tomasello et al., 2007). In the same tradition, domestic
dogs have been tested for their responsiveness to human
gestures, including pointing. However it is far from
clear what point following behavior can tell us about a
dog’s understanding of a human pointer’s intentions, if
anything (including whether dogs actually treat points
as inherently cooperative gestures).
Perspective taking tasks have traditionally come
closer to addressing this type of question. Indeed, pet
dogs have been recognized for their ability to
discriminate between a person looking towards them
and one looking away (or with obscured vision) (E.g.
Forbidden food tasks: Bräuer, Call & Tomasello, 2004;
Call, Bräuer, Kaminski & Tomasello, 2003; Begging
tasks: Cooper et al., 2003; Gácsi, Miklósi, Varga,
Topál, & Csányi, 2003, Udell, Dorey, & Wynne, 2011).
Although occluders, or barriers of attention, used in
both tasks can vary substantially (e.g. reading a book,
bucket over the head, blindfold over the eyes and even
portable wall placement) in the most straight-forward
version of the begging task, a dog is given the choice to
beg from either an attentive experimenter facing the dog
or an inattentive experimenter whose back is turned.
Across studies, dogs have shown sensitivity to the
cooperative nature of begging tasks and the importance
of experimenter attention; reliably approaching the
person looking at them when begging for food – not
approaching the individual with her back turned
(Cooper et al., 2003; Gácsi et al., 2003, Udell et al.,
2011).
On the other hand, the forbidden food task is
clearly not cooperative; instead human attention serves
a competitive or preventative role. For this task, a piece
of food is placed within the dogs reach, and the owner
instructs the dog not to take it. The human is then either
attentive, watching the dogs actions, or inattentive. In
this case dogs’ sensitivity to attentional state has been
demonstrated by dogs increased willingness to steal
food when the human’s back is turned (or when one of
many other possible occluders is used to block the
human’s view of the food or dog), thereby increasing
the dog’s chances of obtaining the food and avoiding
punishment for doing so. Therefore dogs have not only
demonstrated a sensitivity to cues that predict to
attentional state, but also discriminate between contexts
where human attention will facilitate reinforcement
(cooperative scenarios) from contexts where the
absence of human attention is most beneficial (non-
cooperative scenarios).
To date the knower-guesser paradigm provides one
of the few examples of where human pointing and
attentional state measures are combined into a single
task. In this task one experimenter, the ‘knower’,
witnesses the hiding of a piece of food. The subject is
not able to see where the food is hidden, however they
do have visual access to the ‘knower’ during the baiting
phase. The other experimenter, the ‘guesser’, is
prevented from seeing where the food was hidden.
Afterwards both individuals point at a location where
the food might be. The correct response is for the
subject to choose the location indicated by the
‘knower’, and dogs have performed successfully on
several versions of this task (Cooper et al., 2003;
Maginnity, 2007). While interesting in its own right,
this particular methodology is designed to assess what
the dog is knows about the attentional state of the
experimenter with relation to the baiting process; or, in
other words, the experimenter’s knowledge about the
location of the food, not the location of the dog.
Therefore from the dog’s perspective both
experimenters might be attempting to engage in
‘cooperative behavior’, even if one can only provide his
best guess about the location of the food. Begging and
forbidden food tasks are inherently different from the
knower-guesser task in an important way: the human
always knows where the food is (in some cases they are
holding it), the question is whether the person is
attending to the behavior of the dog and whether or not
this is beneficial or problematic for the dog depending
on the nature of the task (cooperative or not). In this

Udell et al. / RACC, 2013, Vol. 5, N°2, 3-20
14
case, behaviors requiring human cooperation, like
begging, should decrease when a human turns her back
(a signal of inattention), conversely behaviors that
compete or conflict with human goals (e.g. human
guarding or forbidding a piece of food), should increase
when a human turns her back. Other behavioral
responses to human-attentional state likely fall
somewhere in between these two ends of the approach-
avoidance continuum; human attention should have less
influence on a dog’s response when the human action is
not perceived as inherently cooperative or competitive.
In Experiment 3 we borrow this perspective taking
methodology to assess whether point following in dogs
is influenced by the attentional state of the human. If
pointing is strictly viewed as a cooperative activity by
dogs, then a cue of inattention (such as turning one’s
back) might be expected to reduce responsiveness to
typically salient gesture types as it does in other
cooperative tasks. On the other hand, dogs may learn
that points can be useful independent of human
cooperative intent (or may not rely on perceptions of
communicative intent or cooperation at all). If this is
true the human’s attentional state may not have reliable
predictive value for point following behavior, but
instead may simply serve as one of many possible
stimulus dimensions that contribute to the overall
salience of the human point.
To test this we revisited three of the point types
from Experiment 1 (dynamic tap, static proximal point,
momentary distal point) comparing dogs’ performance
when the human experimenter faced forward or had his
back turned.
6. Materials and Methods
6.1. Subjects
All subjects were pet domestic dogs reported to be
in good health at the time of testing. Dogs ranged from
6 months to eight years in age (M = 3.4 years, SD = 2.4
years), 16 were male and 16 female, and consisted of a
range of breeds and mixes. All subjects had been
residing in their current human’s home for at least 4
months and were tested indoors by an unfamiliar
experimenter.
Two groups of dogs participated: The experienced
group (E) - made up of eight dogs from Experiment 2,
having previously experience 90 trials of the object-
choice task over the course of all nine forward facing
point types, and the no experience group (NE) - 24
naive dogs, split into three sets of eight dogs- one set
assigned to each back-turned point type.
As in Experiment 2, subjects in the experienced
group (E) experienced all three point types utilized in
Experiment 3 (presented in the following order:
dynamic tap, static proximal, momentary distal). Dogs
experienced these three point types within one 20-30
minute session on a single day. As in Experiment 1,
each dog in the no experience group (NE) only received
10 trials of a single point type during testing to reduce
the possibility of generalization.
6.2. Testing materials, layout, MT, and experimental
trials
Materials, layout, MT, control and experimental
trials were identical to those in Experiment 2, with the
following exceptions:
Only three point type conditions were included in
Experiment 3: Dynamic tap, static proximal point, and
momentary distal point. These conditions were selected
because each point possessed different combinations of
the stimulus dimensions investigated in Experiments 1
and 2 (see table 1), together representing a
comparatively easy, moderate, and difficult form of the
pointing stimulus.
The most significant change to the methods in
Experiment 3 was the shift from forward facing to
backwards facing point presentations. During
experimental and control trials, the testing layout
depicted in figure 1 was utilized, however E1 faced
away from the dog, looking towards a back wall for the
duration of testing. Since the experimenter could not
see the dog from this position, the assistant began the
trial by saying “point,” to indicate that the dog had
oriented towards the experimenter. Once the point had
been presented the dog was released. The assistant was
also responsible for alerting the experimenter once the
dog had made a choice: “yes” indicated a correct
response, “no” indicated an incorrect response, and
“time” indicated that the one-minute timeout period had
passed. Scoring was based on the same choice criterion
described in Experiments 1 and 2. The experimenter
only provided the food reinforcer to the dog if the
assistant indicated a correct choice. This was done to
ensure that a prompt and consistent response was made
to the dog’s behavior during testing even if it occurred
outside of the experimenter’s field of vision.
6.3. Control trials
Control trials were carried out in an identical
manner to Experiments 1 and 2, only the experimenter’s
back was turned during them. Overall, dogs in both the
experienced group (mean of 4.13 out of 12 trials

Udell et al. / RACC, 2013, Vol. 5, N°2, 3-20
15
divided between three sessions; 95% CI [3.24,5.02])
and no experience group (mean 2.08 out of 6 trials
presented during the course of their only session; 95%
CI [1.57,2.59]) did not perform above chance on control
trials.
6.4. Statistical analysis
Performance analysis was based on correct
responses. An individual was considered successful on
the task if it made eight or more correct responses out
of ten trials (binomial test, p < .05). A one-sample t-test
was used to determine if a group of eight dogs
performed better on a point type than would be
predicted by chance.
A two-factor ANOVA (one within, one between),
was used to determine if there were significant
differences in performance across back-turned point
types and between the two subject groups (experience,
no experience). We also independently compared the
mean performance scores of experienced and
inexperienced dogs for the back-turned momentary
distal point using a t-test. Based on the outcomes of
Experiments 1 and 2, we predicted that the momentary
distal point would be the strongest independent
indicator of the effect of prior experience on
performance.
Finally the influence of attentional state in the
context of an object choice task was evaluated
comparing naive dogs experiencing back-turned
dynamic tap, static proximal, and momentary distal
points with the experimenter in a forward facing
orientation (from Experiment 1) with naive dogs
experiencing those same point types with the
experimenter in a back-turned orientation (a cue of
inattention).
All statistical tests were two-tailed and had alpha
set at .05 unless otherwise noted.
7. Results
Dogs in the experienced group (E) performed
above chance on all three of the back-turned point types
(one-sample t-tests, t (7) > 7.17, p < .001). Individually,
out of eight dogs, a total of seven successfully used the
dynamic tap, six dogs used the static proximal point,
and six dogs used the momentary distal point to locate
the target at above chance levels (individual binomial
tests, p < .05). Dogs in the no experience group (NE)
were also successful in using the back-turned dynamic
tap and static proximal point as a group (one sample t-
tests, t (7) > 13.75, p < .001), but did not perform above
chance on the back-turned momentary distal point
condition (one sample t-test, t (7) = .31, p = .77). See
figure 5. Individually, out of the eight dogs in each
condition from the no experience group, all eight
performed above chance (individual binomial tests, p <
.05) on the back-turned dynamic tap and static proximal
point conditions, while only two dogs performed above
chance (individual binomial tests, p < .05) on the back-
turned momentary distal point condition. Differences in
dogs’ average performance by point type (two-factor
ANOVA, F (2, 28) = 19.1, p < .001) and in interactions
between point type and prior experience (two-factor
ANOVA, F (2, 28) = 8.32 p = .001) were identified,
however no significant difference was found between
groups (experience vs no experience) when point types
were pooled (two-factor ANOVA, F (1, 14) = 3.03 p =
.10). However, when the mean group performances
were compared for the back-turned momentary distal
point alone, the group of dogs with prior experience
(8.25/10 correct) performed significantly better than the
naive dogs (5.25/10 correct) (t-test, t (7) = 3.21, p <
.01).
To assess performance differences predicted by the
orientation of the experimenter, the mean performances
of inexperienced dogs witnessing a dynamic tap, static
proximal point, and momentary distal point in
Experiment 1 (forward facing) and Experiment 3 (back-
turned) were compared using a two-factor between-
subject ANOVA. Although differences between the
point types themselves were found (F (2, 42) = 33.1 p <
.001), with dogs performing worst on the momentary
distal point independent of human orientation, a
significant difference was not found between the
performance of dogs in the forward-facing versus back-
turned groups (F (1, 42) = .064, p = .80). See figure 5.
A significant interaction effect between point type and
experimenter orientation was also lacking (F (2, 42) =
.192, p = .82).

Udell et al. / RACC, 2013, Vol. 5, N°2, 3-20
16
Figure 5. Role of experimenter attentional state on point-following performance in nieve and experienced dogs. The mean
number of correct choices out of ten for each back-turned point type and its forward facing counterpart are shown. NE
indicates groups of dogs with no prior experience on the task; these dogs only experienced the single point type indicated. E
indicates groups of dogs with prior experimental Experience; these dogs experienced all point types from Experiments 2 & 3,
however the data shows their first exposure to each particular point type. Solid line indicates 50% chance; error bars represent
+/- SEM. ** indicates one sample t-test, t (7) > 7.17, p < 0.001; * indicates one sample t-test, t (7) = 4.50, p = .002.
Therefore the attentional state of the experimenter
did not appear to significantly influence pet dog
performance in the context of the human- guided object
choice task. Dogs with no prior experience were likely
to succeed on the dynamic tap and static proximal point
conditions and fail on the momentary distal point
condition independent of experimenter orientation.
Dogs with more pointing experience were likely to
succeed on all point types independent of experimenter
orientation. This suggests that dogs do not rely on
traditional cues of cooperative intent (e.g. eye contact)
when responding to human points.
8. General Discussion
It has been suggested that pet dogs’ responsiveness
to human action, including their ability to follow a point
to a target, may contribute to their success in human
environments (Udell & Wynne, 2008). Our results
suggest that while many pet dogs can follow a wide
range of points made with the human arm and hand,
they also show different levels of responsiveness to
points that vary along dimensions of distance and
duration (and possibly to a lesser degree movement).
Therefore different forms of human point should not be
considered interchangeable, as small differences in
topography can have a significant impact on
performance (in some cases predicting success or
failure on the task). Likewise, individual variation
between dogs suggests that it may be equally
problematic to describe ‘dogs’ as proficient on point
following tasks; instead it would be more appropriate to
describe dogs (and other relevant species) as having the
capacity to succeed –or even excel- on human-guided
tasks assuming other lifetime variables (developmental
stage, life experience, home environment, and even
prior experimental exposure) are compatible with such
a response.
The momentary distal point was identified here as
the most challenging point for pet dogs to follow in an
object choice task, a finding that is widely supported in
the literature for the performance of pet dogs (Gásci et
al., 2009), shelter dogs (Udell et al., 2010b) and wolves
(Virányi et al., 2008). The fact that momentary points
eliminate the dog’s ability to view the point, or
discriminative stimulus, at the time of choice, adds a
memory component to the task. This might make the
task more challenging in ways that could be
systematically varied by an experimenter, allowing for
tests of the influence of memory on pointing tasks as
well as another method for assessing short term
memory in domestic dogs (and possibly other species as
well). Therefore momentary points may provide
interesting opportunities for additional study.
Additional experimental manipulations exploring

Udell et al. / RACC, 2013, Vol. 5, N°2, 3-20
17
the effect of distance between the stimulus (human
point) and the target might also be made. While a
change of just 40 cm in distance appeared to make an
significant difference for the performance of dogs in
this study (dogs performed better on points within 10
cm of the target, and worse on those 50 cm from the
target), it would be interesting to know what the
maximum limit for making a connection between the
point and a target might be and whether this could vary
by context or breed.
While the combination of the momentary and distal
components of a point led to the most challenging point
type across all three experiments, dogs were successful
in utilizing a range of other point types possessing
either the momentary or distal component in other
combinations. Therefore predicting the degree of
salience associated with a pointing stimulus may not be
as easy as calculating the sum of its parts.
In Experiment 2 we demonstrated that experience
acquired over the course of an experimental study can
prepare pet dogs to outperform naive dogs on an object
choice task utilizing human points (even when prior
experimental exposure was limited to points containing
some but not all of the stimulus properties associated
with more difficult points). This suggests that dogs can
and do rapidly learn to assimilate new gestures into
their behavioral vocabulary, and can acquire
appropriate responses to new gestures through the
process of generalization. It is possible that dogs might
also develop a learning set with respect to point
following tasks, and that with enough experience dogs
may quickly and seamlessly appear to be proficient at
responding appropriately in the presence of any gesture
within the context of an object choice task (where any
new discrimination can be learned on the first trial).
Determining if this is the case however, will require
further research.
Independent of the type(s) of learning taking place,
within-subject research intended to survey the domestic
dog’s spontaneous success on human-guided tasks (for
examples see Soproni et al., 2001; Virányi et al., 2008)
should carefully consider the effects of learning that
occur over the course of testing, not to mention a
lifetime of learning opportunities present in the pet
dog’s natural environment- the human home. Post-hoc
tests that compare a small portion of trials at the
beginning and end of an experiment after the fact may
not always be sufficient to accurately measure the
influence of learning within the course of an experiment
(Udell et al., 2010b).
It should be noted, however, that generalization
over the course of an experiment may be less likely in
studies utilizing stimuli that differ greatly from one
another (E.g. Miklósi et al., 1998; Udell et al., 2012;
Udell, Giglio et al., 2008). For example, conditions
utilizing a momentary distal point towards a target
followed by a condition using a foot point or a head
turn. While it is possible that subjects may still learn
something about the task (for example, that they should
generally attend to the experimenter’s behavior and
minimally approach one of the target objects each trial),
a drastic shift in stimulus form or location could
potentially decrease performance on the task in this case
(e.g. a dog may still attend to the experimenter’s
neutrally placed hand when the solution is to be found
by looking at the movements occurring with the
experimenter’s foot or head). In contrast, for
Experiment 2, the solution could always be found by
looking at the experimenter’s arm and hand. Indeed a
recent study by Elgier, Jakovcevic, Mustaca, and
Bentosela (2012) demonstrated that dogs who were first
allowed to follow proximal points later performed
above chance when presented with a novel cross-point
(both made with the experimenters arm and hand);
conversely dogs who instead had previous experience
with body position cues (where the experimenter stood
behind the target container) did not perform above
change when later presented with the cross-point. As in
the current study, this finding suggests that dogs can
show improved performance on novel gesture types due
to generalization (i.e. reinforced prior exposure to
simpler point types or gestures from earlier in testing);
however the degree of similarity between the stimuli
also seems to be relevant.
Interestingly, in Experiment 3, the attentional state
of the experimenter did not alter the performance of
dogs on the pointing task. This does not imply that
dogs are insensitive to attentional state or cooperative
actions. To the contrary, there is ample literature
demonstrating that dogs are more likely to approach an
attentive experimenter in tasks that are inherently
cooperative, such as begging tasks (Cooper et al., 2003;
Gácsi et al., 2003; Udell et al., 2011) and are more
likely to steal forbidden food when humans, who might
stop or punish the behavior, are inattentive (Bräuer et
al., 2004; Call et al., 2003). Instead dogs do not appear
to treat point following as a behavior requiring the
attention of the human, or in other words, the responses
of the dogs in Experiment 3 are not consistent with the
hypothesis that dogs view human points as an

Udell et al. / RACC, 2013, Vol. 5, N°2, 3-20
18
inherently cooperative gesture. Alternatively, dogs may
learn that human points are often useful even when not
intended for them, for example when a human points
out a ball to another dog at the park.
A related study recently found that when a pointing
human experimenter called a dog’s name in a
“cooperative tone of voice” dogs were more likely to
reliably follow their point to a target than when the
experimenter gestured towards the target “uttering a
forbidding command in a prohibitive tone of voice”
(Pettersson, Kaminski, Herrmann, & Tomasello, 2011,
p. 236). This finding suggests that additional cues, such
as tone of voice, may allow dogs to discriminate
between contexts where following a point might lead to
reward or punishment. However, while dogs’
performance fell to chance levels in the overtly
competitive situation presented by Pettersson et al.
(2011), the current study suggests that overt cooperative
cues (like eye contact) are unnecessary for above
chance performance.
From a learning perspective, it makes sense that
dogs display flexible point following behavior
independent of human attentional state. Payoff within
the human home may be available for following a point
even when the dog is not the intended recipient.
Pointing used to reprimand a child for dropping food on
the floor is no less laden with information than a point
intended by the human to alert the dog to the location of
the food. This sort of eavesdropping would allow
vigilant dogs to focus their attention on interesting or
important aspects of the environment as signaled by
humans, even when the intended recipient may be
another individual (dog, human, or otherwise).
Eavesdropping may play an especially important role
when interpreting the behavior of dogs who would
likely benefit most by responding to human gestures in
ways counter to the goals of the human (e.g. feral dogs
avoiding capture or harm at the hands of humans, or a
pet dog trying to avoid a bath or shot). This form of
response would not require the dog to understand or
even perceive the intent of the gesture – responding in
accordance with the outcomes of prior context specific
experiences may be sufficient to explain this behavior-
however the possibility that dogs understand the intent
of a gesture but ignore or act counter to it in cases
where it might be beneficial to do so cannot be ruled
out without further research.
While it is possible that some foundational
stimulus properties are necessary for any dog to utilize
a human point as a stimulus- for example, adequate
stimulus size given an individual’s visual acuity (Udell,
Giglio, et al., 2008), or sufficient contrast with a given
background (Lakatos, Dóka, & Miklósi, 2007)-
differences in individual experience with humans and
specific gesture types may account for much of the
variability seen in the literature to date. It is also
possible that the degree to which certain stimulus
dimensions are important could vary by developmental
factors, breed, or population. For example movement as
a stimulus dimension may be more relevant when
testing herding breeds (which have been bred for their
attentiveness to moving stimuli) and less important for
livestock guarding breeds (which are bred for an
inhibited response to movement) (Dorey et al., 2009).
However, these are important empirical questions for
future research.
Ultimately the findings of this study are consistent
with the broader literature on point following and the
use of referential stimuli not only by canids but also by
humans. There is ample evidence that both human
children and dogs learn and develop the ability to
respond to the stimuli of their social companions with
age and experience, both within species (e.g. children:
Carpenter et al., 1998; Mundy et al., 2007; dogs: Fox,
1969; Scott & Fuller, 1965) and between species (e.g.
Bentosela et al., 2008; Elgier et al., 2009; Dorey et al.,
2010). In fact, stimulus type predicts the performance of
young human children in object-choice tasks as well
(Lakatos et al., 2009). Considering the wide variety of
gestures that could be made with the human body, and
the impact that culture, environment, health, growth,
and coordination could have on a human’s gesturing
behavior, flexibility ¬– including the ability to learn and
modify responses to different human stimuli - could
provide many short and long term benefits compared
with a static ability to respond to specific gesture types.
However, more direct acknowledgment and study of the
impact that life experience, environment, and specific
stimulus properties have on the social behavior of pet
dogs may allow us to better appreciate and explain
individual differences as well as species or breed trends.
Additional systematic research on the proximate
variables that influence social behavior may also help
provide a better understanding of how other species,
including humans, develop a sensitivity to the gestures
of others within their lifetime – or even during the
course of experimental testing. A goal that can and
should fit hand in hand with the important research
being done from an evolutionary perspective.

Udell et al. / RACC, 2013, Vol. 5, N°2, 3-20
19
Acknowledgment
Financial support for this research was provided by
UF-Howard Hughes G.A.T.O.R. Program through a
research mentoring fellowship and grant awarded to
Monique A. R. Udell and through two undergraduate
research fellowships and grants awarded to Nathaniel J.
Hall and James Morrison. We thank the many dog
owners and participants, Camp Marlin Doggie Daycare,
Gainesville, Florida and Dogwood Park, Gainesville,
Florida for the use of their facilities, and eleven
anonymous researchers who contributed to the rankings
of point type difficulty utilized in this research.
References
Bentosela, M., Barrera, G., Jakovcevic, A., Elgier, A. M., &
Mustaca, A. E. (2008). Effect of reinforcement,
reinforcer omission and extinction on a communicative
response in domestic dogs (Canis familiaris).
Behavioural Processes, 78(3), 464-469.
doi:10.1016/j.beproc.2008.03.004
Bräuer, J., Call, J., & Tomasello, M. (2004). Visual
perspective taking in dogs (Canis familiaris) in the
presence of barriers. Applied Animal Behaviour Science,
88(3-4), 299-317. doi:10.1016/j.applanim.2004.03.004
Call, J., Bräuer, J., Kaminski, J., & Tomasello, M. (2003).
Domestic dogs (Canis familiaris) are sensitive to the
attentional state of humans. Journal of Comparative
Psychology, 117(3), 257-263. doi:10.1037/0735-
7036.117.3.257
Carpenter, M., Nagell, K., & Tomasello, M. (1998). Social
cognition, joint attention, and communicative
competence from 9 to 15 months of age. Monographs of
the Society for Research in Child Development, 63(4), i-
vi, 1-143.
Cooper, J. J., Ashton, C., Bishop, S., West, R., Mills, D. S.,
& Young, R. J. (2003). Clever hounds: social cognition
in the domestic dog (Canis familiaris). Applied Animal
Behaviour Science, 81(3), 229-244. doi:10.1016/S0168-
1591(02)00284-8
Coppinger, R., & Coppinger, L. (2001). Dogs: A Startling
New Understanding of Canine Origin, Behavior &
Evolution (1st ed.). Scribner, NY.
Dorey, N. R., Udell, M. A., & Wynne, C. D. (2009). Breed
differences in dogs sensitivity to human points: A meta-
analysis. Behavioural Processes, 81(3), 409-415.
doi:10.1016/j.beproc.2009.03.011
Dorey, N. R., Udell, M. A., & Wynne, C. D. (2010). When
do domestic dogs, Canis familiaris, start to understand
human pointing? The role of ontogeny in the
development of interspecies communication. Animal
Behaviour, 79(1), 37-41.
doi:10.1016/j.anbehav.2009.09.032
Elgier, A. M., Jakovcevic, A., Mustaca, A. E., & Bentosela,
M. (2009). Learning and owner-stranger effects on
interspecific communication in domestic dogs (Canis
familiaris). Behavioural Processes, 81(1), 44-49.
doi:10.1016/j.beproc.2008.12.023
Fox, W. M. (1969). Behavioral Effects of Rearing Dogs With
Cats During the 'Critical Period of Socialization'.
Behaviour, 35(3-4), 273-280.
Gácsi, M., Kara, E., Belenyi, B., Topál, J., & Miklósi, A.
(2009). The effect of development and individual
differences in pointing comprehension of dogs. Animal
Cognition, 12(3), 471-479. doi:10.1007/s10071-008-
0208-6
Gácsi, M., Miklósi, Á., Varga, O., Topál, J., & Csányi, V.
(2003). Are readers of our face readers of our minds?
Dogs (Canis familiaris) show situation-dependent
recognition of human’s attention. Animal Cognition, 7,
144–153. doi:10.1007/s10071-003-0205-8
Goldin-Meadow, S. (2007). Pointing Sets the Stage for
Learning Language—and Creating Language. Child
Development, 78(3), 741-745. doi:10.1111/j.1467-
8624.2007.01029.x
Gómez, J. (2007). Pointing Behaviors in Apes and Human
Infants: A Balanced Interpretation. Child Development,
78(3), 729-734. doi:10.1111/j.1467-8624.2007.01027.x
Hare, B., Brown, M., Williamson, C., & Tomasello, M.
(2002). The Domestication of Social Cognition in Dogs.
Science, 298(5598), 1634-1636.
doi:10.1126/science.1072702
Hare, B., Rosati, A., Kaminski, J., Bräuer, J., Call, J., &
Tomasello, M. (2010). The domestication hypothesis for
dogs' skills with human communication: a response to
Udell et al. (2008) and Wynne et al. (2008). Animal
Behaviour, 79(2), E1-E6.
doi:10.1016/j.anbehav.2009.06.031
Lakatos, G., Dóka, A., & Miklósi, Á. (2007). The Role of
Visual Cues in the Comprehension of the Human
Pointing Signals in Dogs. International Journal of
Comparative Psychology, 20(4). Retrieved from:
http://escholarship.org/uc/item/1vm469ff
Lakatos, G., Soproni, K., Dóka, A., & Miklósi, A. (2009). A
comparative approach to dogs' (Canis familiaris) and
human infants' comprehension of various forms of
pointing gestures. Animal Cognition, 12(4), 621-631.
doi:10.1007/s10071-009-0221-4
Maginnity, M. (2007). Perspective Taking and Knowledge
Attribution in the Domestic Dog (Canis familiaris): A
Canine Theory of Mind? University of Canterbury,
Psychology Thesis. doi:10092/1439
Miklósi, Á., Polgardi, R., Topál, J., & Csányi, V. (1998). Use
of experimenter-given cues in dogs. Animal Cognition,
1(2), 113-121. doi:10.1007/s100710050016
Miklósi, Á., & Soproni, K. (2006). A comparative analysis of
animals' understanding of the human pointing gesture.
Animal Cognition, 9(2), 81–93.
Miklósi, Á., Topál, J., & Csányi, V. (2004). Comparative
social cognition: what can dogs teach us? Animal

Udell et al. / RACC, 2013, Vol. 5, N°2, 3-20
20
Behaviour, 67(6), 995-1004.
doi:10.1016/j.anbehav.2003.10.008
Miklósi, Á., Topál, J., & Csányi, V. (2007). Big thoughts in
small brains? Dogs as a model for understanding human
social cognition. NeuroReport, 18(5), 467-471.
doi:10.1097/WNR.0b013e3280287aae
Mundy, P., Block, J., Delgado, C., Pomares, Y., Van Hecke,
A. V., & Parlade, M. V. (2007). Individual Differences
and the Development of Joint Attention in Infancy.
Child Development, 78(3), 938-954. doi:10.1111/j.1467-
8624.2007.01042.x
Nobis, G. (1979). Der älteste Haushunde lebte vor 14000
Jahren. [The oldest domestic dog lived 14,000 years
ago] Umschau 79, 610.
Pettersson, H., Kaminski, J., Herrmann, E., & Tomasello, M.
(2011). Understanding of human communicative
motives in domestic dogs. Applied Animal Behaviour
Science, 133(3–4), 235–245.
doi:10.1016/j.applanim.2011.05.008
Riedel, J., Schumann, K., Kaminski, J., Call, J., & Tomasello,
M. (2008). The early ontogeny of human-dog
communication. Animal Behaviour, 75(3), 1003-1014.
doi:10.1016/j.anbehav.2007.08.010
Scott, J. P., & Fuller, J. L. (1965). Genetics and the Social
Behavior of the Dog (1st ed.). Chicago, IL.: University
Of Chicago Press.
Soproni, K., Miklósi, A., Topál, J., & Csányi, V. (2001).
Comprehension of human communicative signs in pet
dogs (Canis familiaris). Journal of Comparative
Psychology, 115(2), 122-126.
Soproni, K., Miklósi, A., Topál, J., & Csányi, V. (2002).
Dogs' (Canis familiaris) responsiveness to human
pointing gestures. Journal of Comparative Psychology,
116(1), 27-34.
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(2), 141-152. doi:10.1016/S0168-
1591(03)00114-X
Tomasello, M., Carpenter, M., & Liszkowski, U. (2007). A
New Look at Infant Pointing. Child Development, 78(3),
705-722. doi:10.1111/j.1467-8624.2007.01025.x
Udell, M. A. R., Dorey, N. R., & Wynne, C. D. L. (2008).
Wolves outperform dogs in following human social
cues. Animal Behaviour, 76, 1767–1773.
doi:10.1016/j.anbehav.2008.07.028
Udell, M. A. R, Dorey, N. R., & Wynne, C. D. L (2010b).
The performance of stray dogs (Canis familiaris) living
in a shelter on human-guided object-choice tasks.
Animal Behaviour, 79(3), 717-725.
doi:10.1016/j.anbehav.2009.12.027
Udell, M. A. R., Dorey, N. R., & Wynne, C. D. L. (2010a).
What did domestication do to dogs? A new account of
dogs' sensitivity to human actions. Biological Reviews,
85(2), 327-345. doi:10.1111/j.1469-185X.2009.00104.x
Udell, M. A. R., Dorey, N. R., & Wynne, C. D. L. (2011).
Can your dog read your mind? Understanding the causes
of canine perspective taking. Learning & Behavior.
doi:10.3758/s13420-011-0034-6
Udell, M. A. R, Giglio, R. F. & Wynne, C. D. L. (2008).
Domestic dogs (Canis familiaris) use human gestures
but not nonhuman tokens to find hidden food. Journal of
Comparative Psychology, 122, 84-93. doi:
10.1037/0735-7036.122.1.84.
Udell, M. A. R., Spencer, J. M., Dorey N. R. & Wynne, C.
D. L. (2012). Human-Socialized Wolves Follow Diverse
Human Gestures And They May Not Be Alone.
International Journal of Comparative Psychology,
25(2), 97–117.
Udell, M. A. R., & Wynne, C. D. L. (2008). A Review of
Domestic Dogs’ (Canis Familiaris) Human-Like
Behaviors: Or Why Behavior Analysts Should Stop
Worrying and Love Their Dogs. Journal of the
Experimental Analysis of Behavior, 89(2), 247–261.
doi:10.1901/jeab.2008.89-247
Udell, M. A. R., & Wynne, C. D. L. (2010). Ontogeny and
phylogeny: both are essential to human-sensitive
behaviour in the genus Canis. Animal Behaviour, 79(2),
e9-e14.
Wynne, C. D. L., Udell, M. A. R., & Lord, K. A. (2008).
Ontogeny's impacts on human-dog communication.
Animal Behaviour, 76(4), e1-e4.
doi:10.1016/j.anbehav.2008.03.010
Virányi, Z., Gácsi, M., Kubinyi, E., Topál, J., Belényi, B.,
Ujfalussy, D., & Miklósi, Á. (2008). Comprehension of
human pointing gestures in young human-reared wolves
(Canis lupus) and dogs (Canis familiaris). Animal
Cognition, 11(3), 373-387. doi:10.1007/s10071-007-
0127-y.