Journal of Veterinary Behavior 10 (2015) 103e110

Contents lists available at ScienceDirect

Journal of Veterinary Behavior

journal homepage: www.journalvetbehavior.com

Research

Validation of a selection protocol of dogs involved in animal-assisted

intervention
Paolo Mongillo a, Elisa Pitteri a, Serena Adamelli b, Sabrina Bonichini c, Luca Farina d,
Lieta Marinelli a, *
a

Laboratory of Applied Ethology, Department of Comparative Biomedicine and Food Science, University of Padova, Legnaro, Padua, Italy
CSC S.r.l.dCentro di Scienze Comportamentali del Cane, Padova, Italy
Department of Developmental Psychology and Socialisation, University of Padova, Padua, Italy
d
National Reference Centre for Animal Assisted Interventions, Istituto Zooprolattico Sperimentale delle Venezie, Legnaro, Padua, Italy
b
c

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 19 August 2014
Received in revised form
3 November 2014
Accepted 6 November 2014
Available online 22 November 2014

Animal-assisted interventions (AAIs) with dogs are becoming popular worldwide, but there is a lack of
scientic data on dog selection procedures, which prevents the organizations involved from adopting a
uniform assessment method. In the absence of legal regulations and common guidelines, dogs currently
engage in diverse activities, some of which may pose more of a concern for their welfare than others. The
present study sought to standardize and validate a selection protocol of dogs involved in AAIs and to
make it available to interested Italian institutions. To meet this aim, we enrolled dogs already working in
AAIs and qualied by their handlers as fully suitable (N 20) or suitable with reserve (N 20) and
nonworker pet dogs (N 20) in this study. Each dog underwent a behavioral examination followed by
role-playing, simulating an AAI session, presenting various conditions, and unexpected stimuli possibly
occurring in AAIs. In both procedures, blinded experts judged the suitability of dogs evaluating
controllability, reliability, and predictability of their social behavior and considering possible concerns
regarding safety or welfare of patients and of the dogs themselves. Concurrent validity between procedures was fair, whereas reasons for dogs allocation resulted in moderate accordance for dogs being
aggressive, fearful, or avoidant of an unknown person. Moreover, dogs judged suitable, suitable with
reserve, or not suitable by the experts signicantly differed for the relative duration of negative interactions with unknown person, fear, and aggressiveness expressed in the role-playing. Differences in
the ability to cope with stressful situations possibly occurring in AAIs were unnoticeable with the present
protocol, and stress signals shown by dogs during the role-playing were judged by the experts or by the
handlers not to differ between dogs. Given our results, the present protocol could be easily and properly
adopted to identify dogs behavioral prerequisites for AAIs. For the procedure to work properly, each dyad
(dog and handler) should undergo behavioral examination and role-playing simulation in sequence.
2015 Elsevier Inc. All rights reserved.

Keywords:
animal assisted intervention
behavior
behavioral examination
dog
role-playing
selection protocol

Introduction
Animal-assisted interventions (AAIs) are achieving a certain
level of recognition worldwide and this is accompanied by a
growing body of research on the effect of these programs on

* Address for reprint requests and correspondence: Lieta Marinelli, DVM, PhD,
Laboratory of Applied Ethology, Department of Comparative Biomedicine and Food
Science, University of Padova, Viale dellUniversit 16, Legnaro, Padua 35020, Italy.
Tel: (39) 049 641219; Fax: (39) 049 641174.
E-mail address: lieta.marinelli@unipd.it (L. Marinelli).
http://dx.doi.org/10.1016/j.jveb.2014.11.005
1558-7878/ 2015 Elsevier Inc. All rights reserved.

human health and well-being (Urbanski and Lazenby, 2012;

Bernabei et al., 2013; Marcus, 2013; OHaire, 2013). In 2012, the
Italian National Reference Centre for Animal Assisted Interventions
conducted a nationwide survey of organizations offering this
program, nding that such programs were widely distributed
throughout the country and frequently involved dogs (CRN, 2012).
The widespread involvement of dogs in AAIs is embedded in the
outstanding interspecic social ability of this species and in the
dogs ease in adapting to various human environments (Miklsi
and Topl, 2013). Nonetheless, not every dog is suitable to be
involved in AAIs and both international associations (IAHAIO,

104

P. Mongillo et al. / Journal of Veterinary Behavior 10 (2015) 103e110

1998; AVMA, 2014), and Italian authorities (CNB, 2005; DGR n.

4130, 2006; DM G.U., 2009) have noted specic requirements to
be fullled.
Procedures for the selection and screening of dogs involved in
AAIs (thereafter referred as therapy dogs) are more or less
formalized by a number of therapy animal registration organizations (e.g., AKC, 2014). Although diverse criteria are listed by each
organization, most of them follow a general approach, which takes
into account prerequisites, appropriate training, supervised working experiences, and evaluations by experts observing the dogs and
their handlers operating in AAI settings. Essential minimum prerequisites are good health of dogs, absence of behaviors that could
jeopardize participants in AAIs (jumping up or on, mouthing, biting,
dodging, or apparent aggression), and favorable disposition to
interact with unknown people. Having good basic obedience skills
and living with the handler or being familiar with the handler for a
certain period are other common prerequisites. Appropriate
training is aimed at preparing the dog (and the handler) for the
evaluation process used by the organization. Being comfortable in a
medical setting; accepting rough handling, uncontrolled vocalizations, or approach by strangers; being surrounded by a group of
people; disregarding food or toys on cue; and, in general, being able
to cope with stressful situations possibly occurring in AAIs are expected skills of a therapy dog. Moreover, interaction with patients
should appear enjoyable for the dogs, and the dogs must not show
signs of stress, fear, aggression, or shyness, nor should they attempt
to avoid touch.
The great deal of effort put by these organizations into standardizing procedures has provided the framework guiding the selection procedure; however, limitations exist. First, each
organization interprets the requirements from their own perspective thus reecting the biases of that particular association
(Fredrickson-MacNamara and Butler, 2006). Second, in some cases,
organizations have an active role in educating the dog-handler dyad
before they undergo the evaluation process. In spite of the good
faith of the people involved in the training and evaluation practice,
there is no independent objective verication, a concern for reliability across many working-dog disciplines. Third, none of these
selection procedures have been assessed for their scientic validity
to the best of our knowledge. Invalid assessment of dog suitability
to task would pose signicant welfare concerns for therapy dogs
and participants. Moreover, the lack of scientic data on selection
procedures prevents organizations involved in AAIs from recognizing a uniform method of selection and standardizing procedures
in a way that could benet them all.
The present study sought to standardize and validate a protocol
for the selection of therapy dogs, making it available to Italian institutions and organizations working in this eld. As a rst attempt
to validate the selection protocol, we focused on the evaluation of
the behavioral prerequisites (i.e., sociability toward strangers and
absence of behaviors threatening the patients and dogs safety or
welfare). Assessment of these prerequisites was performed through
a behavioral examination and the evaluation of the dogs behavior
in a role-playing simulating an AAI session. Both evaluations are
crucial and represent, respectively, the initial bottleneck and the
nal outcome of the entire process. Because these behavioral prerequisites have to be consistent in a therapy dog, concurrent validity between the outcomes of the 2 evaluations was expected.
Validation also required estimating the sensibility and specicity of
the evaluations. This should be performed by comparing the
outcome of a procedure with a gold standard; however, in the
absence of such standard, we compared the judgments of both
evaluations with the objective assessment of the dogs undesirable
behavior.

Materials and methods

Behavioral examination
Evaluation of the dogs prerequisites through a behavioral examination was the rst step of our evaluation procedure. Clinical
assessments were performed at the Animal Behavior Service of the
Department of Comparative Biomedicine and Food Science (University of Padova). During the examination, the same veterinary
behaviorist evaluated both the history and present behavior of the
dog, being unaware of the specic working experience of the dog.
The veterinarian was asked to evaluate the possible presence of
behavioral problems and the dogs suitability to work in AAIs.
Suitability of dogs was evaluated on controllability, reliability, and
predictability of their social behavior and additionally a possible
concern regarding safety or welfare of patients; safety and welfare
of dogs were also taken into account. The level of technical skills
was not accounted for suitability. At the end of the behavioral examination, the veterinarian assigned the dog to one of the following
categories: unsuitable (UN; dog with possible behavioral problems
and/or concern regarding safety or welfare of patient and/or dog),
suitable with reserve (RV; dog with fair social behavior and without
concern regarding safety or welfare of patient and/or dog), and
suitable (SU; dog with good or excellent social behavior and
without concern regarding safety or welfare of patient and/or dog).
A detailed description of reasons for the allocation was written for
each dog.
Role-playing
The same dog-person dyads were enrolled in a role-play,
simulating an AAI session; in brief, each dog-handler pair was
exposed to various conditions and unexpected stimuli that could
occur during clinical AAI experiences. This second step of the procedure included aspects that could be dened as critical for both the
handler and the animal with regard, in particular, to the maintenance of both the animals and the patients welfare. The roleplaying was performed in the Laboratory of Applied Ethology
(Department of Comparative Biomedicine and Food Science, University of Padova), where a closed-circuit television system allowed
technicians to supervise both recording and timing of the
procedure.
Setup of the role-playing
The environments in which a dog-handler dyad performs AAI
usually vary depending on distractions presented, familiarity of
environment by the dog, and physical characteristics. The Delta
Society (Delta Society, 2002) indicated the following common distracters should be used: food; other people; unexpected, loud, and
sharp noises; and an unexpected approach and interaction with a
stranger in which the stranger pets the dog. We built the roleplaying episodes following the possible situations given below.
To this aim, we selected specic materials (crutches, metal
walker, chairs, table, sunglasses, noisy bracelet, white coat, dog
brush, little ball, chocolate chips) that are common elements in AAI
environments and may constitute sources of distraction and
awkwardness for the dog. One dog brush and 1 ball, widely used by
handlers during AAI sessions to perform various activities, were
made available for the entire time of the simulation.
Besides these materials and based on our clinical experience in
AAI (Marinelli et al., 2009), we selected specic challenging behaviors of both the patient and his or her accompanying health care
professional that can occur during an AAI. For instance, autistic,
schizophrenic, or disabled patients can be uncooperative during the

P. Mongillo et al. / Journal of Veterinary Behavior 10 (2015) 103e110

activities, showing freezing or closure behavior. On the other side,

patients could interact inappropriately, being uncontrolled, noisy,
clumsy, disturbing, and unsafe for themselves and the animal. The
health professional can behave inadequately too by interfering with
the activities and worsening the situation for the dog-handler dyad.
To create the aforementioned situations in the role-play, we trained
3 individuals acting in the simulation: (1) an actor wearing sunglasses, simulating the behavior of a patient with physical and
psychological difculties and an incorrect approach to the animal;
(2) a person wearing a white coat, playing the role of a health care
provider and showing a generally passive and inappropriate attitude during the session; (3) a stranger acting as a distractive and
disturbing presence at a certain point during the AAI session. The
same actors participated in all the simulations and the simulation
setting was standardized and identical for all participants.
The role-playing episodes
The procedure lasted approximately 10 minutes and was articulated in 5 different episodes (see the Supplementary Video). Each
of them had a xed duration, signaled by a technician who calls to
the actors by headphones the starting and ending point of each
episode. All procedures were recorded for subsequent evaluation
and data collection.
Before starting, each handler was given the following information: We will make a simulation session of AAI. You will be left
alone with your dog in the room, allowing you to prepare yourself.
When you feel ready, please raise your hand and we will start the
procedure. Please feel free to do whatever you judge appropriate
during the entire session.
After the agreed signal, the patient and his health care provider
entered the room.
If requested by the dog handler, the health care provider
described the patients condition, always giving the same information: Following a bad accident, the patient reported a head
injury resulting in mental confusion, mood changes, memory
problems, and clumsiness.
Episode 1: freezing and stranger (3 minutes). The patient sat on the
chair after entering the room with the walker, accompanied by the
health care provider, who carried the crutches. The patient initially
stayed in freezing. After 2 minutes, a stranger suddenly opened the
door and said Sorry, I entered the wrong room! Then, he immediately left the room. In the meantime, the patient continued
staying in freezing. The health care provider did not interact.
Episode 2: awkward approach (3 minutes). At the end of freezing,
the patient performed a 3-minute awkward approach to the animal
during which he carried out the following actions: running around
the room, approaching the dog frontally and patting the dog on the
head, hugging the dog, clapping hands with outstretched arms,
sitting on the chair, rocking on the chair, moving the chair noisily,
sitting on the oor and brushing the dog, throwing away the brush,
patting on the head, screaming, clapping hands with outstretched
arms. In each simulation, the behaviors described were performed
in the same order and for the same duration.
Episode 3: seizing of the dog (30 seconds). The patient tried to seize
the dog by the harness. This phase was shorter than 30 seconds if
the handler intervened to protect the animal.
Episode 4: semifreezing (2 minutes). The patient entered a new
phase of semifreezing. The patient did not take any initiative and
was quite passive but looked at the activities proposed by the
handler even if barely collaborating. After 2 minutes, the health care
provider (who had never intervened so far) changed his behavior

105

toward the dog from passive to active, in particular calling or

petting the dog and drawing attention toward himself.
Episode 5: simultaneous interaction (2 minutes maximum). The patient performed the same actions as the health care provider,
challenging the dog with multiple petting. This phase of interaction
could end before 2 minutes on the handlers request.
The role-playing evaluation
Recorded role-playing simulations were evaluated by a second
veterinary behaviorist, who was unaware of the evaluation obtained by the dog in the behavioral examination previously carried
out and of the self-reported level. The veterinarian was asked to
judge the dogs suitability to work in an AAI. Again, the level of
technical skills was not accounted for such judgment, whereas
controllability, reliability, and predictability of the dogs social
behavior were accounted; moreover, possible concerns regarding
safety or welfare of patients and dogs were gathered from behaviors expressed during the simulation. Each dog was evaluated as UN
if the dog expressed behaviors threatening the actors safety or
welfare and/or there was concern regarding the welfare of the dog,
RV if the dog had fair social behavior and there was no concern
regarding the safety or welfare of the actors and the dog, and SU if
the dog had good or excellent social behavior and there was no
concern regarding the safety or welfare of the actors the and dog. In
this case too, a detailed written description of reasons for dog
allocation was required.
Participants
For validation purposes, the protocol should be administered to
a sample of dogs representative of the range of expression of the
assessed behaviors. Accordingly, we recruited therapy dogs with
different levels of suitability in AAI as well as nonworker pet dogs.
Therapy dogs were qualied by their handlers as fully suitable or
suitable with reserve, the latter being incapable of facing challenging interventions. Recruitment of participants was carried out
over a 2-year period among private citizens and professional organizations or individuals providing AAIs. For the inclusion in the
study, dogs had to be aged between 1 and 10 years and the
participating owner or handler had to be their primary caretaker. In
addition, nonworker pet dogs were required to have been living
with their current owner for at least 1 year, and therapy dogs were
required to have been living and/or working mostly with their
current handler for at least 1 year. The last condition was set
because, in some cases, dogs were owned by the organization and
occasionally the same dog could participate in AAIs with different
handlers. To exclude overt medical conditions that might affect
their behavior during the procedure, all dogs underwent a standard
medical examination. Dogs considered dangerous on the basis
direct observation or a biting history toward humans were excluded
from the procedure.
Recruitment ended when 3 groups had been composed:
nonworker pet dogs (nonworking [NW]; mean age  standard
deviation [SD] 3.6  2.4 years; N 20), dogs involved in AAI and
considered fully suitable by their handler (SU; mean age  SD 4.9
 2.2 years; N 20), or suitable with reserve (RV; mean age  SD
3.8  2.6 years; N 20). Fully SU dogs were working in AAIs for an
average of 3.2  1.9 years and RV dogs for 2.4  2.0 years. The SU
group was composed of 7 male (3 neutered) and 13 female dogs (8
spayed); the RV group contained 6 male (1 neutered) and 14 female
dogs (7 spayed). The NW group contained 6 male and 14 female
dogs (7 spayed). There were 21 mongrels and 39 purebred dogs.
Represented breeds included Labrador retriever (N 12), golden
retriever (N 6), cocker spaniel (N 4), American Staffordshire

106

P. Mongillo et al. / Journal of Veterinary Behavior 10 (2015) 103e110

terrier (N 3), Bolognese (N 2), Siberian Husky (N 2), Bernese

mountain dog (N 2), Border collie (N 1), German shepherd (N
1), Belgian shepherd (N 1), West Highland white terrier (N 1),
at-coated retriever (N 1), dachshund (N 1), Lupo Italiano (N
1), and Argentine dogo (N 1).

showed corresponding behavioral parameter) of reasons for allocation of both evaluations were assessed.
Unless otherwise stated, behavioral parameters were reported
as medians (1st-3rd quartiles). All analyses were carried out with
SPSS (SPSS Statistics, version 20.0; IBM Corp., Armonk, NY), and the
statistical signicance was set at 5%.

Data collection and statistical analysis

Concordance of judgment between behavioral examination and
evaluation of role-playing was analyzed using Cohen kappa coefcient comparing across coding groups (SU, RV, UN). Inter-rater
reliability of dogs judgment in role-playing was assessed between the groups (SU, RV, UN) as evaluated by a third blind
observer (N 21 videos; 7 per group of self-reported level) and
found to have a high level of agreement (Cohen kappa 0.807).
Inter-rater reliability of dogs judgment in behavioral examination
was inapplicable.
An in depth examination of written reasons for allocations
identied 6 main reasons that emerged from the behavioral examination (dog being fearful, stressed, excited, aggressive toward
people, aggressive toward other dogs, and avoiding contact with
unknown person) and 5 main reasons that emerged from the roleplaying (dog being fearful, stressed, aggressive toward people, uninterested in interaction with unknown person, and avoiding
contact with unknown person). Reasons were recorded as dichotomous variables (0 absent; 1 present) for each dog, and
concordance of common reasons between examination and roleplaying was analyzed computing Cohens kappa coefcients between the coding (0, 1) of corresponding reasons (Altman, 1991).
Video recordings of role-playing simulations were imported into
Noldus Observer XT software. Data were collected from all videos
by continuous sampling on focal subjects by the third experimenter.
During the entire simulation, the experimenter coded the expression of problematic behaviors threatening the safety or welfare of
the actors or the dog (fear, aggressiveness), and the relative percentages of the time spent in these behaviors were computed. Time
spent interacting with the actors was also coded, and the percentage of time of negative interaction (the dog avoids the stranger
trying to interact or the dog freezes and stiffens while the stranger
interacts) was computed. Finally, stress signals were identied and
counted throughout the role-playing and their total number was
computed and normalized on the total duration of the simulation. A
fourth observer coded one-third of the videos (N 21, 7 per group
of self-reported level), and inter-observer reliability was found to be
good for all parameters (Spearman rho >0.7; P < 0.01 in all cases).
The normality of behavioral parameters was assessed with the
Kolmogorov-Smirnov test, and because none of the behavioral parameters were normally distributed, nonparametric statistical tests
were used for all analyses. With group allocation as a 3-level factor,
dogs behavioral parameters (the relative percentage of the time
spent being fearful, aggressive, interacting negatively with actors,
and the normalized total number of stress signals) were compared
with the Kruskal-Wallis test, appropriately followed by pairwise
comparisons. This analysis was run separately for group allocation
according to behavioral examination, role-playing evaluation, and
self-reported level. Moreover, the Mann-Whitney test was used to
assess differences between dogs with and without a specic reason
for allocation in the correspondent behavioral parameter coded
from the video (e.g., reason of allocation being fearful and duration
of fear). Specicity (number of dogs without a specic reason that
did not show the corresponding behavioral parameter/number of
dogs with or without that reason that did not show the corresponding behavioral parameter) and sensitivity (number of dogs
with a specic reason that showed corresponding behavioral
parameter/number of dogs with or without that reason that

Results
Concordance among suitability assessments
According to the reasons provided by the veterinarian, dogs
judged UN at the examination were fearful (N 7), stressed (N 4),
aggressive (N 6), and/or avoidant of the unknown person (N 6).
Being fearful (N 3), stressed (N 9), excitable (N 12), and/or
avoidant of the unknown person (N 2) were the main reasons to
be allocated to RV. Dogs judged UN and RV presented more than 1
reason for allocation in 76.9% (10 of 13) and 42.1% (8 of 19) of cases,
respectively.
Role-playing allocated dogs to UN because of fear (N 3), stress
(N 11), aggressive behavior (N 3), absence of interest in interaction with the unknown person (N 3), and/or avoidance of the
unknown person (N 2). Fear (N 8), stress (N 4), absence of
interest in interaction with the unknown person (N 4), and/or
avoidance of the unknown person (N 7) were the main reasons to
allocate dogs to RV. Dogs judged UN and RV in the role-playing
presented more than 1 reason for allocation in 80.0% (4 of 5) and
50.0% (10 of 20) of cases, respectively.
Reasons for allocation from the examination and role-playing
resulted in fair-to-moderate accordance for dogs being aggressive
(Cohen kappa 0.397), fearful (Cohen kappa 0.444), or avoidant
of the unknown person (Cohen kappa 0.521), whereas no
agreement was found for dogs being stressed (Cohen kappa
0.109).
The number of dogs judged SU, RV, and UN according to examination or role-playing and correspondent self-reported level by
handlers is listed in Tables 1 and 2. The level of concordance between judgments of the 2 evaluations was fair (examination and
role-playing: Cohen kappa 0.310).
Of self-reported SU dogs, 85 % (N 17) were judged suitable in
both evaluations and in 75% (N 15) of the cases were the same
dogs. Notably, 1 of the 2 dogs self-reported as RV was considered
UN by both evaluations because of aggressive behavior toward
people, which was evident during the role-playing simulation. The
other self-reported RV dog was judged UN at the examination for a
suspected phobia of conned spaces. Nine dogs judged UN at the
clinical assessment were considered SU (N 2) or RV (N 7) according to role-playing. Three of them were considered UN in the

Table 1
Number of dogs judged suitable (SU), suitable with reserve (RV), and unsuitable (UN)
according to the behavioral examination or the role-playing and the correspondent
self-reported level by handlers
Type of evaluation

Suitability level

Self-reported
SU

RV

NW

Total

Behavioral examination

SU
RV
UN
Total

17
3
0
20

10
8
2
20

1
8
11
20

28
19
13
60

Role-playing

SU
RV
UN
Total

17
3
0
20

15
4
1
20

3
13
4
20

35
20
5
60

NW refers to the group of nonworker pet dogs enrolled in the study.

P. Mongillo et al. / Journal of Veterinary Behavior 10 (2015) 103e110

Table 2
Number of dogs judged suitable (SU), suitable with reserve (RV), and unsuitable (UN)
according to behavioral examination and role-playing
Type of evaluation

Behavioral examination

Suitability level

SU
RV
UN
Total

Role-playing

Total

SU

RV

UN

23
10
2
35

5
8
7
20

0
1
4
5

28
19
13
60

examination because of intraspecic aggressiveness, 5 were

suspected of poor socialization toward humans, and 1 was the
aforementioned dog phobic of conned spaces.
Behavioral parameters of dogs according to suitability assessments
Statistical analysis on coded behavioral parameters indicated
that group allocation did not affect relative expression of stress
signals, regardless of the type of assessment. In contrast, group
allocation signicantly affected the relative duration of negative
interaction, fear, and aggressiveness (Table 3).
As expected, variability of behavioral parameters within NW
dogs was higher than within self-reported RV or SU dogs with the
exception of the number of stress signals emitted. Compared to selfreported RV dogs, dogs self-reported as SU did not express
any difference in coded behavioral parameters. However, aggressiveness and fear expressed by self-reported RV dogs did not differ
from NW dogs.
Allocation of dogs to suitability groups according to the behavioral examination resulted in consistent differences between UN
dogs and both SU and RV dogs in the relative duration of both fear
and aggressiveness (P < 0.001 for all comparisons between UN vs.
SU and P 0.02 for all comparison between UN vs. RV); negative
interaction was different only between UN and SU dogs (P < 0.001).
None of the behavioral parameters were different between SU
and RV.
Evaluation of role-playing was the only assessment that clearly
differentiated SU from RV dogs at least for fear (P 0.001) and
negative interaction (P < 0.001). Nevertheless, UN dogs did not
differ from RV dogs in both parameters during role-playing,

107

whereas UN dogs aggressiveness was signicantly longer than

that of SU (P < 0.001) and RV (P 0.02) dogs.
Dogs judged stressed at the examination or during the roleplaying were not signicantly different from dogs that did not
present this reason for allocation in the expression of stress (total
number of stress signals: behavioral examination, 23.5% [16.0%31.0%] vs.17.5% (15.5%-58.3%); Mann-Whitney test, P 0.642;
role-playing, 27.0% [17.0%-31.0%] vs. 21.0% [15.5%-32.5%]; MannWhitney test, P 0.313). A longer expression of fear was present
when being fearful was the reason given by the behavioral examination (9.7% [0.0%-28.8%] vs. 0.0% [0.0%-0.0%]; Mann-Whitney test,
P < 0.001) and by the role-playing (1.4% [0.0%-22.8%] vs. 0.0% [0.0%0.0%]; Mann-Whitney test, P < 0.001). The specicity of the reason
being fearful was 0.94 for each of the 2 evaluations and 0.96
considering data of both evaluations together. The sensitivity of
being fearful reached 0.54 for each evaluation and 0.77 for the 2
together. Aggressive behaviors toward the actors were longer in
dogs with aggressiveness as a reason given by the role-playing
and the behavioral examination than in dogs without this reason
(examination: 4.2% [1.8%-8.9%] vs. 0.0% [0.0%-0.0%]; Mann-Whitney
test, P < 0.001; role-playing: 0.4% [0.0%-8.7%] vs. 0.0% [0.0%-0.0%];
Mann-Whitney test, P < 0.05). The reason aggressiveness had a
specicity and sensitivity of 0.96 and 0.50 in the behavioral examination and 0.96 and 0.75 in the role-playing evaluation. Taking
together data of both evaluations, specicity and sensitivity
reached 0.94 and 0.88, respectively. Finally, the reason being
avoidant of the unknown person according to the behavioral examination and the role-playing predicted a longer negative interaction with the actors (examination: 8.1% [5.2%-10.2%] vs. 1.8%
[0.4%-4.0%]; Mann-Whitney test, P < 0.001; role-playing: 6.6%
[3.7%-9.5%] vs. 1.8% [0.4%-5.3%]; Mann-Whitney test, P 0.006).
Specicity and sensitivity could not be calculated here because dogs
that did not avoid the actors for the entire duration of the simulation were too few (N 8).
Discussion
The results of the present study show that our protocol,
including an assessment of dogs behavioral prerequisites and an
in-context evaluation, could clearly identify dogs expression of

Table 3
Median (1st-3rd quartile) of the relative duration of fear, aggressiveness, negative interaction, and the total number of stress signals in dog groups (SU, suitable; RV, suitable
with reserve; UN, unsuitable) allocated by self-reported level, role-playing, and behavioral examination
Behavioral parameter

Self-reported
SU

Fear (%)
Aggressiveness (%)
Negative interaction (%)
Stress signals (N)

0.0
0.0
1.0
20.5

NW

RV
(0.0-0.0)A
(0.0-0.0)A
(0.3-2.7)A
(9.0-44.0)A

0.0
0.0
1.8
21.0

(0.0-0.5)AB
(0.0-0.0)AB
(0.3-4.9)A
(6.0-78.0)A

0.2
0.0
6.2
25.5

(0.0-8.9)B
(0.0-2.7)B
(3.5-9.5)B
(9.0-66.0)A

Role-playing
SU
Fear (%)
Aggressiveness (%)
Negative interaction (%)
Stress signals (N)

0.0
0.0
1.0
20.0

P
RV

(0.0-0.0)A
(0.0-0.0)A
(0.2-3.4)A
(6.0-56.0)A

0.5
0.0
5.5
26.5

UN
(0.0-8.3)B
(0.0-0.1)A
(2.4-8.9)B
(11.0-78.0)A

0.0
3.8
6.9
23.0

(0.0-15.7)AB
(0.4-8.2)B
(3.3-11.0)B
(9.0-41.0)A

Behavioral examination
SU
Fear (%)
Aggressiveness (%)
Negative interaction (%)
Stress signals (N)

0.0
0.0
1.4
21.0

0.001
0.001
0.001
NS
P

RV
(0.0-0.0)A
(0.0-0.0)A
(0.3-3.7)A
(6.0-44.0)A

0.006
0.003
0.001
NS

0.0
0.0
3.2
24.0

UN
(0.0-0.4)A
(0.0-0.0)A
(0.6-5.3)AB
(11.0-78.0)A

6.3
2.4
6.8
31.0

(0.0-16.8)B
(0.0-5.8)B
(4.2-10.7)B
(9.0-66.0)A

NW refers to the group of nonworker pet dogs enrolled in the study.

Different superscript letters within a row indicate groups with signicantly different distributions. Adjusted pairwise comparisons after Kruskal-Wallis test.

0.001
0.001
0.001
NS

108

P. Mongillo et al. / Journal of Veterinary Behavior 10 (2015) 103e110

fear, aggressiveness, and negative interaction during a role-playing

simulation of an AAI session.
Both evaluations identify dogs behavioral expressions that
could jeopardize the participants in AAI (e.g., aggressiveness), thus
permitting the selection of dogs unsuitable for AAI. Notably, reported or observed dangerous dogs were pre-emptively excluded
from the present study; therefore, the predictive validity of the
behavioral examination to exclude aggressive dogs can be considered fairly good. Moreover, specicity and sensitivity of the reason
aggressiveness, when data of both evaluations were considered,
support the validity of the procedure to identify dogs expression of
aggressive behavior during an AAI session. These parameters were
fairly high and slightly higher for our protocol compared to other
behavioral tests assessing dogs aggressiveness toward an unknown
person (specicity 0.73, sensitivity 0.76, Klaus et al., 2014;
specicity 0.81, sensitivity 0.33, van der Borg et al., 2010;
specicity 0.81, sensitivity 0.84, Planta and De Meester, 2007).
According to the fact that absence of aggressiveness is an
essential minimum prerequisite (see also Lucidi et al. 2005), both
evaluations clearly differentiated unsuitable from suitable dogs
(with or without reserve) because of this behavioral trait. Notably,
the handlers of our study did also recognize aggressive behavior
among their dogs but allocated in the category with reserve those
dogs expressing aggressiveness to the same extent as the NW
group. Therefore, handlers seemed to tolerate a certain degree of
aggressiveness in dogs self-judged as RV, disregarding the absence
of aggressiveness as an essential prerequisite. The reason of this
choice may be an underestimation of the risks associated with
aggressive behavior or, more likely, the belief of being able to
adequately manage the expression of such behavior.
Fear expression was easily identied by both evaluations, again
with fairly high specicity and sensitivity. We could not nd any
data to compare specicity and sensitivity of behavioral tests
assessing fear in dogs. Despite accuracy in fear identication, the
behavioral examination and the role-playing evaluation allocated
fearful dogs differently. Absence of the expression of fear was
treated as a prerequisite in the behavioral examination, whereas a
certain degree of fear was considered acceptable in the roleplaying, where dogs expressing fear could be judged RV. The
reason of such difference may be grounded in the modality of fear
assessment. Assessing this emotional state in the context gives
direct access to more detailed information about specic fearevoking stimuli, intensity of expression, temporal evolution, and
time for recovery (Palestrini, 2009). In this sense, quality of fear
signs may have being considered acceptable in RV dogs and unacceptable in dogs judged UN according to the role-playing. The same
consideration could have guided the handlers who accepted a
certain degree of fear in dogs self-judged as RV. Behavioral examination assessed fear by history and present signs during consultation, which would not allow access to the same direct and
contextualized information. This shortage may have led to a more
cautious judgment in this evaluation. However, expression of fear
per se is not a prerequisite of a therapy dog unless it jeopardizes
either AAI patients (e.g., preluding aggression) or dog welfare. In
line with this point of view, one previous study (Lucidi et al., 2005)
evaluated therapy dogs fear expression as an ordinal variable,
excluding only dogs that reached a determined score of fear
expression.
Negative interaction (e.g., avoiding contact or freeze) with the
actors present in the AAI simulation is another undesirable
behavioral characteristic in a therapy dog. This trait was clearly
noticed in both evaluations and by handlers with diverse consequences on the judgment. Notably, this is the only behavioral
parameter that differentiated dogs self-reported as suitable (with
or without reserve) from the NW group of family dogs. Because no

difference emerged between dogs self-reported as RV or not, the

absence of such behavioral expression seems to be the prerequisite to the eyes of the handlers. On the contrary, expression of
negative interaction of RV dogs was at an intermediate level or at a
level similar to UN dogs according to the behavioral examination
and the role-playing evaluation, respectively. Moreover, roleplaying evaluation is the only assessment that could identify fully
SU dogs. Freeze or ight behavioral reactions are nonaggressive
strategies of conict resolution in dogs (Walker et al., 1997; Lindsay,
2005) as in others animals. A passive strategy (e.g., freeze) is
believed to be, to a certain extent, a measure of dogs sociability in
puppies (Riemer et al., 2013), whereas the expression of ight reaction (e.g., active avoidance) during interspecic interaction is less
} ri et al., 2010). Accordingly, negative
consistent (Vas et al., 2008; Gyo
interactions expressed by the dogs in the present study could
represent both a temperament trait of the dogs and/or a dogs
exible response to an interaction perceived as threatening. Even if
both conditions are undesirable in therapy dogs, the latter may be
overcome with appropriate training (Lucidi et al., 2005). Therefore,
different judgments on the condition underlying such behavioral
expression may have accounted for differences in judgment of dogs
expressing negative interaction between the behavioral examination and the role-playing.
Different ability to cope with stressful situations possibly
occurring in AAI was unnoticeable with the present protocol. Both
evaluations failed in differentiating stress expression in dogs with
this reason for allocation, and no differences of levels emerged
between dogs judged differently, even by the handlers. Most likely,
a ceiling effect could explain such result. This is further supported
by the similar variability in stress signals emission between therapy
and NW dogs and the quite high expression of stress signals
(approximately 2.5 stress signals/minute) elicited by our procedure.
Studies on stress-related expressions in therapy dogs give inconsistent results (Haubenhofer and Kirchengast, 2006; Haubenhofer
and Kirchengast, 2007; Marinelli et al., 2008; Marinelli et al.,
2009; King et al., 2011; Glenk et al., 2014) being greatly affected
by the AAI conditions (duration, frequency, site and type of interventions, number and age of participants, parameters used to
assess the dogs stress). It is then likely that in our protocol being in
an unknown place with a stranger behaving in an unusual and
inconsistent manner was rather a challenging condition for those
dogs. However, similar and even more challenging situations occur
in real AAI sessions (see the Supplementary Video to appreciate the
actual condition the dogs had to face), and an easier simulation
would have been improper for the scope of the present research.
Nonetheless, the lack of better coping in therapy dogs is unexpected
as experience had proved to positively affect dogs stress parameters during AAI (King et al., 2011; Glenk et al., 2013). Several reasons
could explain this lack: handlers were unwilling or unable to
advocate for their dogs welfare in this experimental setting or to
perceive their stress signals; the levels of our dogs technical skills
were equal regardless their working experience; the ability to cope
in stressful AAI sessions is not accounted for in therapy dogs selection; in challenging AAI situations, therapy dogs are just as
stressed as NW dogs. The rst and second hypotheses are currently
under investigation by our group, whereas the last two will pose
ethical questions regarding the involvement of dogs in challenging
AAI sessions.
The concordance of judgment between the behavioral examination and the role-playing evaluation was not optimal. This
outcome could likely arise from the different behavioral aspects
these procedures are able to investigate and from the diverse information about the same aspect possibly obtained. As discussed
previously for fear expression, the quality of information is quite
different. Some rationales for allocation of dogs are not shared by

P. Mongillo et al. / Journal of Veterinary Behavior 10 (2015) 103e110

109

and Eleonora Simoni for helping with experiments and to Carlo

Poltronieri and Tommaso Brogin for technical assistance.
This study was funded by the University of Padova and by the
Istituto Zooprolattico Sperimentale delle Venezie (Legnaro,
Padova, Italy).
Lieta Marinelli, Sabrina Bonichini, and Luca Farina conceived the
idea for the study. Lieta Marinelli, Sabrina Bonichini, and Paolo
Mongillo designed the experiments. Serena Adamelli, Elisa Pitteri,
Lieta Marinelli, and Paolo Mongillo performed the experiments.
Lieta Marinelli and Paolo Mongillo analyzed the data. Paolo Mongillo, Sabrina Bonichini, and Lieta Marinelli wrote the article. The
study complied with relevant Italian laws.

the 2 evaluations (e.g., being excitable, intraspecic aggressiveness,

absence of interest in interaction with unknown person). Moreover,
the presence of behavioral problems would never be assessable by
role-playing evaluation. Given these constraints, it would be proper
to consider these evaluations as 2 necessary steps of a selection
protocol. This approach would also guarantee a better specicity
and sensitivity in identifying dogs behavioral prerequisites and a
better safeguard of patients and dogs welfare. As suggested by
therapy animal organizations (e.g., AKC, 2014), dogs could undergo
behavioral examination and role-playing simulation in sequence in
appropriate moments. This would allow excluding dogs with
behavioral problems from further assessment and accounting for
dogs and handlers technical skills in role-playing evaluation.
Moreover, handlers of dogs judged as RV at clinical examination
could get proper recommendations for helping their dogs and reach
a fully SU judgment at role-playing. Finally, a 2-step evaluation
allows involvement of animal behavior experts with different
professional proles. According to the Federation of Veterinarians
of Europe (FVE, 2008), veterinarians are the only professionals
allowed to determine the health and welfare status of an animal.
Even more importantly for the aim of an ofcial protocol, veterinarians are the only ones who can take legal responsibility for this
act, and the behavioral examination performed by a veterinarian
expert in animal behavior and in AAI could provide a certicate to
be included in the selection protocol. Evaluation of the technical
skills in the role-playing could be performed with the support of
professional dog trainers with expertise in AAI. At present, these
professionals are not legally recognized in Italy, and this shortcoming could deserve action by the Italian governmental bodies.

Elisa Pitteri was supported by a PhD grant funded by the University of Padova, Italy. The other authors declare no conicts of
interest.

Conclusions

Supplementary data

The protocol of therapy dogs selection assessed by the present

research is the rst scientic attempt to standardize and validate a
recognized procedure applicable by organizations working in the
AAI eld. Given our results, the protocol could be easily and properly adopted to assess the presence of behavioral prerequisites in
dogs directed toward working in AAIs. However, the possible applications extend beyond the initial selection of dogs; for instance,
the protocol could be used to evaluate dogs already working in AAIs
on a regular basis to verify the maintenance of a desirable behavioral standard over time. Moreover, the adoption of standardized
evaluation criteria could provide a useful reference base for the
training of veterinarians and other professional gures that are
called to judge therapy dog-handler pairs.
In the absence of legal regulations and common guidelines, AAIs
currently comprise very diverse activities, and welfare concerns for
the involved dogs are justied and benecial. To this regard, further
research aiming at standardizing procedures for dogs selection,
training, and monitoring should be highly encouraged. The adoption of common procedures by institutions working in this eld will
also prompt research on the topic, currently dissuaded by the high
variability of AAI practices.

Supplementary data related to this article can be found at http://

dx.doi.org/10.1016/j.jveb.2014.11.005.

Acknowledgments
The authors express their sincere gratitude, rst and foremost,
to the dogs owners and all the associations and volunteers who
participated in this study. Being judged is never easy and all the
handlers enrolled proved to be very dedicated to their profession
and willing to improve their skills for the benet of their dogs and
patients. The authors also thank Soa Farina for her help in revising
the use of the English language in this article. We are very grateful
to the students Matteo Candaten, Lavinia Eddy, Rosanna Blonda,

Ethical considerations
Compliance with the policy of the journal on ethical consent
and standards of animal care was not applicable. All procedures
involving dogs and owners were performed in compliance with
relevant Italian legislations and have been approved by the Ethical
Committee of Istituto Zooprolattico Sperimentale delle Venezie.
The owners and handlers undertook the procedure on a voluntary
basis. They were informed about the aim of the study and gave their
explicit consent for collecting data from the video.
Conict of interest

References
AKC, 2014. American Kennel Club, Inc. Therapy Dog Organizations. Available at:
https://www.akc.org/akctherapydog/organizations.cfm. Accessed August 2014.
Altman, D.G., 1991. Practical Statistics for Medical Research, 1st ed. Chapman and
Hall, Oxford, pp. 1e611.
AVMA, 2014. American Veterinary Medical Association, Service and Therapy Animals: Wellness Guidelines for Animals in Animal-assisted Activity, Animalassisted Therapy and Resident Animal Programs. Available at: https://
www.avma.org/KB/Policies/Pages/Wellness-Guidelines-for-Animals-in-AnimalAssisted-Activity-Animal-Assisted-Therapy-and-Resident-Animal-Programs.aspx.
Accessed August 2014.
Bernabei, V., De Ronchi, D., La Ferla, T., Moretti, F., Tonelli, L., Ferrari, B., Forlani, M.,
Atti, A.R., 2013. Animal-assisted interventions for elderly patients affected by
dementia or psychiatric disorders: a review. J. Psychiatr. Res. 47, 762e773.
CNB, 2005. Comitato Nazionale per la Bioetica. Problemi bioetici relative allimpiego
di animali in attivit correlate alla salute e al benessere umani. Documento
approvato nella seduta Plenaria del 21 ottobre 2005.
CRN, 2012. Possibili sviluppi dei progetti di ricerca in relazione al censimento - CRN
per gli Interventi Assistiti con gli Animali, 27 gennaio 2012. Available at:http://
www.centroreferenzapet-therapy.it/js/cknder/userles/les/Rel.%20Seminario
%202701/2_%20Possibili%20sviluppi%20dei%20progetti%20di%20ricerca%20in%
20relazione%20al%20censimento_Dott_%20Muraro.pdf. Accessed August 2014.
Delta Society, 2002. Minimum Standards for Service Dogs. Delta Society, Renton,
WA.
DGR n. 4130, 2006. Deliberazione della giunta regionale n. 4130 del 19 dicembre
2006, Bollettino Ufciale Regione del Veneto, XXXVIII n. 10 del 30 gennaio 2007.
DM G.U., 2009. Decreto Minsteriale 18 giugno 2009. Ministero del lavoro, della
salute e delle politiche sociali. Istituzione di nuovi Centri di referenza nazionali
nel settore veterinario. (09A11290) Gazzetta Ufciale della Repubblica Italiana,
Serie Generale n. 225 del 28 settembre 2009.
Fredrickson-MacNamara, M., Butler, K., 2006. The art of animal selection for animalassisted activity and therapy programs. In: Fine, A.H. (Ed.), Handbook on
Animal-assisted Therapy: Theoretical Foundations and Guidelines for Practice,
2nd ed. Academic Press, San Diego, CA, pp. 121e147.
FVE, 2008. Federation of Veterinarians of Europe, Denition of Veterinary Act.
Available at: http://www.fve.org/uploads/publications/docs/fve_08_009_vetact.
pdf. Accessed August 2014.

110

P. Mongillo et al. / Journal of Veterinary Behavior 10 (2015) 103e110

Glenk, L.M., Kothgassner, O.D., Stetina, B.U., Palme, R., Kepplinger, B., Baran, H., 2013.
Therapy dogs salivary cortisol levels vary during animal-assisted interventions.
Anim. Welf. 22, 369e378.
Glenk, L.M., Kothgassner, O.D., Stetina, B.U., Palme, R., Kepplinger, B., Baran, H., 2014.
Salivary cortisol and behavior in therapy dogs during animal-assisted interventions: a pilot study. J. Vet. Behav.: Clin. Appl. Res. 9, 98e106.
} ri, B., Gcsi, M., Miklsi, ., 2010. Friend or foe: context dependent sensitivity to
Gyo
human behaviour in dogs. Appl. Anim. Behav. Sci. 128, 69e77.
Haubenhofer, D.K., Kirchengast, S., 2006. Physiological arousal for companion dogs
working with their owners in animal-assisted activities and animal-assisted
therapy. J. Appl. Anim. Welf. Sci. 9, 165e172.
Haubenhofer, D.K., Kirchengast, S., 2007. Dog handlers and dogs emotional and
cortisol secretion responses associated with animal-assisted therapy sessions.
Soc. Anim. 15, 127e150.
IAHAIO, 1998. The IAHAIO Prague Declaration, Approved by the General Assembly at
the International Association of Human-animal Interaction Organizations
Conference, 10e12 September, Prague, Czech Republic.
King, C., Watters, J., Mungre, S., 2011. Effect of a time-out session with working
animal-assisted therapy dogs. J. Vet. Behav.: Clin. Appl. Res. 6, 232e238.
Klaus, B., Kis, A., Persa, E., Miklsi, ., Gcsi, M., 2014. A quick assessment tool for
human-directed aggression in pet dogs. Aggress. Behav. 40, 178e188.
Lindsay, S.R., 2005. Handbook of Applied Dog Behavior and Training. Procedures
and Protocols, Vol. 3. Blackwell Publishing, IA.
Lucidi, P., Bernab, N., Panunzi, M., Dalla Villa, P., Mattioli, M., 2005. Ethotest: a new
model to identify (shelter) dogs skills as service animals or adoptable pets.
Appl. Anim. Behav. Sci. 95, 103e122.
Marcus, D.A., 2013. The science behind animal-assisted therapy. Curr. Pain. Headache Rep. 17, 322.

Marinelli, L., Salvadoretti, S., Normando, S., Moro, L., Ravarotto, L., Bono, G., 2008.
Monitoraggio funzionale ed etologico del benessere di cani impiegati in attivit
assisitite dagli animali. Obiettivi e Documenti Veterinari 6, 36e45.
Marinelli, L., Normando, S., Siliprandi, C., Salvadoretti, M., Mongillo, P., 2009. Dog
assisted interventions in a specialized centre and potential concerns for animal
welfare. Vet. Res. Commun. 33, 93e95.
Miklsi, ., Topl, J., 2013. What does it take to become best friend? Evolutionary
changes in canine social competence. Trends Cogn. Sci. 17, 287e294.
OHaire, M., 2013. Animal-assisted intervention for autism spectrum disorder: a
systematic literature review. J. Autism. Dev. Disord. 43, 1606e1622.
Palestrini, C., 2009. Situational sensitivities. In: Horowitz, D.F., Mills, D.S. (Eds.),
BSAVA Manual of Canine and Feline Behavioural Medicine, 2nd ed. British Small
Animal Veterinary Association, Gloucester, pp. 169e181.
Planta, J.U.D., De Meester, R.H.W.M., 2007. Validity of the Socially Acceptable
Behavior (SAB) test as a measure of aggression in dogs towards non-familiar
humans. Vlaams Diergen Tijds. 76, 359e368.
Riemer, S., Mller, C., Virnyi, Z., Huber, L., Range, F., 2013. Choice of conict resolution
strategy is linked to sociability in dog puppies. Appl. Anim. Behav. Sci. 149, 36e44.
Urbanski, B.L., Lazenby, M., 2012. Distress among hospitalized pediatric cancer patients modied by pet-therapy intervention to improve quality of life. J. Pediatr.
Oncol. Nurs. 29, 272e282.
van der Borg, J.A.M., Beerda, B., Ooms, M., Silveira de Souza, A., van Hagen, M.,
Kemp, B., 2010. Evaluation of behaviour testing for human directed aggression
in dogs. Appl. Anim. Behav. Sci. 128, 78e90.
}ri, B., Miklsi, ., 2008. Consistency of dogs reactions to
Vas, J., Topl, J., Gyo
threatening cues of an unfamiliar person. Appl. Anim. Behav. Sci. 112, 331e344.
Walker, R., Fisher, J., Neville, P., 1997. The treatment of phobias in the dog. Appl.
Anim. Behav. Sci. 52, 275e289.