Journal Pre-proof Fear/Reactivity in working dogs: an analysis of 37 years of behavioural data from the Mira Foundation’s future service dogs ´ Paulus, Noel ¨ Champagne, Nicolas Nicolas Dollion, Amelie ´ St-Pierre, Eric St-Pierre, Marcel Trudel, Pierrich Plusquellec
PII:
S0168-1591(19)30014-0
DOI:
https://doi.org/10.1016/j.applanim.2019.104864
Article Number:
104864
Reference:
APPLAN 104864
To appear in:
APPLAN
Received Date:
28 January 2019
Revised Date:
13 August 2019
Accepted Date:
28 August 2019
´ Trudel Please cite this article as: Dollion N, Paulus A, Champagne N, St-Pierre N, St-Pierre E, M, Plusquellec P, Fear/Reactivity in working dogs: an analysis of 37 years of behavioural data from the Mira Foundation’s future service dogs, Applied Animal Behaviour Science (2019), doi: https://doi.org/10.1016/j.applanim.2019.104864
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2019 Published by Elsevier.
Fear/Reactivity in working dogs: an analysis of 37 years of behavioural data from the Mira Foundation’s future service dogs
Nicolas Dolliona, b, Amélie Paulusa, Noël Champagneb, Nicolas St-Pierrea, Éric St-Pierreb, Marcel Trudelc,
a
of
Pierrich Plusquelleca, d
Laboratoire d’Observation et d’Éthologie Humaine du Québec, Research Centre, Montreal Mental Health
ro
University Institute, CIUSSS Est, Montreal, Canada b
Department of Psychoeducation, University of Sherbrooke, Montreal, Quebec, Canada
re
c
-p
Mira Foundation, Sainte-Madeleine, Quebec, Canada
d
ur na
Corresponding author:
lP
School of Psychoeducation, University of Montreal, Montreal, Quebec, Canada
Pierrich Plusquellec
School of Psychoeducation
Jo
University of Montreal
PO Box 6128 Centre-ville STN Montréal (Québec) H3C 3J7 Canada
1
E-mail:
[email protected]
Highlights
Using dimension reduction analyses we were able to extract Fear/Reactivity dimensions This personality dimension is relatively stable across development and measurement tools Fear/Reactivity is an important contributor and predictor of dogs’ disqualification Fear/Reactivity influences the type of programme dogs will certify into
of
Defining and predicting a dog’s personality is a major concern for groups providing service and/or
ro
working dogs. The Mira Foundation is a non-profit organisation offering service dogs to autistic children and individuals with motor or visual disabilities. Since its establishment in 1981, the Foundation has donated
-p
thousands of dogs and registered behavioural data on its dogs by relying on rigorous and standardised procedures. First, at 6 and 12 months of age, while the dogs are in a foster family, data are collected using
re
questionnaires to assess the presence or absence of specific behaviours (QFFs). Then, at 1 year of age, just prior to receiving professional training, dog trainers conduct a set of short behavioural tests to evaluate dogs
lP
for various traits that are key for training. Exploiting the unique database created by the Foundation, this study aimed to investigate whether, as pointed out in the literature, fearfulness can be a reliable predictor of
ur na
adult dog personality traits and of qualification as a service dog. More specifically, three goals were pursued in the present study: 1) to study the possibility of extracting fearfulness dimensions at 6 months and 1 year of age; 2) to assess the stability of this dimension across development and across the types of evaluations; 3) to assess its predictive value for future qualification as a service dog (i.e. guide dog, assistance dog, dog
Jo
for ASD children, or breeder); and 4) to investigate the impact of sex and breed on the likelihood of achieving qualification as a service dog and on Fear/Reactivity at 1 year of age. The analysis revealed the presence of a Fear/Reactivity personality dimension in dogs across the three evaluations. The results confirmed the consistency of this dimension between 6 and 12 months of age (QFFs), and in the behavioural evaluation at 1 year of age. Furthermore, the Fear/Reactivity dimension significantly predicted dog disqualification from training programmes and also the subtype of the programme in which dogs will be
2
certified. Finally, the results indicated that a dog’s Fear/Reactivity and likeliness to qualify in a specific programme varied according to its sex and breed.
Keywords: guide dogs, service dogs, Fear/Reactivity, prediction
of
1. Introduction
ro
Service dogs are a class of working dogs specifically dedicated to aiding and assisting people suffering from various forms of physical and mental disabilities. In order to do this, the dogs must receive
-p
specific training and be selected based on their health, personality, and working abilities. At least half of the candidate dogs will not be selected or qualify to be service dogs (Duffy and Serpell, 2012; Sinn et al., 2010).
re
Behavioural issues are the primary reasons for this rejection. The main rationale for this high attrition rate is to avoid qualifying dogs with behaviours that could be problematic or unsuitable for their future work
lP
(Carere and Locurto, 2011; Gazzano et al., 2008). It is thus critical for organisations to assess a dog’s personality as early as possible, and to be able to predict its ability to undergo training.
ur na
Personality can be defined as the behavioural tendencies that vary between individuals, and that are stable in the same individual across time and situations (De Palma et al., 2005; Gosling, 2001; Miklósi et al., 2014; Riemer et al., 2016; Stamps and Groothuis, 2010). Three assessment methods are commonly used to assess dog personality: a battery of behavioural tests, ad hoc observations, and/or questionnaires given to
Jo
the caretaker (Mirkó et al., 2012). Numerous studies have been conducted on this topic using these assessment methods (Fratkin et al., 2013; Miklósi et al., 2014; Svartberg, 2007) and several models derived from factor-analytic approaches have been proposed (e.g. De Palma et al., 2005; Diedrich and Giffroy, 2006; Gosling and John, 1999; Hsu and Serpell, 2003; Ley et al., 2008; Svartberg and Forkman, 2002). Among these studies, many find similar personality dimensions across dog breeds: activity/excitability, aggression, confidence/fear, extraversion/sociability, and obedience/trainability (for review see Fratkin, 2017). One of
3
the most commonly accepted models of dog personality was proposed by Jones and Gosling (2005). After performing a meta-analysis of 51 studies and an expert-sorting of all the listed descriptors of dog personality, the authors extracted seven broad personality dimensions: reactivity, fearfulness, activity, sociability, responsiveness to training, submission, and aggression. These dimensions were subsequently reduced to six dimensions, with a merging of reactivity and fearfulness into a single Fear/Reactivity dimension (Fratkin et al., 2013).
of
Numerous studies have attempted to determine whether some dog personality traits might be
ro
significantly correlated with exclusion from different types of working dog programmes. In guide dogs (i.e. dogs for blind or visually impaired people), it has been demonstrated that dog disqualification can be
-p
predicted based on personality traits, notably: distractibility, activity, aggression, fear, sensitivity, and docility (Arata et al., 2010; Batt et al., 2008; Duffy and Serpell, 2012; Goddard and Beilharz, 1982, 1983;
re
Serpell and Hsu, 2001; Tomkins et al., 2011). In military and police dogs, other personality traits, such as
Sinn et al., 2010; Svartberg, 2002).
lP
vivacity, object focus, boldness, and trainability, have been identified as such predictors (Foyer et al., 2014;
One of the holy grails of dog research is the prediction of adult personality based on behavioural
ur na
dimensions as a puppy (Asher et al., 2013), since this might provide ways for organisations to optimise their selection of dogs early in the process. In guide dogs, some personality traits have been found as potential predictors in puppies as young as 6–8 weeks of age; these traits are: activity, fearfulness, aggression, sensitivity to humans, and distractibility (Asher et al., 2013; Batt et al., 2008; Duffy and Serpell, 2012;
Jo
Kobayashi et al., 2013). Among the different personality traits, fearfulness has been demonstrated to be one of the most consistent across development and a reliable predictor of disqualification from guide dog training programmes (Fratkin, 2017; Goddard and Beilharz, 1982, 1984, 1986). However, some authors question the consistency of personality traits between assessments conducted at an early stage of development and assessments conducted in adulthood (Bremhorst et al., 2018).
4
The Mira Foundation is a Quebec non-profit organisation, founded in 1981, which is dedicated to the donation of service dogs to people with visual and physical disabilities, and to families of autistic children. Mira has always relied on a rigorous methodology in order to select dogs with the most suitable personality. First, while in foster families, data on the dogs’ behaviour are collected through questionnaires addressed to the caretakers when the dogs are 6 and 12 months old. Second, the dogs receive a full behavioural evaluation at 1 year of age, just before assignment to training. Despite the long history of the
of
Mira Foundation and the successful selection of its breeds, the rate of exclusion remains significant at approximately 38 %. Using the considerable database on dogs’ behaviour registered by the Foundation over
ro
the years, the aim of this study was to test whether, as suggested in the literature, a dog’s fearfulness would emerge as a behavioural dimensions at 6 and 12 months of age, and whether this specific personality
-p
dimension would be consistent across development and represent a reliable predictor of adult dog
re
personality and of qualification/disqualification as a service dog.
Numerous studies have demonstrated differences in dogs’ Fear/Reactivity according to sex and
lP
breed (e.g. Asp et al., 2015; Fratkin, 2017; Willis, 1995; Wilsson and Sundgren, 1997), and it has been suggested that some breeds have a higher potential than others to present a personality profile that is suited
ur na
to becoming a working or service dog (Batt et al., 2008). We thus also investigated whether variation in qualification/disqualification as a service dog according to sex and breed could be observed, and whether those differences coincided with variations in Fear/Reactivity at 1 year of age.
Jo
2. Materials and methods 2.1. Subjects
The total sample in this study included 5340 dogs (2777 males and 2563 females). All subjects were owned and/or donated by the Mira Foundation, except for 129 breeding dogs who were owned by other breeding facilities affiliated with the Foundation. All the dogs were born between 1987 and 2016. Among
5
the 5340 dogs, 1555 were Labradors, 417 were Bernese Mountains, 2563 were Labrenese (i.e. a crossbreed between Bernese Mountain and Labrador), 799 were Saint-Pierres (i.e. at least third generation of Labernese), 404 were from four other breeds used by the Mira Foundation in its early years (176 Royal Poodles, 127 Golden Retrievers, 97 Golden Labradors, and 4 Bouvier des Flandres), and 6 dogs were from unknown breeds (i.e. their breed was not registered in Mira’s database). The four other breeds used early on by the Foundation were suited to the weather in Québec and known to be trainable. In its early stages, the
of
Foundation tested their suitability as service dogs.
ro
Among this sample, 2020 dogs were disqualified (38 %), 3210 were qualified (60 %), and the qualification/disqualification status of the remaining 110 dogs was not specified in the database. Among the
-p
disqualified dogs, 1261 dogs were disqualified for behavioural and personality issues (62 %) and 759 for health issues (38 %). The issues leading to disqualification for the remaining 92 dogs were not specified in
re
the database. It is important to note that the dogs disqualified for behavioural issues corresponded either to dogs disqualified based on the behavioural evaluation (BE) at 1 year of age, dogs disqualified during training
lP
for behavioural issues, or dogs disqualified due to behavioural issues after being assigned to a beneficiary (representing approximately 6–7 % of the assigned dogs). When qualified, the dogs were either selected as
ur na
breeder or certified as service dogs in one of Mira’s four main service dog programmes: guide dog (i.e. dogs trained to guide a blind or visually impaired individuals during daily travels), assistance dog (i.e. dogs trained to assist physically impaired people in their daily activities, such as opening doors, picking up objects, pulling the wheelchair etc.), or dogs for children with autism spectrum disorders (ASD) (i.e. dogs
Jo
trained to escort children with ASD in their daily activities). Among the qualified dogs, 865 dogs became guide dogs, 955 became assistance dogs, 619 became dogs for ASD children, 386 became breeders, and 162 were destined to other types of experimental programmes and work (e.g. court dogs, ergotherapy centre dogs, promotion dogs, and dogs for beneficiaries with multiple deficits). The exact programme to which the remaining 223 dogs were assigned was not specified in the database.
6
Out of the 5340 dogs considered in the present study, 1809 and 3216 dogs had registered data extracted from the questionnaire given to the foster family (QFF) and completed when the dogs were 6 and 12 months old, respectively, and 3499 dogs had registered data collected through the behavioural evaluation (BE) at 1 year of age. Furthermore, among these dogs, 1585 dogs had results from both the QFF at 6 and 12 months of age, 1594 had data from both the QFF at 6 months and the BE at 1 year of age, and 3044 had data from both the QFF at 12 months and the BE at 1 year of age.
of
2.2. Procedure
ro
All dogs bred at the Mira Foundation underwent the same procedures. From birth until they reached 8 weeks of age, puppies were raised with their dam and littermates at the Foundation nursery. The Mira
-p
Foundation possesses its own veterinarians who provide all standard veterinary care and monitoring for all puppies during this period. From 8 weeks of age (8.19 ± 1.27 weeks of age), the dogs were placed and raised
re
in foster family homes, who were volunteers as part of a puppy-walking programme, with the purpose of socialising the dogs, exposing them to diverse types of situations and environments, and providing their first
lP
training skills (e.g. to respond to their name, walk on a leash, and house training). While with the foster family, all veterinary care (i.e. check-ups, vaccines, and neutering) was provided by Mira’s veterinarians.
ur na
Except for future breeders, all male dogs were castrated at approximately 7 months of age (mean age 7.28 ± 2.96 months) and female dogs were spayed at approximately 1 year of age (11.22 ± 4.71 months). Breeders were neutered at retirement, which occurred when they reached approximately 6 years of age (mean age 74.83 ± 22.36 months). During their time with the foster families, two series of data collection were
Jo
performed on dogs through questionnaire surveys addressed to the foster families when the dog reached 6 and 12 months of age. The questionnaires were completed during the week preceding the visit of the foster family to the foundation (i.e. at 6 months old, when visiting for the veterinary check-up, and at 12 months old, when returning the dog). At 1 year of age, the dogs were returned to the Mira Foundation, where they received a full veterinary check-up (notably checking for dysplasia issues) and went through a BE carried out by trainers (mean age at BE 12.59 ± 1.66 months). The decision of whether to induct a dog into the
7
training programme was based on the results of both the BE and the veterinary check-up. If disqualified, the dog was removed from Mira’s working dog programmes and placed into a regular family home. If qualified, the dog was either kept as a breeder or put into training to become a guide dog (7 months of training), an assistance dog (5 months of training), or a dog for autistic children (3 months of training) (see https://www.mira.ca/en/ for more details on training). These training programmes are mutually exclusive
of
(i.e. if trained in one programme, a dog cannot work as a service dog in another programme).
ro
2.3. Behavioural measurements 2.3.1 Questionnaire survey
-p
The questionnaire survey was addressed to foster families when the dog reached 6 and 12 months
re
of age. The questionnaires assessed the presence or absence of specific behaviours through closed-ended questions. Responses were provided by the foster family by ticking the appropriate answer. The respondents
lP
were free to add any comments for each of their answers. The survey included questions covering various topics: the dog’s house training, its general activity and behaviours, and its attractiveness and fear toward various types of stimuli. The questionnaires also included more general questions concerning the living
ur na
habits of the dog, in addition to more specific questions (e.g. the dog’s ease in cars, attraction to various people and animals, and response to its name). However, these last questions were not recorded in the database constituted by the Mira Foundation since the answers were too broad and varied, or due to the answers overlapping with others. The respondents were asked to answer according to their whole experience
Jo
with the foster dog since its arrival at their home. The QFFs at 6 and 12 months of age contained 25 and 37 items, respectively. Twenty items were common to both questionnaires, with some topics covered by one question in one questionnaire, and by multiple questions in the other (in those cases the answers were fused into a single answer as specified in the caption for Table i). The questions are provided as supplementary material (Table i). 2.3.2 Behavioural evaluation
8
Over its 37 years of existence, the Mira Foundation has tried and applied various methodologies in order to select dogs with the most suitable personality traits for training and becoming service dogs. Twelve behavioural tests, carried out by trainers, constitute the core of the tests applied and recorded by the Foundation since its creation. It is important to note that these tests have always been performed by relying on a standardised procedure. The set of 12 behavioural tests is as follows: attraction to cats, walking on staircases, noise
of
sensitivity (walking close to a running tractor), reaction to odd stimuli (a lion statue), walking on a bridge with traffic, simple walking, stranger encounter, activity when left alone, gluttony, competition for resource,
ro
walking in a pack, and the mannequins’ corridor test (Table ii in the supplementary materials provides a full description of each test). These tests were designed to assess different aspects of the dogs’ personality that
-p
are relevant for the selection of future service dogs: social attraction and aggression (toward other dogs,
re
animals, and unfamiliar humans), noise sensitivity, vertigo, reactivity/sensitivity to odd stimuli and situations, submissiveness, reactivity to isolation, exploratory behaviours, global activity level, and docility.
lP
All behavioural tests for the evaluation of the dogs’ personalities were conducted at the Mira Foundation. They were performed and video recorded by a dog trainer, and all video footage was then observed and
ur na
assessed by a second dog trainer during the week following the behavioural evaluation. The measurement methodology corresponds to a semi-quantitative assessment of behaviour. For each test, the dog received a score between 0 and 10 according to its performance during the test, based on specific criteria (higher scores reflected more desirable behaviours/responses in the test; the scoring criteria are specified in Table ii). If the test required that the dog trainer hold the dog on a leash, the dog trainer had to remain passive and not
Jo
intervene during the entire test. If a dog obtained a score less than or equal to five in one of the twelve tests, it was not selected to go into training. Similarly, if a dog obtained scores below 6 or 7 in multiple tests, disqualification was considered and the dog trainer evaluating the videos would have performed a closer investigation of the footage of the 12 tests. Although these tests and their specific procedures were specifically developed by the Mira Foundation in order to address their specific needs concerning dog behavioural traits, they share numerous
9
similarities with some dog personality tests described in the dog personality literature (for a review, see Diederich and Giffroy, 2006). Indeed, moving dolls or dummies, strange unfamiliar humans, dogs, and cats have frequently been used in various studies and dog selection protocols to assess social attraction and aggressiveness toward humans, conspecifics, and other animals (Bollen and Horowitz, 2008; Schalke et al., 2008; van der Borg et al., 1991; Weiss, 2002; Wilsson and Sudgren, 1997). A simple walk on a leash outside has also been added to some protocols in order to assess dog obedience (e.g. van der Borg et al., 1991;
of
Wilsson and Sudgren, 1997). Furthermore, dog noise sensitivity has often been tested in studies of dog personality, and even more so in studies of service dogs, which have generally used loud metallic noises or
ro
gunshots (Diedrich and Giffroy, 2006; Goddard and Beilharz, 1984; Svartberg, 2002; Taylor and Mills, 2006; Wilsson and Sundgren, 1997). Finally, a dog’s propensity to compete for a resource has often been
-p
tested through tests assessing food-guarding behaviours (defending access to a bowl of food), whereas
re
exploratory behaviours have been investigated through the placement of the dog in an empty and unfamiliar
2.4. Statistical analyses
lP
room (Taylor and Mills, 2006; van der Borg et al., 1991).
First, all the answers to both QFFs were coded into categorical scores to enable the statistical
ur na
analyses to be performed (“No” answers were coded as 1, “Yes”, ‘Sometimes” and “Excessively” were coded as 2; “< 5 min,” “5–10 min,” and “> 10 min” were coded as 1, 2, and 3, respectively). Only common items between both QFFs were considered (i.e. the 20 common question items shown in bold and italics in Table i). All data analyses were conducted using SPSS Statistics 25 (IBM Corp, Chicago, IL, USA). The
Jo
dimension reduction technique was applied since it allows for the characterisation of the common and opposite patterns of results present in a set of data and for the identification of combinations of variables (i.e. factors) that account for most of the variability. Since our data set concerns behaviours and tests reflecting the dogs’ personalities, this type of statistic allowed for the extraction of factors/dimensions characterising Mira dogs’ personalities, and thus the identification of a Fear/Reactivity dimension, if present. Multiple correspondence analyses (MCA) were performed on the data from the 20 common items
10
from the QFF at 6 and 12 months of age. The factors were selected according to their inertia (i.e. the inertia of the last extracted dimension had to be higher than the average inertia). MCAs were chosen since all the variables extracted through the questionnaires were categorical. However, principal component analyses (PCA) were performed on the scores of the 12 tests from the BE at 1 year of age since all the data collected through the BE were continuous. A varimax rotation was applied to the PCA. The extracted factors were determined based on their eigenvalue (i.e. the eigenvalue for the last extracted factor had to be more than
of
1.0). To assess the internal consistency of factors, we also calculated the Cronbach’s alpha reliability for items belonging to each factor. Factor scores were then calculated for each factor extracted through the
ro
MCAs and the PCA for all dogs. The regression method was applied for the extraction of scores of factors obtained through the PCA. Sometimes the foster families did not answer every question, or the trainers did
-p
not perform every behavioural tests in dogs due to weather conditions or technical issues, thus creating
re
missing data. In these cases, missing values were replaced by the mean of the observed values (i.e. mean imputation) during factor extraction with the MCAs and the PCA (2.9 % and 14.5 % of the dogs had missing
lP
values on the QFF at 6 and 12 months of age, respectively, and missing values represented 4.3 % of the total data collected with the BE at 1 year of age).
ur na
Pearson correlations were used to test the associations between the scores of each dog for the Fear/Reactivity factor extracted from the MCA on the results from the QFF at 6 and 12 months, and the scores extracted from the PCA on the results from the BE at 1 year of age. These correlation analyses were performed to assess the consistency of the Fearful/Reactivity factors, and thus, the consistency of a dog’s fearfulness over time. In cases where multiple Fear/Reactivity factors were extracted from a QFF (i.e. two
Jo
Fear/Reactivity factors were extracted from the QQF), linear regressions were applied in order to explore whether those Fear/Reactivity factors were predictive of the Fear/Reactivity factor extracted from the next measurement (i.e. the Fear/Reactivity factor from the BE at 1 year of age). Since fearfulness has been shown to be associated with working dog disqualification from different types of programmes, we investigated whether the Fear/Reactivity dimension assessed through the questionnaires and the BE could predict a dog’s disqualification from a training programme or its success
11
to be certified by a specific type of training programme (such as a guide dog, a dog for ASD children, an assistance dog, or a breeder). Binary logistic regression tests were applied to the Fear/Reactivity dimension extracted from both QFFs and the BE at 1 year of age. For each measurement, distinct binary logistic regression tests were applied to explore whether Fear/Reactivity was predictive of: disqualification for behavioural issues, qualification as a guide dog, qualification as an assistance dog, qualification as a dog for ASD children, or qualification as a breeder. Chi-square results from the omnibus test of model coefficients
of
were used to assess the predictive value of the Fear/Reactivity dimension. A close inspection of the results on Wald test was also performed, since it allows to inspect if the predictive variable contributes significantly
ro
to the predictive validity of the model.
Finally, the impact of sex and breed on disqualification due to behavioural issues and on
-p
qualification in one of Mira’s three main service dog programmes (i.e. guide dog, assistance dog, and dog
re
for ASD children) was tested using chi-square tests. Differences on the Fear/Reactivity dimension from the BE at 1 year of age according to those variables were assessed using a T test and an ANOVA followed by
lP
post-hoc Tukey’s tests. In these analyses, the breeds used by the Mira foundation in its early years (i.e. Royal Poodles, Golden Retrievers, Golden Labrador, and Bouvier des Flandres) were considered as a single
ur na
group while the four breeds currently used by the foundation (i.e. Bernese Mountain, Labrador, Labernese, and Saint-Pierre) were considered individually.
3. Results
3.1. Extraction of the Fear/Reactivity dimension in Mira dogs
Jo
Multiple correspondence analyses of the question items from the QFF at 6 months of age (N = 1757)
resulted in the extraction of seven factors, accounting for 48.9 % of the variance (the mean inertia was equal to 0.053). However, only the first four factors were considered. Indeed, factors 5–7 were not considered for further analyses since they overlapped with the first factors (i.e. there was a redundancy of items in factors 5–7 with items in factors 1–4) and their item loadings were weak. The remaining factors (i.e. Factors 1–4) accounted for 33.6 % of the variance. Factor 1 was loaded by question items reflecting a fear of children,
12
women, men, other animals, noises, and traffic; it was thus designated as the “Fear/Reactivity” factor. This factor accounted for 13.2 % of the variance. Dogs with lower scores for this factor had more Fear/Reactivity issues. Factor 2 was loaded by the tendency to steal/beg for food, to climb on furniture, to jump on people, destroy objects, and chew, and the time to calm down with visitors and outside. The third factor was loaded by the following items: does its business on leash, does its business on hard surfaces, and at ease on
of
staircases. Finally, the fourth factor was loaded by the time to calm down with visitors and outside.
ro
The multiple correspondence analysis based on the data from the QFF at 12 months (N = 3142) extracted six factors, accounting for 41.9 % of the total variance (the mean inertia was equal to 0.056).
-p
However, factors 4–6 were excluded from the analysis since they shared common items with factors 1, 2,
re
and 3. Factors 1–3 accounted for 31.1 % of the total variance. The first factor was loaded by the following question items: time to calm down with visitors; time to calm down outside; destroys objects; chews; afraid
lP
of children, woman, men, other animals, new places, noises, and traffic. The second factor was loaded by the fear of children, women, and men; the tendency to jump on people; and the time to calm with visitors
ur na
and outside. The third factor was loaded with items assessing the dog’s ability to do its business on a leash and on hard surfaces, and the time the dog needs to calm down with visitors and outside (Table 3). Factors 1 and 2 were both designated as “Fear/Reactivity” factors, with factor 1 (Reactivity/Activity) appearing to reflect general fear, and factor 2 (Fear of human/Reactivity) appearing to be more specific to the fear of human beings. Those two factors accounted for 10 % and 8.8 % of the variance, respectively. Dogs with
Jo
lower scores on these factors had more Fear/Reactivity issues.
The PCA of the results from the 12 tests of the BE performed on 1-year-old dogs (N = 3499) resulted
in the extraction of four factors, accounting for 48.2 % of the total variance. The factors were labelled according to the patterns of the loadings of the tests. The Kaiser-Meyer-Oklin value was 0.682 and Barlett’s test of sphericity was significant (p < 0.0001). The first factor was loaded by the following tests: activity
13
alone, cat attraction, gluttony, walking in a pack, simple walking, and noise sensitivity. The second factor was loaded by the bridge with traffic test, the mannequin’s corridor test, the noise sensitivity test, and the stranger encounter test. The third factor had negative loadings from the competition for resource test, the bridge with traffic test, and the walking on staircases test. The fourth and final factor was negatively loaded by the walking in a pack test and positively loaded by the odd stimuli test (Table 3). Based on its loadings, the second factor was thus labelled as the Fear/Reactivity factor. Dogs with higher scores for this factor had
ro
3.2. Consistency and predictability of the Fear/Reactivity dimension
of
fewer Fear/Reactivity issues. It accounted for 14.1 % of the variance.
Pearson correlations and linear regressions were used on the Fear/Reactivity factors extracted from
-p
the questionnaire at 6 and 12 months of age, and from the BE at 1 year of age, in order to test the association
re
of the factors and the predictability of Fear/Reactivity across the measurements. The results from the correlation analyses indicated that the Fear/Reactivity factor at 6 months of age
lP
was significantly correlated with both the Fear/Reactivity factors from the QFF at 12 months of age (F1: r1585 = 0.312, p ≤ 0.001; F2: r1585 = 0.339, p ≤ 0.001) and with the Fear/Reactivity factor from the BE at 1
ur na
year of age (r1594 = 0.231, p ≤ 0.001). Furthermore, the first and second Fear/Reactivity factors extracted from the QFF at 12 months of age were both significantly correlated with the Fear/Reactivity factor extracted from the BE at 1 year of age (r3044 = 0.223, p ≤ 0.001; r3044 = 0.243, p ≤ 0.001, respectively). The model of linear regression performed on both Fear/Reactivity factors from the QFF at 12 months of age and the BE at 1 year of age was also statistically significant (F(2, N = 3044) = 186.21, p < 0.0001 R2= 0.109).
Jo
It appears that the Fear/Reactivity dimension is consistent across development (i.e. from 6 months old to 1 year old), and across measurement tools, at least moderately (i.e. QFF and BE).
3.3. Prediction of behavioural disqualifications and qualification status from Fear/Reactivity dimensions
14
In order to assess whether the Fear/Reactivity dimensions from the QFF at 6 and 12 months, and from the BE at 1 year of age might predict the dogs’ qualification/disqualification, binary logistic regressions were performed. It is important to underline that in these analyses, ‘disqualified dogs’ corresponds only to dogs that were disqualified due to behavioural issues and/or unfit personalities, and did not include dogs that were disqualified due to health problems. First, binary logistic regressions were performed to assess whether a dog’s Fear/Reactivity factor at
of
6 months of age predicted behavioural disqualification and the dog’s assignment into the four main programmes. A higher score on this factor was indicative of fewer Fear/Reactivity issues. Concerning 1778=
25.86, p < 0.0001), indicating that
ro
disqualification, the model was statistically significant (X²1,
Fear/Reactivity at 6 months was significantly predictive of behavioural disqualification. For every one-point
-p
increase on the Fear/Reactivity dimension at 6 months of age, dogs were 0.78 times less at risk of being
re
disqualified. Concerning qualification in the four main programmes, the models for qualification as a guide dog (X²1, 1778= 7.52, p = 0.006), as a dog for ASD children (X1, 1778= 11.78, p = 0.001), and as a breeder (X²1, = 5.33, p = 0.021) were all significant, whereas the model for qualification as an assistance dog failed
lP
1778
to reach significance (X²1, 1778 = 0.93, p > 0.05). However, the results of the Wald test on the qualification
ur na
as a breeder revealed a close to significant result (p= 0.059). For every one-point increase on the Fear/Reactivity dimension at 6 months of age, dogs were 1.27 and 1.29 times more likely to be qualified as a guide dog or a dog for ASD children, respectively (see Table 4).
Second, binary logistic regressions were performed on both factors of Fear/Reactivity at 12 months
Jo
of age in order to test whether they could predict behavioural disqualification and assignment to the four main programmes. As previously, a higher score on this factor was indicative of fewer Fear/Reactivity issues. Here, the model was also statistically significant (X²2, 3169 = 149.62, p < 0.0001) (see Table 5). The results indicated that for every one-point increase on the first and second Fear/Reactivity dimensions extracted from the QFF at 12 months, dogs were 0.65 and 0.85 times less at risk of being disqualified for behavioural issues, respectively. Concerning binary regressions on qualification in the four main
15
programmes, all models were statistically significant (all X² ranged from 8.108 to 42.833 and all p values ranged from 0.017 to < 0.001). For every one-point increase on the first Fear/Reactivity factor, dogs had 1.40, 1.29, and 1.50 times more chance of being qualified as guide dog, dog for ASD children, and breeder, respectively, whereas they had 1.23 and 1.13 times more chance of being qualified as a guide and assistance dog, respectively, for every one-point increase in the second Fear/Reactivity factor (see Table 5).
of
Finally, a binary logistic regression performed on the Fear/Reactivity factor extracted from the BE indicated that this factor was also a significant predictor of behavioural disqualification (X² 1, 3463= 245.21, p
ro
< 0.0001). For every one-point increase in the Fear/Reactivity factor, dogs were 0.56 times less at risk of being disqualified (a higher score on this factor was indicative of fewer Fear/Reactivity behaviours observed
-p
during the tests). Concerning the binary regressions on qualification in the four main programmes, all the
re
models failed to reach significance (all p > 0.05), except for the regression on qualification as guide dogs (X²1, 3463= 267.05, p < 0.0001) (see Table 6). For every one-point increase in the Fear/Reactivity factor, dogs
lP
were 2.59 times more likely to be qualified as guide dogs.
The results from these regression analyses highlight that, as suggested in the previous literature, the
ur na
Fear/Reactivity dimension is predictive of dogs’ disqualification, across three measurements. They also reveal that this dimension is predictive of the type of programme/work dogs will qualify into. Notably, we observed that Fear/Reactivity is significantly predictive of qualification as a guide dog across the three
Jo
measurements.
3.3. Differences on Fear/Reactivity dimension and in behavioural disqualifications/qualification
status according to sex and breed An ANOVA followed by posthoc Tukey’s tests were used to explore whether scores on the Fear/Reactivity dimension from the BE varied according to breed, and a T test was used to explore whether differences according to sex were observed for this variable. Chi-square tests were used in order to determine
16
whether the proportion of dogs disqualified for behavioural issues or of dogs qualified in each specific programme (i.e. guide dog, assistance dog, or dog for ASD children) varied according to dogs’ sex and breed. Concerning breed, results from the ANOVA showed a significant difference on Fear/Reactivity dimension according to breed (F4,3490= 66.74, p < 0.001). Posthoc Tukey’s tests indicated that the breeds used by the Mira Foundation in its early years (i.e. Royal Poodles, Golden Retrievers, Golden Labradors,
of
and Bouvier des Flandres) had significantly more Fear/Reactivity issues (M= -0.90) compared to all the other breeds (M from -0.01 to 0.20, all p< 0.001), whereas Labradors had significantly fewer Fear/Reactivity
ro
issues (M= -0.20) than Labernese (M = -0.01, p <0 .001) and Saint-Pierre breeds (M = 0.04, all p = 0.012). The chi-square test performed on the impact of breed on the likelihood of behavioural disqualification
-p
indicated a significant association between breed and behavioural disqualification (X²4, 5003 = 9.89, p =
re
0.042). However, the results revealed that, compared to theoretical distribution, no breed had a stronger propensity to be disqualified (all z ≥ 2.82; all p > 0.05). The chi-square test performed on the breed and the
lP
programme the dogs qualified into indicated a significant association between the two variables (X²8, 2434 = 133.80, p < 0.001). More Labernese became guide dogs compared to theoretical distribution (z= 5.29, p <
ur na
0.001), whereas less Labradors became guide dogs (z = -4.67, p = 0.005). Concerning assistance dogs, less Saint-Pierres were certified into this programme (z = -5.05, p = 0.0010). Finally, more Saint-Pierre and Labrador became dogs for ASD children compared to theoretical distribution (z = 7.62, p < 0.001; z = 4.73, p = 0.004, respectively), whereas less Labernese and Bernese Mountain dogs certified in this programme (z = -7.39, p < 0.001; z = -4.18, p < 0.001).
Jo
Concerning sex, no significant association with disqualification for behavioural issues was observed
(X²1, 5007 = 2.879, p > 0.05), while the chi-square test revealed a significant association with the subtype of programme in which dogs were certified (X²2, 2439 = 179.92, p < 0.001). Significantly more males certified as guide and assistance dogs than in the theoretical distribution (z = 7.96, p < 0.001; z = 3.95, p < 0.001, respectively), whereas more females certified as dog for ASD children (z = 13.18, p < 0.0010). The T test
17
revealed a significant difference according to sex (t3494 = -4.37, p = 0.001), with males having fewer Fear/Reactivity issues (M = 0.07) than females (M = -0.08).
4. Discussion Based on the analysis of behavioural data collected by the Mira Foundation on potential future service dogs, the present study aimed to investigate whether, as suggested in the previous literature,
of
fearfulness/emotional reactivity would emerge as a significant behavioural dimension, and whether it influences the chance of qualification as a service dog. Using dimension reduction analyses, we were able
ro
to extract Fear/Reactivity dimensions from both QFFs at 6 and 12 months of age, and from the BE at 1 year of age. Correlation analyses confirmed the relative consistency of this personality dimension across
-p
development and measurement tools. Furthermore, the results confirmed that not only is the Fear/Reactivity
re
dimension an important contributor and predictor of a dog’s disqualification, but that it also influences the programme in which the dogs will be certified. Finally, the results highlight differences in the specific
lP
programme dogs qualify into and in the dogs’ Fear/Reactivity according to their sex and breed.
ur na
4.1. Fear/Reactivity dimension in dogs
By relying on questionnaires and behavioural observations, multiple studies on dogs’ personalities have identified a Fear/Reactivity dimension among the different extracted personality dimensions (Dowling-Guyer et al., 2011; Ley et al., 2008; McGarrity et al., 2015; Serpell and Hsu, 2001; Svartberg and
Jo
Forkman, 2002; see also Fratkin, 2017; Jones and Gosling, 2005; Svartberg, 2007), and have revealed that this dimension could be reliably measured early in dogs’ development. For example, Goddard and Beilharz (1984, 1986) found that consistent individual differences in Fear/Reactivity could be assessed as early as 8 weeks of age. More recently, Riemer et al. (2016) reported extracting this trait in Border collies as early as 6 months of age. In the present study, the results of the MCAs on data from the QFF at 6 and 12 months of age, and of the PCA on data from the BE at 1 year of age confirmed the presence of a Fear/Reactivity
18
personality dimension in the population of Mira foundation dogs. Furthermore, it also confirmed our hypothesis that Fear/Reactivity is an important personality dimension to account for behaviour differences between potential future service dogs. It is worth noting the distinction of two Fear/Reactivity factors extracted from the QQF at 12 months of age, with one appearing to be a more general fear dimension (i.e. the first factor) and the second appearing to be a more socially-oriented fear dimension (i.e. the second factor). The distinction of these two personality
of
dimensions is no surprise, since multiple studies have revealed a distinction between social and non-social
ro
fear in dogs (e.g. Riemer et al., 2016; Serpell and Hsu, 2001). As an example, using PCAs on data collected through the C-BARQ (Hsu and Serpell, 2003), Serpell and Duffy (2014) reported the presence of a
-p
distinction between two fear factors: a social fear factor, defined by fear of unfamiliar humans and dogs;
lP
4.2. Consistency of Fear/Reactivity
re
and a non-social fear factor, defined by fear of unanimated items (e.g. wind, thunder, traffic, loud noises).
Fear/Reactivity seemed to be relatively consistent between the QFF at 6 and 12 months and the BE at 1 year of age. Indeed, the Fear/Reactivity dimension significantly correlated between the three
ur na
measurements, and the Fear/Reactivity factors from the QFF at 12 months were predictive of the Fear/Reactivity factor from the BE at 1 year of age. This result is consistent with the results reported by Goddard and Beilharz (1984, 1986), who identified fearfulness as a stable personality trait between 6 and 12 months of age. Furthermore, the magnitude of the observed correlations is relevant to the published
Jo
literature assessing the correlation of dogs’ personality traits assessed at different ages, which all together show a mean correlation of 0.34 (see Fratkin et al., 2013; Harvey et al., 2016). Some behaviours and personality traits have been demonstrated to be more stable than others across development, such as aggression and submissiveness (Fratkin, 2017; Harvey et al., 2016; Riemer et al., 2014a, 2016; Slabbert and Odendaal, 1999). Experiences throughout ontogeny and the maturation processes can shape and influence
19
future behavioural tendencies and personality profiles, and can thus account for the low correspondence of personality traits between puppyhood and adulthood (Riemer et al., 2014b, 2016). Since the QFF at 12 months of age and the BE were carried out within the same timeframe (i.e. within a one to two weeks interval), it would have been expected that the correlation between the Fear/Reactivity dimensions from these two measurement tools would have been stronger. The presence of a rather moderate correlation might indicate that the two measurements capture different aspects of
of
Fear/Reactivity. Evaluators (i.e. dog trainers) might give more importance to components that are relevant
ro
to a guide and/or serving dog than foster families do. Furthermore, during the BE, dogs might be exposed to and evaluated in situations that are not assessed through the QFFs and vice versa.
-p
Even though significant and comparable to previous literature, the correlation between the Fear/Reactivity factors according to the three measurements remains low (r ranging from 0.231 to 0.339).
re
This might underline the crucial role of the foster family in exposing dogs to a rich environment and stimulations during the first year. Furthermore, the fact that Fear/Reactivity remains a factor that contributes
lP
significantly to inter-individual variability in highly selected lineages and breeds of dogs remains surprising, since previous studies have shown that this trait can be relatively well selected for (e.g. Boissy, 1995). The
ur na
cause of Fearfulness/Reactivity in dogs is multifactorial and might include inherited characteristics, environmental influences, and learnt experiences throughout ontogeny. Indeed, multiple studies have demonstrated the impact of environmental enrichment and stressor exposure during puppyhood, as well as stimulation and mother exposure during the postnatal and weaning periods (for a review see Rooney et al.,
Jo
2016). It would be of interest for future studies to further investigate the impact of early exposure and environmental enrichment in our population of service dogs.
4.3. Prediction of dog qualification/disqualification
20
The Fear/Reactivity dimensions from both QFFs at 6 and 12 months and the BE at 1 year of age were all significantly associated with the success/failure of qualifying as a service dog. In the previous literature, Fear/Reactivity has been similarly identified as a reliable trait to predict adult dog disqualification from guide dog programmes (Batt et al., 2008; Duffy and Serpell, 2012; Goddard and Beilharz, 1982, 1983, 1984, 1986; Riemer et al., 2016; Tomkins et al., 2011). The main rationale for personality traits such as Fear/Reactivity contributing to dog disqualification is that fearfulness reduces a dog’s reliability and ability
of
to work. Indeed, during their work, service dogs encounter many types of situations and stimuli, and excessive fear reactions and restlessness will increase distractibility and unsuitability for work (Goddard
ro
and Beilharz, 1984; Tomkins et al., 2011).
-p
More than just predicting the qualification/disqualification status, the results also highlight that the Fear/Reactivity dimension was also predictive of the future type of work the dogs will be assigned to.
re
Indeed, the Fear/Reactivity dimension extracted from both QFFs at 6 and 12 months of age was predictive of assignment as a dog for ASD children, as a breeder, and as guide dog, whereas it was only predictive of
lP
assignment as a guide dog when extracted from the BE at 1 year of age. The present results seem to indicate that stronger requirements on the fearfulness trait are needed to qualify as a guide dog. Applying a stronger
ur na
requirement for guide dogs makes sense, since blind beneficiaries might be exposed to a higher risk if paired with a more fearful dog, whereas, for example, beneficiaries with reduced mobility have the ability (through sight) to anticipate, interpret, and avoid a fearful situation/stimulus, and to handle the dog. Our results suggest that Fear/Reactivity might play a different role according to the programme of working dogs and
Jo
calls for studies to investigate the distinct influence of personality dimensions on those various programmes. These results emphasise the importance of additional in-depth studies of fearfulness and emotional
reactivity in domestic and familiar animals, as noted by Boissy (1995). They also highlight the importance of early consideration and measurement of fearful personality traits in potential future service dogs for organisations providing this type of dog, since it might help to avoid heavy investments (i.e. approximately $25000–$35000 per dog) in dogs at risk of developing unsuitable personality profiles.
21
4.4. Differences according to sex and breed Neither the dog’s sex nor breed had any association with the dog’s disqualification for behavioural issues. However, differences according to sex on qualification in a specific programme were observed, with males qualifying more as guide and assistance dog, and females qualifying more as dogs for ASD children. In line with the stronger requirement on Fear/Reactivity in guide dogs and the higher likelihood of males to qualify as guide dogs, we also observed that at 1 year of age, males were significantly less fearful than
of
females. This result is in agreement with the previous literature indicating that female dogs present more
ro
Fear/Reactivity issues than males (Asp et al., 2015; Fratkin, 2017; Willis, 1995; Wilsson and Sundgren, 1997). These results are likely due to sex differences. However, we cannot exclude the possibility that the
-p
difference in age of neutering between males and females could have had some effects, since impacts of steroids hormones on emotional reactivity have been demonstrated in other species (e.g. Boissou, 1990;
re
Morgan and Pfaff, 2001).
The results highlighted that variations in certification in specific service programmes according to
lP
breed were present: Labradors and Saint-Pierres qualified more as dogs for ASD children, Saint-Pierres qualified less as assistance dogs, and Labernese qualified more as guide dogs. Furthermore, in line with
ur na
those variations and as referred in the previous literature, differences in Fear/Reactivity at 1 year of age according to breed were also observed in the present study (Asp et al., 2015; Fratkin, 2017; Serpell & Duffy, 2014; Wilsson and Sundgren, 1997). Indeed, breeds used by the Foundation in its early years (i.e. Royal Poodles, Golden Retrievers, Golden Labrador, and Bouvier des Flandres) were more fearful than the breeds
Jo
currently used by the Foundation (i.e. Labrador, Bernese Mountain, Labernese, and Saint-Pierre), while Labradors were less fearful. Since the selection of dog breeds has been made based on behavioural characteristics relevant to service dog requirements, it is thus no surprise that breeds differ on a characteristic as fundamental as fear, and that their likelihood to be selected in a specific programme vary in line with this characteristic. However, Labradors being less fearful and certifying less as guide dogs might seem surprising, since the dog’s Fear/Reactivity dimension from the BE at 1 year of age was significantly
22
predictive of qualification as a guide dog. This result likely indicates that components other than Fear/Reactivity, such as physical characteristics (notably a dog’s height so that its harness is at hand level for the beneficiary) or other personality traits, contribute to selection as a guide dog. There is a lot of variation in the breeds used by different organisations providing service dogs. It would thus be of importance for future studies on the personalities of working or service dogs to consider the impact of sex and breed on their results. Furthermore, it would be of interest for future studies to
of
investigate the impact of those parameters on early predictors of a dog’s personality. Indeed, some
ro
researchers suggest that behaviours and personality traits might develop at different rates and in different
-p
directions according to the breed (see Riemer et al., 2016).
re
4.5. Limitations and future directions
In the previous literature, and notably in the Jones and Gosling model (2005), four other personality
lP
traits defining a dog’s personality have been commonly reported (see Fratkin, 2017): activity, responsiveness to training, submission, and sociability and aggression. These traits have been demonstrated
ur na
to be predictive of a puppy’s and adult dog’s future disqualification (Arata et al., 2010; Asher et al., 2013; Batt et al., 2008; Duffy and Serpell, 2012; Kobayashi et al., 2012; Serpell and Hsu, 2001). Personality dimensions rely heavily on the instruments used to assess them (Diedrich and Giffroy, 2006; DowlingGuyer et al., 2011; Hsu and Serpell, 2003; McGarrity et al., 2015; Mirkó et al., 2012; Serpell and Hsu,
Jo
2005). Our results are thus limited by the tools built and used by the Mira Foundation, which could prevent the extraction of other meaningful personality dimensions. Another limitation is that inter-rater reliability could not be tested, since the scoring of only one family member/dog trainer for each dog was registered in the database created by the Mira Foundation. Therefore, as with many previous investigations, the present study relied on data acquired through tools developed by the organisation itself. Rather than completely changing their methods, it would be of interest for organisations providing service dogs to integrate
23
additional standardised tests and questionnaires (i.e. tests/questionnaires that have been demonstrated to be reliable and valid), such as the C-BARQ (Hsu and Serpell, 2003). However, concerning behavioural assessment, as pointed out by numerous authors (e.g., Diedrich and Giffroy, 2006; McGarrity et al., 2015), many behavioural tests reported in the literature lack a clear description of their materials and procedures, and no consensus on a standardised method to test a dog’s behaviour and personality has been reached. However, adding this type of test to their standard procedure would not only allow organisations to draw
of
more reliable conclusions, but would also facilitate the development of inter-organisation studies.
ro
When considering the reliability within factors, one might notice that Cronbach’s alphas on the extracted factors were less than 0.7, which could indicate that the behavioural assessments were highly
-p
variable and thus resulted in a lot of heterogeneity. In the literature, most studies relying on dimension reduction either had Cronbach alphas less than 0.7 (e.g. McGaritty et al., 2016) or did not report them (e.g.
re
de Palma et al., 2005; Kobayashi et al., 2013; Svartberg and Forkman, 2002; Wilsson and Sinn, 2012). Compared to the few studies reporting Cronbach alphas above 0.7, the present study used questionnaires
lP
with closed-ended question instead of Likert scales (Ley et al., 2008; Serpell and Hsu, 2001) and did not rely on scoring by trainers familiar with the dog for dogs over 1 year of age (Arata et al., 2010). Furthermore,
ur na
the heterogeneity might also originate from the long time period over which data were acquired (i.e. from 1987 to 2016), during which the assessment and scoring might have evolved. To conclude, our study suggests that data provided directly by a non-academic organisation have a huge potential to help us better understand the factors influencing dogs’ personalities and their suitability
Jo
for training. Despite the limitations inherent to the situation, our results indicate that Fear/Reactivity remains an important factor to consider even in a population of dogs selected for generations as working dogs, and thus emphasises the importance of the field of research dedicated to emotions and personality in animals.
AUTHOR DECLARATION
24
1) We wish to confirm that there are no known conflicts of interest associated with this publication and there has been no significant financial support for this work that could have influenced its outcome. 2) We confirm that the manuscript has been read and approved by all named authors and that there are no other persons who satisfied the criteria for authorship but are not listed. We further confirm that the order of authors listed in the manuscript has been approved by all of us.
of
3) We confirm that neither the entire paper nor any of its content has been submitted, published, or accepted by another journal. The paper will not be submitted elsewhere if accepted for publication in the Journal.
-p
ro
4) We understand that the Corresponding Author is the sole contact for the Editorial process (including Editorial Manager and direct communications with the office). He is responsible for communicating with the other authors about progress, submissions of revisions and final approval of proofs.
re
Authors and affiliations:
Dr. Dollion, N., Laboratoire d’Observation et d’Éthologie Humaine du Québec, School of Psychoeducation, University of Montreal, Montreal, Quebec, Canada; Mira Foundation, SainteMadeleine, Quebec, Canada
-
Amélie Paulus, Laboratoire d’Observation et d’Éthologie Humaine du Québec, School of Psychoeducation, University of Montreal, Montreal, Quebec, Canada.
-
Dr. Noël Champagne, Mira Foundation, Sainte-Madeleine, Quebec, Canada
-
Nicolas St-Pierre, Mira Foundation, Sainte-Madeleine, Quebec, Canada
-
Éric St-Pierre, Mira Foundation, Sainte-Madeleine, Quebec, Canada
-
Dr. Marcel Trudel, Department of Psychoeducation, University of Sherbrooke, Montreal, Quebec, Canada
Jo
ur na
lP
-
-
Pr. Pierrich Plusquellec, Laboratoire d’Observation et d’Éthologie Humaine du Québec, School of Psychoeducation, University of Montreal, Montreal, Quebec, Canada.
Acknowledgements
25
We gratefully thank all the trainers and all the staff from the Mira Foundation for helping us to conduct this research, for their cooperation, and for their work in the creation of the considerable database we analysed in the present study. This work was supported by the Marcelle and Jean Coutu Foundation, the Chagnon Foundation, and the Jean-Louis Lévesque Foundation. This work was also supported in part by a grant, Mitacs Elevate, for a postdoctoral student. Finally, we owe special thanks to E. Saint-Pierre, the creator of the Mira Foundation, and N. Saint-Pierre, the general director of the Mira Foundation, who
Jo
ur na
lP
re
-p
ro
of
allowed us to perform the present study.
26
References
Arata, S., Momozawa, Y., Takeuchi, Y., Mori, Y., 2010. Important behavioral traits for predicting guide dog qualification. J. Vet. Med. Sci. 72(5), 539–545. https://doi.org/10.1292/jvms.09-0512
Asher, L., Blythe, S., Roberts, R., Toothill, L., Craigon, P. J., Evans, K.M., Green, M.J., England, G.C.W., 2013. A standardized behavior test for potential guide dog puppies: Methods and association with
of
subsequent success in guide dog training. J. Vet. Behav. 8, 431–438.
ro
https://doi.org/10.1016/j.jveb.2013.08.004
Asp, H.E., Fikse, W.F., Nilsson, K., Strandberg, E., 2015. Breed differences in everyday behaviour of
-p
dogs. Appl. Anim. Behav. Sci. 169, 69–77. https://doi.org/10.1016/J.APPLANIM.2015.04.010
re
Batt, L.S., Batt, M.S., Baguley, J.A., McGreevy, P.D., 2008. Factors associated with success in guide dog
lP
training. J. Vet. Behav. 3, 143–151. https://doi.org/10.1016/j.jveb.2008.04.003
Boissy, A., 1995. Fear and fearfulness in animals. Q. Rev. Biol. 70(2), 165–191.
ur na
https://doi.org/10.1086/418981
Bouissou, M.F., 1990. Effects of estrogen treatment on dominance relationships in cows. Horm. Behav. 24(3), 376–387. https://doi.org/10.1016/0018-506X(90)90016-Q
Bollen, K.S., Horowitz, J., 2008. Behavioral evaluation and demographic information in the assessment of
Jo
aggressiveness in shelter dogs. Appl. Anim. Behav. Sci. 112, 120–135. https://doi.org/10.1016/j.applanim.2007.07.007
Bremhorst, A., Mongillo, P., Howell, T., Marinelli, L., 2018. Spotlight on assistance dogs—legislation, welfare and research. Animals 8, 129–148. https://doi.org/10.3390/ani8080129
27
Carere, C., Locurto, C., 2011. Interaction between animal personality and animal cognition. Curr. Zool. 57(4), 491–498. https://doi.org/10.1093/czoolo/57.4.491
De Palma, C., Viggiano, E., Barillari, E., Palme, R., Dufour, A.B., Fantini, C., Natoli, E., 2005. Evaluating the temperament in shelter dogs. Behaviour 142, 1307–1328. https://doi.org/10.1163/156853905774539337
of
Diederich, C., Giffroy, J.M., 2006. Behavioural testing in dogs: aA review of methodology in search for
ro
standardisation. Appl. Anim. Behav. Sci. 97, 51–72. https://doi.org/10.1016/j.applanim.2005.11.018 Dowling-Guyer, S., Marder, A., D’Arpino, S., 2011. Behavioral traits detected in shelter dogs by a
re
https://doi.org/10.1016/j.applanim.2010.12.004
-p
behavior evaluation. Appl. Anim. Behav. Sci. 130, 107–114.
Duffy, D.L., Serpell, J.A., 2012. Predictive validity of a method for evaluating temperament in young
lP
guide and service dogs. Appl. Anim. Behav. Sci. 138, 99–109. https://doi.org/10.1016/j.applanim.2012.02.011
ur na
Foyer, P., Bjällerhag, N., Wilsson, E., Jensen, P., 2014. Behaviour and experiences of dogs during the first year of life predict the outcome in a later temperament test. Appl. Anim. Behav. Sci. 155, 93–100. https://doi.org/10.1016/j.applanim.2014.03.006
Jo
Fratkin, J.L., 2017. Personality in dogs, in: Vonk, J., Weiss, A., Kuczak, S.A. (Eds.), Personality in Nonhuman Animals. Springer, Gewerbestrasse, Switzerland, pp. 205–224. https://doi.org/0.1007/978-3-319-59300-5
Fratkin, J.L., Sinn, D. L., Patall, E.A., Gosling, S.D., 2013. Personality consistency in dogs: A metaanalysis. PLoS ONE 8(1), e54907. https://doi.org/10.1371/journal.pone.0054907
28
Gazzano, A., Mariti, C., Sighieri, C., Ducci, M., Ciceroni, C., McBride, E.A., 2008. Survey of undesirable behaviors displayed by potential guide dogs with puppy walkers. J. Vet. Behav. 3, 104–113. https://doi.org/10.1016/j.jveb.2008.04.002
Goddard, M.E., Beilharz, R.G., 1982. Genetic and environmental factors affecting the suitability of dogs as Guide Dogs for the Blind. Theor. Appl. Genet. 62(2), 97–102.
of
https://doi.org/10.1007/BF00293339
Goddard, M.E., Beilharz, R.G., 1983. Genetics of traits which determine the suitability of dogs as guide-
ro
dogs for the blind. Appl. Anim. Ethol. 9, 299–315. https://doi.org/10.1016/0304-3762(83)90010-X
-p
Goddard, M.E., Beilharz, R.G., 1984. A factor analysis of fearfulness in potential guide dogs. Appl. Anim.
re
Behav. Sci. 12, 253–265.
Goddard, M.E., Beilharz, R.G., 1986. Early prediction of adult behaviour in potential guide dogs. Appl.
lP
Anim. Behav. Sci. 15(3), 247–260. https://doi.org/10.1016/0168-1591(86)90095-X
Gosling, S.D., 2001. From mice to men: What can we learn about personality from animal research?
ur na
Psychol. Bull. 127(1), 45.
Gosling, S.D., John, O.P., 1999. Personality dimensions in nonhuman animals: A cross-species review. Curr. Dir. Psychol. Sci. 8(3), 69–75. https://doi.org/10.1111/1467-8721.00017
Jo
Harvey, N.D., Craigon, P.J., Sommerville, R., McMillan, C., Green, M., England, G.C.W., Asher, L., 2016. Test-retest reliability and predictive validity of a juvenile guide dog behavior test. J. Vet. Behav. 11, 65–76. https://doi.org/10.1016/j.jveb.2015.09.005
29
Hsu, Y., Serpell, J.A., 2003. Development and validation of a questionnaire for measuring behavior and temperament traits in pet dogs. J. Am. Vet. Med. Assoc. 223(9), 1293–1300. https://doi.org/10.2460/javma.2003.223.1293
Jones, A.C., Gosling, S.D., 2005. Temperament and personality in dogs (Canis familiaris): A review and evaluation of past research. Appl. Anim. Behav. Sci. 95, 1–53.
of
https://doi.org/10.1016/j.applanim.2005.04.008
Kobayashi, N., Arata, S., Hattori, A., Kohara, Y., Kiyokawa, Y., Takeuchi, Y., Mori, Y., 2013.
ro
Association of puppies’ behavioral reaction at five months of age assessed by questionnaire with
-p
their later ‘distraction’ at 15 months of age, an important behavioral trait for guide dog qualification. J. Vet. Med. Sci. 75(1), 63–67. https://doi.org/10.1292/jvms.12-0148
re
Ley, J., Bennett, P., Coleman, G., 2008. Personality dimensions that emerge in companion canines. Appl.
lP
Anim. Behav. Sci. 110, 305–317. https://doi.org/10.1016/j.applanim.2007.04.016
McGarrity, M.E., Sinn, D.L., Gosling, S.D., 2015. Which personality dimensions do puppy tests measure?
ur na
A systematic procedure for categorizing behavioral assays. Behav. Processes 110, 117–124. https://doi.org/10.1016/j.beproc.2014.09.029
Miklósi, Á., Turcsán, B., Kubinyi, E., 2014. The Personality of Dogs, in: Kaminski, J., Marshall-Pecini, S. (Eds.), The Social Dog: Behavior and Cognition. Elsevier Science, Burlington, pp. 191–222.
Jo
https://doi.org/10.1016/B978-0-12-407818-5.00007-3
Mirkó, E., Kubinyi, E., Gácsi, M., Miklósi, Á., 2012. Preliminary analysis of an adjective-based dog personality questionnaire developed to measure some aspects of personality in the domestic dog (Canis familiaris). Appl. Anim. Behav. Sci. 138, 88–98. https://doi.org/10.1016/j.applanim.2012.02.016
30
Morgan, M.A., Pfaff, D.W., 2001. Effects of estrogen on activity and fear-related behaviors in mice. Horm. Behav. 40(4), 472–482. https://doi.org/10.1006/hbeh.2001.1716
Riemer, S., Mills, D.S., Wright, H., 2014a. Impulsive for life? The nature of long-term impulsivity in domestic dogs. Anim. Cogn. 17(3), 815–819. https://doi.org/10.1007/s10071-013-0701-4
Riemer, S., Müller, C., Virányi, Z., Huber, L., Range, F., 2014b. The predictive value of early behavioural
of
assessments in pet dogs—a longitudinal study from neonates to adults. PLoS ONE 9(7), e101237.
ro
https://doi.org/10.1371/journal.pone.0101237
Riemer, S., Müller, C., Virányi, Z., Huber, L., Range, F., 2016. Individual and group level trajectories of
re
https://doi.org/10.1016/j.applanim.2016.04.021
-p
behavioural development in Border collies. Appl. Anim. Behav. Sci. 180, 78–86.
Rooney, N. J., Clark, C. C., Casey, R. A., 2016. Minimizing fear and anxiety in working dogs: A review.
lP
J. Vet. Behav. 16, 53–64. https://doi.org/10.1016/j.jveb.2016.11.001
Schalke, E., Ott, S.A., von Gaertner, A.M., Hackbarth, H., Mittmann, A., 2008. Is breed-specific
ur na
legislation justified? Study of the results of the temperament test of Lower Saxony. J. Vet. Behav. 3, 97–103. https://doi.org/10.1016/j.jveb.2007.10.004
Serpell, J.A., Duffy, D.L., 2014. Dog Breeds and their Behavior, in: Horowitz, A. (Ed.), Domestic Dog
Jo
Cognition and Behavior: The Scientific Study of Canis familiaris. Springer, Berlin, pp. 31–57. https://doi.org/10.1007/978-3-642-53994-7_2
Serpell, J.A., Hsu, Y., 2001. Development and validation of a novel method for evaluating behavior and temperament in guide dogs. Appl. Anim. Behav. Sci. 72, 347–364. https://doi.org/10.1016/S01681591(00)00210-0
31
Serpell, J.A., Hsu, Y., 2005. Effects of breed, sex, and neuter status on trainability in dogs. Anthrozoos 18(3), 196–207. https://doi.org/10.2752/089279305785594135
Sinn, D.L., Gosling, S.D., Hilliard, S., 2010. Personality and performance in military working dogs: Reliability and predictive validity of behavioral tests. Appl. Anim. Behav. Sci. 127, 51–65. https://doi.org/10.1016/j.applanim.2010.08.007
of
Stamps, J., Groothuis, T.G., 2010. The development of animal personality: Relevance, concepts and
ro
perspectives. Biol. Rev. 85(2), 301–325.
Slabbert, J.M., Odendaal, J.S.J., 1999. Early prediction of adult police dog efficiency—a longitudinal
-p
study. Appl. Anim. Behav. Sci. 64, 269–288. https://doi.org/10.1016/S0168-1591(99)00038-6
re
Svartberg, K., 2002. Shyness-boldness predicts performance in working dogs. Appl. Anim. Behav. Sci.
lP
79, 157–174.
Svartberg, K., 2007. Individual Differences in Behaviour-Dog Personality, in: Jensen, P. (Ed.), The
ur na
Behavioural Biology of Dogs. CAB International, Cambridge, pp. 182–206.
Svartberg, K., Forkman, B., 2002. Personality traits in the domestic dog (Canis familiaris). Appl. Anim. Behav. Sci. 79, 133–155. https://doi.org/10.1016/S0168-1591(02)00121-1
Taylor, K.D., Mills, D.S., 2006. The development and assessment of temperament tests for adult
Jo
companion dogs. J. Vet. Behav. 1, 94–108. https://doi.org/10.1016/j.jveb.2006.09.002
Tomkins, L.M., Thomson, P.C., McGreevy, P.D., 2011. Behavioral and physiological predictors of guide dog success. J. Vet. Behav. 6, 178–187. https://doi.org/10.1016/j.jveb.2010.12.002
32
van der Borg, J.A.M., Netto, W.J., Planta, D.J.U., 1991. Behavioural testing of dogs in animal shelters to predict problem behaviour. Appl. Anim. Behav. Sci. 32(2–3), 237–251. https://doi.org/10.1016/S0168-1591(05)80047-4
Weiss, E., 2002. Selecting shelter dogs for service dog training. J. Appl. Anim. Welf. Sci. 5, 43–62. https://doi.org/10.1207/S15327604JAWS0501_4
of
Willis, M.B., 1995. Genetic aspects of dog behaviour with particular reference to working ability, in: J. Serpell (Ed.), The Domestic Dog: Its Evolution, Behaviour, and Interactions With People.
ro
Cambridge University Press, Cambridge, pp. 51–64.
-p
Wilsson, E., Sinn, D.L., 2012. Are there differences between behavioral measurement methods? A comparison of the predictive validity of two ratings methods in a working dog program.
re
Appl. Anim. Behav. Sci. 141, 158–172. doi:10.1016/j.applanim.2012.08.012
lP
Wilsson, E., Sundgren, P.E., 1997. The use of a behaviour test for the selection of dogs for service and breeding, I: Method of testing and evaluating test results in the adult dog,
ur na
demands on different kinds of service dogs, sex and breed differences. Appl. Anim. Behav.
Jo
Sci. 53, 279–295. https://doi.org/10.1016/S0168-1591(96)01174-4
33
Table 1. Factor loadings for each question item from the questionnaire addressed to the foster family when the dog was 6 months old. Question item and main topic F3
F4
F5
F6
F7
0.009 0.004 0.023 0.006
0.027 0.000 0.536 0.424
0.034 0.007 0.003 0.000
0.177 0.000 0.010 0.001
0.000 0.151 0.002 0.021
0.249 0.168 0.048 0.153
0.008 0.007 0.015 0.000 0.000 0.001
0.404 0.388 0.269 0.199 0.227 0.062
0.020 0.044 0.034 0.005 0.029 0.036
0.619 0.626 0.016 0.000 0.002 0.000
0.237 0.266 0.010 0.091 0.002 0.085
0.014 0.016 0.001 0.090 0.003 0.028
0.019 0.024 0.031 0.013 0.008 0.067
0.001 0.002
0.130 0.132
0.472 0.625 0.577 0.227 0.337 0.257 0.274 0.001 2.806 0.132 13.16 0.677 0.140 Fear/ Reactivity
0.011 0.014 0.014 0.008 0.004 0.026 0.016 0.003 1.950 0.084 8.39 0.513 0.098
ro
0.012 0.001
0.004 0.008
0.065 0.076
0.108 0.052
0.145 0.016
0.004 0.005 0.001 0.000 0.003 0.001 0.007 0.279 1.468 0.063 6.31 0.336 0.073
0.000 0.000 0.000 0.001 0.000 0.001 0.002 0.011 1.333 0.057 5.73 0.263 0.067
0.037 0.033 0.021 0.042 0.049 0.061 0.033 0.001 1.295 0.056 5.57 0.240 0.065
0.080 0.104 0.086 0.059 0.040 0.206 0.119 0.000 1.179 0.051 5.07 0.160 0.059
0.000 0.000 0.000 0.029 0.005 0.003 0.001 0.101 1.081 0.047 4.65 0.079 0.054
re
lP
of
0.001 0.000 0.001 0.000
ur na
Personality factors
F2
-p
F1 House training Issues with toilet training Eats its own faeces Does its business on leash Does its business on hard surfaces Activity/Behaviours Time to calm down with visitors Time to calm down outside Jumps on people Destroys objects Chews Runs away Attractions Steals or begs for food Climbs on furniture Fears Afraid of children Afraid of women Afraid of men Afraid of other animals Afraid of new places Afraid of noises Afraid of traffic At ease on staircases Eigenvalue Contribution ratio Percentage of explained variance Cronbach’s Alpha Inertia
Jo
No absolute values under 0.1 are referred to in this table. The underlined numbers refer to those that were selected as the composing items for each factor.
34
Table 2. Factor loadings for each question item from the questionnaire addressed to the foster family when the dog was 12 months old. Question item and main topic F4
F5
F6
0.063 0.001 0.016 0.006
0.000 0.006 0.006 0.000
0.019 0.011 0.358 0.247
0.007 0.001 0.227 0.190
0.040 0.018 0.000 0.000
0.000 0.032 0.003 0.013
0.161 0.164 0.084 0.115 0.113 0.049
0.191 0.182 0.186 0.089 0.078 0.089
0.229 0.245 0.021 0.006 0.078 0.089
0.147 0.066
0.076 0.060
0.273 0.244 0.002 0.083 0.000 0.010
0.317 0.250 0.003 0.026 0.002 0.010
0.016 0.003
0.071 0.144
0.065 0.040
0.017 0.020
0.004 0.000 0.002 0.001 0.030 0.007 0.023 0.266 1.499 0.065 6.45 0.350 0.075
0.002 0.007 0.007 0.004 0.006 0.018 0.013 0.098 1.362 0.059 5.86 0.280 0.068
0.015 0.024 0.009 0.017 0.086 0.081 0.113 0.144 1.266 0.054 5.44 0.221 0.063
0.025 0.063 0.043 0.062 0.046 0.090 0.130 0.077 1.230 0.053 5.29 0.197 0.062
ro
0.210 0.272 0.007 0.066 0.000 0.010
-p
re
lP
0.169 0.263 0.216 0.390 0.217 0.347 0.149 0.017 0.165 0.027 0.182 0.005 0.110 0.006 0.059 0.021 2.252 2.040 0.100 0.088 10.04 8.77 0.585 0.537 0.113 0.102 Reactivity/ Fear of human/ Activity Reactivity
of
F3
Jo
Personality factors
F2
ur na
House training Issues with toilet training Eats its own faeces Does its business on a leash Does its business on hard surfaces Activity/Behaviours Time to calm down with visitors Time to calm down outside Jumps on people Destroys objects Chews Runs away Attractions Steals or begs for food Climbs on furniture Fears Afraid of children Afraid of women Afraid of men Afraid of other animals Afraid of new places Afraid of noises Afraid of traffic At ease on staircases Eigenvalue Contribution ratio Percentage of explained variance Cronbach’s Alpha Inertia
F1
No absolute values under 0.1 are referred to in this table. The underlined numbers refer to those that were selected as the composing items for each factor.
35
Table 3. Factor loading for each test of the behavioural evaluation at 1 year of age.
F4 -0.007 0.006 -0.039 -0.505 0.143 0.749 0.235 -0.033 0.027 0.124 0.044 0.392 1.008 0.084 8.40 0.082
of
F3 0.020 0.060 0.012 -0.090 -0.040 0.054 -0.695 -0.425 -0.037 0.301 -0.161 -0.532 1.033 0.086 8.61 0.043
ro
F2 -0.034 -0.112 -0.019 0.218 0.131 0.194 0.017 0.462 0.749 0.591 0.670 0.087 1.740 0.145 14.50 0.538 Fear/ Reactivity
-p
Activity alone Cat attraction Gluttony Walking in a pack Simple walking Odd stimuli Walking on staircases Bridge with traffic Mannequin's corridor Noise sensitivity Stranger encounter Competition for resource Eigenvalue Contribution ration Percentage of explained variance Cronbach's alpha
F1 0.733 0.601 0.607 0.446 0.617 0.101 0.141 -0.011 0.024 0.088 -0.114 0.191 1.997 0.167 16.65 0.510
re
Behavioural test
Personality factors
Jo
ur na
lP
No absolute values under 0.4 are referred to in this table. The underlined numbers refer to those that were selected as the composing items for each factor.
36
Table 4. Results of the binary logistic regression of the Fear/Reactivity factor extracted from the questionnaire addressed to the foster family when the dog was 6 months old.
No 1233 1552 1500 1409 1719
QFF at 6 months of age B SE OR CI -0.249 0.049 0.780 0.708–0.858 0.237 0.096 1.267 1.050–1.529 0.066 0.071 1.013 0.931–1.227 0.237 0.076 1.287 1.092–1.470 0.454 0.240 0.922 0.984–2.523
p <0.001 0.013 0.347 0.002 0.059
of
Disqualified Guide dog Assist. Dog ASD Dog Breeder
Yes 555 226 278 369 59
Jo
ur na
lP
re
-p
ro
Abbreviations: Yes: the number of dogs assigned to the corresponding programme or disqualified based on behaviour; No: the number of dogs not assigned to the corresponding programme or not disqualified based on behaviour; B: unstandardised regression coefficients; SE: standard error; OR: odds ratio; CI: 95 % confidence interval; p: significance probability.
37
Table 5. Results of the binary logistic regressions of the Fear/Reactivity factor extracted from the questionnaire addressed to the foster family when the dog was 12 months old.
Disqualifi 101 215 0 9 ed Guide dog Assist. Dog ASD Dog Breeder
270 8 256 609 0 262 545 4 303 130 9 461
Factor 1 QFF at 12 months of age B SE OR CI p 0.03 0.65 0.602– <0.00 0.43 9 1 0.703 1 0 0.33 0.06 1.40 1.248– <0.00 9 0 4 1.580 1 0.06 0.04 1.06 0.971– 0.185 2 7 4 1.167 0.25 0.05 1.29 1.169– <0.00 7 1 3 1.430 1 0.40 0.11 1.49 1.207– <0.00 3 0 7 1.857 1
Yes 101 0 461 609 545 130
Factor 2 QFF at 12 months of age No B SE OR CI 215 0.03 0.84 0.7860.16 9 8 6 0.911 7 270 0.20 0.06 1.22 1.0888 6 2 9 1.388 256 0.12 0.04 1.12 1.028– 0 1 8 9 1.240 262 0.01 0.05 1.01 0.917– 4 5 2 5 1.122 303 0.12 0.11 1.13 0.908– 9 5 3 3 1.415
p <0.00 1 0.001
of
No
ro
Yes
0.011 0.777 0.269
Jo
ur na
lP
re
-p
Abbreviations: Yes: the number of dogs assigned to the corresponding programme or disqualified based on behaviour; No: the number of dogs not assigned to the corresponding programme or not disqualified based on behaviour; B: unstandardised regression coefficients; SE: standard error; OR: odds ratio; CI: 95 % confidence interval; p: significance probability.
38
Table 6. Results of the binary logistic regressions of the Fear/Reactivity factor extracted from the behavioural evaluation at 1 year of age.
Yes 1040 637 718 555 157
No 2423 2826 2745 2908 3306
p <0.001 <0.001 0.859 0.113 0.094
of
Disqualified Guide dog Assist. Dog ASD Dog Breeder
BE at 1 year of age B SE OR CI -0.584 0.039 0.558 0.517–0.602 0.950 0.068 2.586 2.264–2.955 -0.007 0.042 0.993 0.914–1.078 0.076 0.048 1.079 0.982–1.185 0.149 0.089 1.161 0.975–1.382
Jo
ur na
lP
re
-p
ro
Abbreviations: Yes: the number of dogs assigned to the corresponding programme or disqualified based on behaviour; No: the number of dogs not assigned to the corresponding programme or not disqualified based on behaviour; B: unstandardised regression coefficients; SE: standard error; OR: odds ratio; CI: 95 % confidence interval; p: significance probability.
39