DEVELOPMENT AND VALIDATION OF A BEHAVIORAL ASSESSMENT TOOL TO MONITOR SHELTER
DOG COPING BEHAVIOR
By
Anna Katherine Breithaupt
A THESIS
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Animal Science – Master of Science
2023
ABSTRACT
Dogs that enter animal shelters experience a variety of well-documented environmental
stressors that may lead to chronic stress. Stress response can become maladaptive when
shelter dogs are unable to effectively cope with their environments. Often pharmaceutical
intervention is necessary to prevent further behavioral decline. However, access to
pharmaceuticals is dependent on shelter resources with many lacking veterinary behaviorists
on staff. Furthermore, a reliable assessment that would accurately communicate shelter dogs’
needs to clinicians does not exist. In response, a novel behavior assessment was developed
based on extant literature and expertise of a board-certified veterinarian behaviorist and
doctorate in ethology, designed to provide enrichment while remaining feasible for shelter staff
to implement. During the preliminary pilot, the tool went through several iterations based on
evaluation of reliability and clinical relevance.
June-October 2022, 91 single-housed shelter dogs, ≥ 12 weeks of age were assessed
either indoors (n = 43) or outdoors (n = 48). Dogs were assessed in real-time by two raters for
inter-rater reliability and video recorded for intra-rater reliability. Inter- and intra-rater percent
agreement was moderate to near perfect. To establish validity criterion, a board-certified
veterinarian behaviorist blinded to coping score diagnosed participating dogs as either adaptive
coping (AC), maladaptive coping anxious-avoidant (MC-AA) or excessive-aroused (MC-EA) using
assessment video. There was no evidence of a difference in coping score between assessment
areas; therefore, indoor and outdoor assessments were pooled for validity analysis. At
statistical significance, the tool was able to differentiate MC-AA dogs from AC and MC-EA but
was unable to differentiate MC-EA from AC, although MC-EA dogs had higher marginal mean
total score than AC dogs. As expected, the marginal mean for MC-AA dogs was negative and
MC-EA positive, between which fell the marginal mean for AC dogs. Based on reliability and
validity, the tool was further refined for use in future studies including, establishment of a
diagnostic scale for interpreting coping score, and evaluation of tool’s ability to track changes in
coping behavior over time and in response to pharmaceutical interventions.
This thesis is dedicated to Boone, forever in my heart.
Thank you for allowing me to experience the depth of your soul.
iii
ACKNOWLEDGEMENTS
I would like to express my deepest gratitude to the following individuals and groups,
without whom this thesis would not have been possible. Dr. Jacquelyn Jacobs, for her invaluable
guidance, unwavering support, and co-development of this research. Her expertise and
mentorship were instrumental in shaping this work. Dr. Marie Hopfensperger, for her significant
contributions as a co-developer and committee member. Her insightful feedback and dedication
to this project enriched its quality. Dr. Hanne Hoffmann and Dr. Janice Siegford, for their
valuable insights and contributions as committee members. Their expertise enhanced the depth
and scope of this study. I extend my heartfelt thanks to the dedicated team of research
assistants - Kaitlyn Tree-Macneill, Nattawipa Ampaiwan, Taylor Stehouwer, Mary Gardella,
Kathryn Hurt, Kennedi Robinson, Brianna Yi, and Sophia Yabut. Their hard work and
commitment played a crucial role in the successful execution of this research. A special thanks
goes to the MSU CANR Statistical Consulting Center, specifically Guanqi Lu, Rabin KC, Sakib
Hasan, and Dr. Robert Tempelman, for their expertise in data analysis and statistical support. I
am grateful to Kyle MacMillan for his unofficial role as an IT consultant, providing crucial
technical support. I would also like to acknowledge the invaluable support of the Capital Area
Humane Society of Lansing staff including Julia, Lexy, Sam, Makenzie, Emma, Sarah, and
Elizabeth. Their cooperation and assistance in facilitating access to the shelter and the dogs
involved in the study were indispensable. The shelter dogs that participated in this study
deserve a special mention. Their willingness to be a part of this research provided essential
data, and their welfare is always at the forefront of our efforts.
To my family and friends, whose encouragement and belief in my abilities sustained me
throughout this journey, I am profoundly thankful. Lastly, I am indebted to my companions -
Boone, bunny Zo, and hermit crabs Hermes and Del Rey, whose antics have brought me joy.
Boone especially is my inspiration; our relationship motivated me to dedicate myself to
exploring the emotional complexity of dogs.
Thank you all for your invaluable contributions.
iv
TABLE OF CONTENTS
CHAPTER 1: LITERATURE REVIEW OF SHELTER STRESSOR EFFECTS ON BEHAVIOR AND
DEVELOPMENT OF BEHAVIOR ASSESSMENTS ................................................................................ 1
CHAPTER 2: DEVELOPMENT OF A RELIABLE BEHAVIOR ASSESSMENT TO MEASURE COPING
BEHAVIOR OF SHELTER DOGS ....................................................................................................... 19
CHAPTER 3: VALIDITY OF A BEHAVIOR ASSESSMENT TO MEASURE COPING BEHAVIOR OF
SHELTER DOGS .............................................................................................................................. 79
BIBLIOGRAPHY ............................................................................................................................ 120
APPENDIX A: ASSESSMENT FORM DRAFTS ................................................................................. 137
APPENDIX B: PILOT DOCUMENTS ............................................................................................... 148
APPENDIX C: TRAINING DOCUMENTS ........................................................................................ 150
APPENDIX D: ASSESSMENT SCORE RUBRIC ................................................................................ 168
APPENDIX E: REFINED ASSESSMENT FORM ................................................................................ 171
v
CHAPTER 1: LITERATURE REVIEW OF SHELTER STRESSOR EFFECTS ON BEHAVIOR AND
DEVELOPMENT OF BEHAVIOR ASSESSMENTS
1.1: Introduction
1.1.i: Domestication Juxtaposition: Dynamics of the Human-dog Co-evolution
Dogs (Canis lupus familiaris) have long forged an unparalleled bond with humans,
becoming integral members of our community. The profound connection between humans and
dogs tells the tale of co-evolution that has left an indelible mark on our emotional and
physiological ties (for reviews see Gee et al., 2021; McNicholas et al., 2005). Yet, while
domestication has endowed dogs with unique adaptations suited to enhancing our relationship,
including the seemingly innate ability to read human facial expressions and preference for
human proximity (Albuquerque & Resende, 2023; Lazzaroni et al., 2020; vonHoldt & Driscoll,
2017), they still retain ancestral traits that can put canine welfare at odds with modern
expectations. Working and companion dogs often find themselves in environments far removed
from the village settings that initially shaped our cross-species relationship. This disjunction
arises, in part, from our advanced interconnectedness. In our efforts to integrate animals into
our lifestyles, we frequently resort to physical confinement or impose restrictions on their
natural behaviors. Nowhere is this contrast more apparent than in the environments where
dogs are temporarily housed for re-homing. The stark disparity between the evolutionarily
familiar surroundings of the ancestral dog and the contemporary confines of animal shelters
often leaves many individuals struggling to navigate and adapt to their new circumstances.
1.1.ii: Animal Shelter Genesis
As invaluable as dogs are to the people they support, a myriad of complex factors
contributes to the pervasive occurrence of pet relinquishment which leads to the necessity of
re-homing shelters. While there are a handful of documented attempts to quantify shelter dogs
nationwide, estimating the total population in any given region is difficult due to the lack of a
complete national database (Downes et al., 2013; Patronek & Glickman, 1994; Rowan & Kartel,
2018). However, even considering the broad range possible indicated by the limited data
available, a significant number of individuals enter shelters or animal rescues with the most
recent scholarly data conglomerations floating around 5 million dogs annually (Rowan & Kartel,
1
2018; Woodruff & Smith, 2020), although national census data estimates that closer to 3.1
million dogs enter shelters nationwide (Pet statistics, ASPCA). However, the source census for
ASPCA Pet Statistics is the Shelter Animals Count database which relies on volunteer reporting
and therefore likely underestimates national intake. Often these shelters provide temporary
housing for surrendered dogs, strays, and those removed from owners for reasons of neglect or
abuse. During this transitional period, dogs are kenneled and cared for in these establishments,
until they leave under circumstances of return-to-owner, new adoptions, transfers, or death. In
the past, shelter overpopulation, exacerbated by resource shortages including staff labor,
funding, and space, has contributed to high euthanasia rates (Salman et al., 1998). Concerned,
people began to challenge the concept of the traditional shelter, defined as re-homing
establishments that euthanize to control population in addition to medical and behavioral
reasons.
Thereafter, many shelters enacted policies restricting instances of euthanasia to cases of
critical illness or poor temperament; these shelters, in contrast to traditional shelters, are
labeled “no-kill†shelters through definition by the American Veterinary Medical Association
(Brown et al., 2013). In response to restricted use of euthanasia, communities have
implemented various strategies to address overpopulation in shelters such as encouraging pet
retention through supportive programs and moving animals to alternative housing through
foster systems. However, relinquishment in combination with stray intake continues to fill
shelters with available dogs. Data indicates that national euthanasia rates are declining (Bartlett
et al., 2005; Protopopova, 2016; Rowan & Kartal, 2018), but concern for the potential increase
of extended stays as available dogs await adoption has led researchers to investigate factors
that impact length of stay. While there have not been studies directly comparing the length of
stay of dogs in no-kill versus traditional shelters, many researchers have relied on induced cause
and effect as the basis for identifying influential factors that may improve adoption rates
(Bradley & Rajendran, 2021; Brown et al., 2013, Normando et al., 2006). Behavioral factors such
body position in-kennel and activity have been shown to predict longer length of stay in
shelters, supporting the belief that behavior influences adopter choice (Protopopova & Wynne,
2014; Protopopova et al., 2014; Wells et al., 2002). Aiming resources at targeted behavioral
2
modification programs may augment the movement of adoptable dogs from shelters to homes,
leading many shelters to implement obedience and training classes, although few experimental
studies on behavior modification and likelihood of adoption currently exist (Herron et al., 2014;
Luescher & Medlock, 2009), as reviewed by Protopopova and Gunter (2017).
Compounded with previously established efforts at decreasing pet overpopulation, the
Covid-19 pandemic ushered in a shift in national pet ownership characterized by fewer dogs
entering shelters (Powell et al., 2021). Additionally, many sources anecdotally reported an
increase in adoptions through the onset of the pandemic (Szydlowski & Gragg, 2020), although
one study found that overall adoption rates have not changed with evidence of a decrease in
number of adoptions, likely reflecting decreased intake (Powell et al., 2021). However, it is
difficult to assess the degree to which the pandemic has influenced shelter trends due to
conflicting results across studies with some showing evidence that shelters were already
experiencing decreased intake and increased live release rates (Hawes et al., 2019) while others
indicate a pre-pandemic increase in total dog intake (Hawes et al., 2021). In one retrospective
study that spanned five years from 2016 through the first year of the pandemic in 2020,
researchers found that total intake decreased, and live release rates increased (Rodriguez et al.,
2022).
Contradictory to these findings, the most recent Shelter Animals Count report concludes
that intake has been steadily increasing since 2021 (Q2 2023 Analysis, Shelter Animals Count).
However, general intake/outcome trends may indicate that shelters are observing improved
occupancy flow. Despite this, dogs that enter shelters continue to experience widely variable
length of stays, creating situations in which dogs may spend several days to months housed at
an animal shelter. On average, census data indicates that shelter dogs are experiencing longer
length of stays (Brown et al., 2013; Wenstrup & Dowidchuk, 1999). Enabled to look beyond
overpopulation control, the shelter community is eager to direct resources toward improving
shelter dog welfare not only to appeal to potential adopters but also to provide individual dogs
with the best possible opportunity to successfully transition through the shelter system.
3
1.2: Shelter Environment
1.2.i: Environmental and Social Stressors
The effects of shelter-specific environmental and social factors have been studied in
dogs using physiological measures of stress such as cortisol, heart rate variability (HRV), and
immune function, as well as behavioral measures including activity level and ethograms, as
reviewed by Beerda and colleagues (1997) and Polgar and colleagues (2019) Studies of
physiological outcomes in kenneled dogs predominately measure hypothalamic-pituitary-
adrenal (HPA) axis activity (for review see Hennessy, 2013). The HPA axis is understood to be the
primary system activated in response to a stressor. Upon perceiving a threat to homeostasis, the
paraventricular nucleus (PVN) of the hypothalamus, a region of the brain, secretes
corticotropin-releasing factor (CRF) to the anterior pituitary gland located near the
hypothalamus. The anterior pituitary responds by releasing adreno-corticotropic hormone
(ACTH) into circulation, principally targeting the adrenal cortex. There, glucocorticoids are
released into circulation to receptive tissues and organs to prep the body for fight or flight
responses (for review see Smith & Vale, 2006). Activation of the HPA axis can be measured
through levels of glucocorticoids as cortisol concentration in plasma, saliva, urine, feces, or hair
(for review see Polgar et al., 2019). Several environmental and social factors inherent to shelters
have thus been identified as stressors that may illicit physiological stress responses and negative
affective states in dogs (for review see Protopopova, 2016).
Kennel environments, as experienced not only by shelter dogs but also working dogs,
laboratory animals, veterinary patients, and pets temporarily boarded away from their humans,
present environmental stressors inherent to confined housing (for review see Taylor & Mills,
2007). Restricted kennel areas, typically consisting of limited outdoor access and potentially
austere conditions, has been shown to increase stress behaviors and cortisol levels (Beerda et
al., 1999a; Hubrecht et al., 1992; Normando et al., 2014). Limited exercise and barren
environments may increase stress by restricting the extent of which a dog can express natural
behaviors. In one study, 25-minutes of human-guided exercise lowered salivary cortisol and
improved scores on a behavior test (Menor-Campos et al., 2011). Sensory stimuli can further
contribute to shelter dog stress as shelter design and management can lead to olfactory and
4
auditory overload. Upon entering a shelter, dogs are assailed with multiple strong odors
including conspecific pheromones and cleaning chemicals along with persistent exposure to
noise from multiple dogs barking, exacerbated in buildings lacking acoustic design (Coppola et
al., 2006a; Sales et al., 1997; Schiefele et al., 2012). Ambient shelter noise can reach over 100
dB, which in one study resulted in hearing damage for all dogs exposed for six months (Schiefele
et al., 2012). Diet can have profound effects on physical welfare, often investigated in farm
animals (Manteca et al., 2008) and companion animals within the framework of pet nutrition
(Buff et al., 2014). Because of practicality, shelters often have a standardized feeding formula.
However, there is only one known study directly investigating the effects of diet on shelter dog
cortisol levels in which the authors found that dogs provided with both a premium diet and 20
minutes of supplemental human interaction per day saw a moderate decrease in plasma cortisol
in response to a novel object, suggesting that both human interaction and nutritional
enrichment can moderate HPA axis activity (Hennessy et al., 2002). The moderating effects of
human interaction found by Hennessy and colleagues (2002) has been corroborated by other
studies (Coppola et al., 2006b, Menor-Campos et al., 2011; Shiverdecker et al., 2013), although
the effect is likely context dependent (Hennessy, 2013).
Social isolation, both intra-specific and inter-specific, typical of single-housed kennels
and constricted time with human contact, respectively, can increase stress, particularly if the
dog is isolated from a bonded individual (Coppola et al., 2006b; Walker et al., 2014). In most
cases, shelter life comes as an abrupt change from the dog’s previous living arrangement,
whether that be in a home or as a stray. Novelty in and of itself has been shown to stimulate
physiological stress responses (Ader & Cohen, 1993). Loss of control and predictability,
characteristic of routine disruption and adherence to novel husbandry practices, cause
activation of the HPA axis (for review see Hennessy, 2013). This not only impacts the dog
physiologically, but when stress becomes chronic, also induces a negative affective state leading
to poor welfare (for review see Taylor & Mills, 2007). Additionally, dogs that enter the shelter
experience sudden disruption of relationships and separation from attachment figures (Tuber et
al., 1996). Due to the social nature of the species this too can cause stress, and it has been
shown that the presence of a caregiver can reduce canine stress when in a novel environment
5
(Tuber et al., 1996). Further illustrating the social needs of the species, pair or group housing
has been shown to have a positive impact on behavior and separating bonded dogs can have
negative impacts on immune system indicators (Mertens & Unshelm, 1996; Walker et al., 2014).
Studies have found that solitary housing can exacerbate stereotypies (Beerda et al., 1999a;
Hetts et al., 1992; Mertens & Unshelm, 1996), although differences in behavior descriptions
complicates the ability to compare across studies (for review see Table 1 in Protopopova, 2016).
While shelter-typical stressors have been identified (refer to the introduction of Gunter
et al., 2019 for succinct summary), staff are limited in their ability to mitigate exposure to said
factors. Both available resources and practicality can impede environmental change. Land and
building constraints dictate kennel size, sanitation requires the use of cleaning agents, and dogs
will bark loudly. Additional design consideration requires substantial funding and increasing
social interaction requires labor. While shelter resources vary, it is unrealistic to assume
environmental stressors can be completely eliminated. To target resources effectively,
substantial effort has gone into understanding the effect of environmental factors on the
welfare of shelter dogs.
1.2.ii: Effects of Stress on Welfare
Definitions of animal welfare encompass diverse perspectives on the human-animal
relationship (Carenzi & Verga, 2009). Scientifically, it pertains to biological, psychological, and
natural aspects. Broom, oft referenced, writes “the welfare of an individual is its state as regards
its attempt to cope with its environment†(Broom, 2019). Failure and difficulty to cope with
environmental challenges, whether actual or perceived as threatening, leads to poor welfare
(Broom, 1991). The shelter environment introduces external stressors documented to activate
the HPA axis, discussed in the previous section and reviewed by Hennessy (2013). Exposure to
shelter-typical stressors can interact with the length of stay and lead to dysregulation of the HPA
axis (Hennessy et al., 2006). It is also shown that physiological response to stressors varies
depending on an individual’s genetics and past experiences (Beerda et al., 1999b, Boxal et al.,
2004; Hiby et al., 2006; Rooney et al., 2007). Therefore, dogs react uniquely to equivalent
external cues, as well documented by studies evaluating behavioral and physiological measures
to stressful situations.
6
Since Hennessy’s 2013 review, extant literature has expanded to include investigations
on long-term effects of the shelter environment and validation of stress measurements, with a
focus on differentiating between acute versus chronic stress. Acute stress is indicative of a
normal regulatory response to a physical or perceived threat followed by a return to
homeostasis after the danger has passed. However, as length of stay increases, acute stress has
the potential to become chronic. Chronic stress leads to dysregulation of the HPA axis and
immune suppression (Hennessy, 2013). Yet, studies have failed to irrefutably determine whether
the shelter environment induces chronic stress. This is largely due to the ambiguity of
physiological measures, discussed by Protopopova (2016) in her review of the methodologies
used to quantify shelter dog welfare through measures of physiology, immune function, and
behavior. Prolonged exposure to stressors can lead to a state of chronic stress characterized by
dysregulation of the HPA axis often accompanied by physiological and psychological disorders.
However, it is unclear whether the average shelter dog is experiencing dysregulation or
habituation, which similarly leads to an increase in cortisol followed by a decrease. Cortisol
measures can increase during states of stress or arousal, the latter of which does not suppose
valence (for review on the dysregulation hypothesis and habituation hypothesis, see
Protopopova, 2016). In the instance of dysregulation due to chronic stress, immune function is
also impaired but direct studies of immunosuppression in shelter dogs are complicated by
individual immune differences and high exposure to infectious diseases. Extensive research has
been done on the effects of sheltering on behavior with studies generally indicating differences
in stereotypical and repetitive behaviors, stress and fear behaviors, and activity level. In
conclusion, Protopopova (2016) recommended that future research focus on individual dogs
rather than group averages when evaluating welfare.
As recently as 2020, Hennessy and colleagues reviewed laboratory literature as the
guiding principle in describing the effects of shelter environments on dogs and the formulation
of methods to reduce stress. While the review predominately focused on research regarding
social buffering as stress remediation, the authors also discuss the difficulty in designing
experiments to evaluate behavioral consequences of stress in shelter dogs due to obstacles such
as accommodating shelter protocols, confounding environmental factors, and divergent past
7
experiences that influence individual dogs’ behaviors. Despite the difficulties in researching
shelter dog stress behavior, Hennessy and colleagues (2020) found evidence that psychological
stress as experienced by shelter dogs may compromise sociality and cognition, both considered
desirable traits in adoptable dogs. Additionally, the authors demonstrated how “insidiousâ€
consequences of shelter stress may be masked by developmental patterns that result in
increased susceptibility to mental and physical disorders later in life. The “two-hit†model,
applied in various scientific fields (Feigenson et al., 2014; Gold et al., 1988; van Rooij et al.,
2018) and often first attributed to Knudson (1971), provides a framework for exploring long-
term effects of shelter stress on dogs. In the model proposed by Hennessy and colleagues
(2020), a sensitization of the stress response occurs via exposure to a strong or chronic stressor
(first hit) which leads the individual susceptible to a more profound reaction to a later instance
of stress (second hit) that causes dysregulation of the stress response with mental or physical
consequences. Chronic stress or trauma, as could be experienced by a dog while in shelter,
could be the first hit that leaves the individual susceptible to behavioral consequences post-
adoption.
Despite the challenges of evaluating long-term effects of stress in shelter dogs, the wide
variability in physiological and behavioral response exhibited by shelter dogs indicates that
while some transition through the shelter with minimal long-term consequences, others may
have lasting adverse reactions. Furthermore, the inability to cope with stressors in the shelter
environment reduces welfare state, leading to immediate welfare concerns. Therefore, to
address current shelter dog welfare, it is important to evaluate coping ability at the individual
level.
1.2.iii: Coping Styles
Identifying dogs that struggle to cope in the shelter setting is complicated by the
behavioral variation seen among individuals. Coping has been defined as the behavioral
reaction to situations that activate neuroendocrine pathways involved in physiological response
to aversive situations (Wechsler, 1995). When successful, coping behavior reduces physiological
measures of stress and is considered adaptive. However, if the behavior is ineffective at either
removing the animal from the aversive situation or restoring physiological measures to baseline,
8
an individual is at risk for chronic stress at which point coping strategies can become
maladaptive. Research on stress response in rodents and farm animals supports the adaptive
evolution of two different coping styles characterized by consistent behavioral and physiological
reactions in response to environmental challenges (for review see Koolhaas et al., 1999). The
dichotomous framework of coping styles, first introduced by Henry and Stephens (1977),
synthesized early work on the active fight-flight responses by Cannon (1915) and the
conservation-withdrawal response originally described by Engel and Schmale (1972). However,
it is important to note that response to stress lies along a continuum of active to inactive
physiological and behavioral responses and one individual may rely on a variety of strategies
depending on the environmental context and past experiences. Subsequent scientific discussion
differentiated between the two coping styles as either active or passive, but use of this
terminology masks the agency of individuals that react to aversive stimuli by decreasing activity
(Wechsler, 1995).
More recently, scientists have categorized coping styles as either proactive in lieu of
active or reactive in lieu of passive (Koolhaas et al., 1999). Understanding that coping style can
be flexible dependent on adaptive value in current situations, when behavioral response fails to
eliminate the perception of danger, individuals will tend toward the extreme ends of the
spectrum, resulting in either a learned helplessness (reactive) or excessive activation of
energetic output often recognized as stereotypies or loss of impulse control (proactive). Here
the terms reactive and proactive refer to the coping styles representative of the maladaptive
behavior patterns seen in shelter dogs.
1.2.iv: Maladaptive Coping Behavior
Evolutionarily, if the coping behavior is successful in achieving environmental control,
then fitness of the individual increases and the coping style is adaptive (Wechsler, 1995).
However, if the animal’s behavior fails to either remove the aversive stimuli or return
neuroendocrine systems to homeostasis, then the coping response can have deleterious effects
on fitness rendering the behavior maladaptive. When animals are placed in situations that are
evolutionarily unnatural, such as housing systems that restrict natural behavior, individuals may
lack innate or learned adaptive coping strategies. Several studies have investigated changes in
9
cortisol levels over time as a measure of acclimation to the shelter environment (Coppola et al.,
2006b; Hennessy et al., 1997; Hennessy et al., 1998). While some shelter studies found an
increase of cortisol followed by a decrease before plateauing, others found less defined patterns
(for review see Lamon et al., 2021, section 3.4). The variability found across studies could be
due to a variety of factors such as collection methods, individual variability, and environmental
factors, but it may also suggest that each dog experiences a unique stress response
independent of average trends. For the dogs that are incapable of adjusting to the shelter
environment, coping style may impact which behavior patterns manifest while in-shelter.
Dogs with a predominately reactive coping style would likely demonstrate anxious-
avoidant behaviors and may become despondent or withdrawn. Animals with a reactive coping
style will respond to aversive stimuli with decreased physical activity, commonly demonstrated
in animal studies as longer attack latencies and conditioned immobility along with increased
HPA axis and parasympathetic reactivity; those with a predominately proactive coping style can
progress to excessive-arousal behaviors and stereotypic activity such as pacing in-kennel or
persistent barking. Proactive coping is associated with increased activity, demonstrated by
decreased attack latency, and increased active avoidance along with decreased HPA axis and
parasympathetic reactivity (Benus et al., 1989; De Boer et al., 1990; Hessing et al., 1994; Korte
et al., 1992; Korte et al., 1999; Ruis et al., 2000; Schouten et al., 1997; van Oortmerssen &
Bakker, 1981). For comparison of behavioral and physiological differences between reactive and
proactive coping styles, see review by Koolhaas et al. (1999).
The exact combination and extremity of behaviors is influenced not only by the
individual’s coping style but also genetics and past experiences. Individual coping style may
correlate with personality. Considering the shy-bold continuum, proactive behaviors, such as
increased locomotion, align with bold type personalities characterized in part by exploration
and risk-taking (Sloan Wilson et al., 1994). Shy type personalities, generally less likely to engage
in risk-taking, are frequently identified using behavioral responses associated with reactive
coping styles in studies using novel object or novel environment tests. In dogs, boldness has
been shown to correlate with playfulness, interest in chase, exploratory behavior and sociability
towards strangers while shyness is correlated with avoidance behavior (Svartberg & Forkman,
10
2002). In a study on coping response of police dogs during an acute challenge, the authors
found fearfulness or aggressiveness to be primary personality factors that differentiate between
coping styles (Horvath et al., 2007). Fearfulness, measured by behaviors related to flight or
withdrawal, was predominant in individuals that responded to the threatening approach of a
human with behavior demonstrative of a reactive coping style, the characteristics of which align
with previous findings on shy personalities (Horvath et al., 2007; Sloan Wilson et al., 1994).
Aggressiveness, measured by behaviors such as barking, was predominant in individuals that
demonstrated a proactive coping style, which has also been shown to align with bold
personalities (Horvath et al., 2007; Sloan Wilson et al., 1994). While the authors identified a
third primary coping factor of ambivalence which included behavioral responses that indicate
uncertainty, the described behavior patterns, i.e., paw lifting, mouth licking, and looking away,
have also been categorized elsewhere as displacement behaviors, hypothesized to function as
appeasement signals (Overall, 2017). Displacement activity includes behaviors without an
apparent function and are correlated with situations of psychosocial stress (Maestripieri et al.
1992; Zeigler 1964). Appeasement behaviors are intended to communicate non-aggressive
intent during conflict (Kuhne et al. 2014; Pastore et al. 2011). There is an ongoing debate on
whether dog displacement behaviors do function as appeasement signals (Pedretti et al., 2023);
however, it is possible that by measuring displacement behaviors during testing, the researchers
may be capturing indicators of affective state in lieu of coping response.
Past experiences can also affect individuals’ ability to cope. Hiby and colleagues (2006)
found that cortisol/creatinine ratios (C/C) tended to decrease over time in dogs returned to a
shelter while increasing in those relinquished from homes. Rooney and colleagues (2007) found
that C/C increased in all dogs entering a training establishment but was significantly higher in
dogs that were not previously habituated to kenneling. Therefore, there is evidence that prior
experience with kenneling may mitigate the stress response. Additionally, several studies have
found that early life experiences can have a profound impact on behavior later in life (Appleby
et al., 2002; Boxall et al., 2004). Beerda and colleagues (1999b) found that weather influenced
physiological stress response of beagles entering restricted housing, suggesting that certain
environmental conditions may even impact individual response to kenneling. Because of the
11
complex relationship between coping style, genetics, environmental context, and previous
experience, individual dogs may be better or worse suited for adapting to the shelter
environment.
Dogs experiencing physiological and psychological states of stress will show repeated or
extreme behaviors correlated with proactive or reactive coping, such as hyper-activity or
despondence respectively. Relying on behavior alone as an indicator of stress is complicated
due to considerable individual variability (Part et al., 2014). However, species-specific stress
signals have been well-documented in kenneled dogs (for review see Polgar et al., 2019).
Increased activity or avoidance indicates stress in dogs, both of which are considered
undesirable in-kennel behaviors found to increase length of stay (Protopopova et al., 2014). In a
questionnaire distributed to potential adopters after interactions with shelter dogs, people
chose not to adopt based on perception of high activity level and insufficient attentiveness
while those who did adopt described the dog as calm, friendly, and/or playful (Protopopova &
Wynne, 2014). Adoptable dogs generally viewed as less desirable are often those that are
struggling to cope with environmental stressors and are exhibiting maladaptive coping
behaviors. It may take considerably longer for these dogs to get adopted, prolonging exposure
to an environment the dog finds stressful.
Furthermore, to fully consider the success of adoption, pet retention must also be
considered. While relinquishment is a complex dilemma, surveys that examine factors related to
relinquishment have validated the impact behavior can have on the decision to surrender a pet,
with behavioral problems frequently cited as a determinant (for review see Marston & Bennett,
2003). Additionally, strength of attachment has been shown to correlate with likelihood of
relinquishment (Patronek et al., 1996; Serpell, 1996) and there is some evidence that shelter
dog human-attachment may be characterized by higher levels of anxiety when compared to
pets who were homed as puppies (Previde & Valsecchi, 2007). Long-term behavioral
consequences of exposure to shelter stressors may impact the development of the human-pet
bond post-adoption, creating additional challenges for pet ownership.
To encourage successful adoptions, shelters that have the capacity to implement
behavioral modification plans to support those that exhibit concerning behavior (signals
12
indicative of fear, aggression, or high arousal) have begun to use a holistic approach toward
assessment, including context specific observations (in-kennel, outside, in-building),
supplemented with behavioral and/or medical interventions with the goal of decreasing
undesirable behaviors. Dogs that struggle to cope will continue to perform maladaptive
behaviors and may be overlooked by potential adopters, ultimately decreasing their welfare due
to prolonged stays in the shelter. Behavioral or medical interventions may help these dogs cope
with their environment and ultimately improve their welfare, however, these cases need to be
accurately identified by shelter staff.
1.3: Interventions to Assist with Kennel Coping
1.3.i: Enrichment and Behavioral Modification
Behavioral intervention encompasses strategies aimed at mitigating the effects of
external stressors. In a review on shelter adoption and relinquishment interventions,
Protopopova and Gunter (2017) categorize behavioral intervention strategies under object
enrichment, sensory enrichment, conspecific interaction, or human interaction. While object
enrichment has been shown to influence dog activity, the impact on dog behavior and adopter
choice remains unclear (Protopopova & Gunter, 2017). Sensory enrichment, such as the use of
odors and music, shows potential to alter behavior but more research is needed to understand
the magnitude of effect (Wells, 2009). Conspecific interaction, such as pair-housing, has been a
particular area of interest due to the social nature of the species. Many studies have found that
conspecific contact can reduce abnormal behavior, but confounding factors such as pen size and
personality complicate conclusions. Other forms of social enrichment include increased positive
human interaction (Shiverdecker et al., 2013), including behavior training methods (for review
see Protopopova & Gunter, 2017), to foster behaviors attractive to potential adopters and
indicative of improved welfare. Habituation and desensitization programs aim to decrease
sensitivity to specific stimuli that cause an individual stress.
Popular support for environmental enrichment to improve welfare has caused a
burgeoning collection of scientific literature regarding the benefits and methodology of
encouraging species-specific behaviors and maintenance of the healthy physiological reactions
essential to homeostasis. While many studies have been conducted on environmental
13
enrichment for kennels (for reviews see Taylor & Mills, 2007; Wells, 2004), sample populations
are often restricted by the region served by the shelter, demographic representation of sex,
neuter status, age, breed, and disposition as most participants consist of adoptable dogs
deemed medically healthy and available for adoption. Furthermore, as a safety precaution,
individuals exhibiting aggressive behavior are often excluded from studies. Therefore, there
remains a portion of the shelter dog population that require additional support for behavior
intervention beyond conventional intervention programs.
In instances where behavioral intervention programs are ineffective at decreasing
concerning behaviors, shelter dogs may benefit from pharmaceutical intervention. Anxiolytic
treatment on dogs experiencing anxiety disorder has shown promise (Ballantyne, 2018; Gilbert-
Gregory et al., 2016; Gruen & Sherman, 2008). Anxiolytic medications encompass
pharmaceuticals designed to target neural pathways involved in physiological stress responses.
However, there have been few studies on use of drug therapy specifically on shelter dogs
(Abrams et al., 2020; Corsetti et al., 2021; Tod et al., 2005). This may be due largely to the many
obstacles in providing shelter dogs with anxiolytic medications. For a dog to receive
pharmaceutical treatment, a veterinarian must write a prescription. Shelters vary greatly in
resources and not all have veterinarians on staff. Illustrating the disparity in available resources
and need, through a survey intended to evaluate community programming offered by animal
shelter organizations, Russo and colleagues (2021) found that reasons for discontinuing
community assistance included lack of funding, improper staffing, and lack of space;
inconclusion, collaboration between shelters, veterinarians, and animal behaviorists would
benefit communities by effectively addressing canine behavior problems.
To facilitate collaboration, accurate assessment of individual dogs from shelter staff may
save veterinarians consultation time and decrease professional costs. This would alleviate some
of the financial burden on shelters to provide dogs with anxiolytic medications. Veterinarian
services are partially reliant on third-party observations and clinical appointments include pet-
owner interview for a better understanding of the animal’s clinical needs. This is necessary as
clinicians are unable to directly observe the animal in all situations. However, communication
between veterinarians and caretakers inherently poses a potential for disconnect between
14
actual behavior and perceived meaning. In one study where dog-owners, veterinarians, dog
trainers, and non-owners were asked to describe dog behavior using adjectives, there was little
agreement when classifying behaviors associated with aggression, confidence, and actual-play
(Tami & Gallagher, 2009). While this study only involved nine dogs and 60 observers, other
studies have shown how experience and education may influence interpretation of animal
behavior (Diesel et al., 2008; Fidler et al., 1996). On the contrary, one study found no difference
in ability to interpret behaviors based on experience factors between 47 veterinary students;
however, given that participants were veterinary students, a degree of training and/or
familiarity could be assumed (Meyer et al., 2014). Therefore, to ensure quality standard of care,
a reliable behavior assessment for shelter staff use would provide veterinarians with accurate
information by minimizing observer bias. However, a validated behavior assessment intended to
diagnose coping behavior does not exist. Instead, current behavior assessments used in shelters
are focused on temperament testing or broad indication of welfare within shelter
establishments.
1.4: Assessments
1.4.i: Existing In-shelter Assessments
Temperament assessments are frequently used in shelters by staff, often to assist in
decisions on suitability for adoption. Despite the prevalence of assessments in shelter
management, exact methodology and use varies by shelter (Clay et al., 2020; Marston &
Bennett, 2003; Taylor & Mills, 2006). Evaluating the public safety risk of shelter dogs is an
important responsibility of staff and previous efforts have gone into developing assessments
that rate temperament to assist in decision-making (Luicidi et al., 2005; Mornement et al., 2014;
Mugenda et al., 2019). To assess companion pet suitability, temperament tests focus on the
dog’s response to certain scenarios designed to mimic potential at-home situations. However,
studies evaluating the predictive validity of temperament testing for shelter dogs have found no
correlation between behaviors observed during assessments to post-adoption behavior
(Christensen et al., 2007; Mornement et al., 2014; Mornement et al., 2015), with the
implication that several dogs are falsely identified for aggression and deemed unsuitable for
adoption (Patronek & Bradley, 2016). While some studies have utilized principal component
15
analysis to determine the ability of temperament testing to detect personality factors, e.g.,
fearfulness, friendliness, and aggressiveness with limited success (de Palma et al., 2005;
Dowling-Guyer et al., 2011), in their review on validation of shelter assessments, Patronek and
colleagues (2019) conclude that existing temperament assessments used to screen shelter dogs
for adoption suitability do not meet accepted standards of validity due to confusion between
colloquial and scientific definitions of measurement, limitations in analysis, and the inability to
extrapolate beyond the research subjects to shelter dogs . The discrepancy between in-shelter
and post-adoption behavior within an individual can largely be explained by the effects on
behavior of the shelter environment and unnatural testing scenarios. Despite the inability to
predict behavior, many justify the use of temperament tests as one source of information that
can contribute to a broader behavior profile consisting of multiple observations of the dog
made in several situations. However, it is important to note that studies have failed to show
practical benefits of temperament testing for either the dog or for adopters.
As such, current temperament assessments used in shelters are not suited for evaluating
shelter dog welfare. Beyond temperament tests, there are few shelter-specific Quality-of-Life
(QoL) assessments developed to either assess acclimation to the environment or measure
impact of an intervention program, primarily measuring effect of human interaction but also
used in some instances for measuring the impact of varying housing environments. In a review
on shelter dog QoL assessments, Lamon and colleagues (2021) identified five validated
ethogram-based assessments, one validated interaction test, and one validated behavior test
(Table 1.1). While these assessments provide a way to compare the welfare of dogs across
shelters, results do not indicate individual needs in terms of additional mental support.
Furthermore, several are time-consuming and unpractical for staff to repeatedly administer to
all dogs. Therefore, there remains a need for a practical behavior assessment that can
communicate current coping state and indicate anxiolytic needs.
16
Table 1.1. Existing shelter dog QoL assessments. Validation methods for assessments used for
evaluating shelter dog welfare published between 2000 and 2020 (adapted from Table 1 in
Lamon et al., 2021).
Method
Assessment Tool
Study Focus
Validation
Reference
Ethogram-
Intervention:
Quality of Life (QOL)
Quality of life (QL) score
Kiddie & Collins,
based
Human
Assessment
calculated using positive
20141; Popescu et
interaction1
Intervention:
Socialization
program2
and negative indicator
al, 20182
behaviors validated against
treatment groups
Ethogram-
Acclimation
Shelter Quality
Management, resource, and
Barnard et al.,
based
Protocol (SQP)
animal-based measures
2016; Arena et al.,
used to identify welfare
2019a
hazards; individual
measures validated
observing individual twice
(at a distance and again
close-up)
Ethogram-
Acclimation
Shelter Quality
Refined version of SQP
Berteselli et al.,
based
Protocol 2 (SQP 2)
validated against climatic
2019
conditions
Ethogram-
Acclimation3
Qualitative
Descriptive terms analyzed
Arena et al, 20173;
based
Intervention:
Behavioral
for consensus dimensions
Arena et al., 2019b;
Different housing
Assessment (QBA)
to describe emotional state
Berteselli et al.,
environments and
length of stay4
of shelter dogs
20193; Walker et
al., 20164
Ethogram-
Acclimation
Multi-operator
Stress level score based on
Menchetti et al.,
based
Qualitative
Behavioral
brief description of the
2019
dog’s behavior; stress level
Assessment
score did not correlate
between raters, suggesting
the score is situation
dependent and not a
comprehensive indication
of welfare
17
Table 1.1 (cont’d)
Interaction
Intervention:
Approach test
Approach behavior
Arhant &Troxler,
test
Human
interaction
categorized as “contact
2014
possible†or “no contact
possible†validated against
attitude of shelter staff via
questionnaire
Behavior test
Intervention:
Behavior test
Test is not described in
Exercise and
human contact
study and previously
validated by authors (not
published)
Menor-Campos et
al., 2011
1.4.ii: Development of a Novel Coping Behavior Assessment
Due to the wide variation of dogs’ ability to adapt to the shelter environment, not all
dogs successfully acclimate, resulting in the manifestation of maladaptive coping behaviors that
are indicative of a negative affective state and contributory to decreased desirability to potential
adopters. When behavioral intervention programs focused on environmental and social
enrichment fail to decrease concerning behaviors, anxiolytic pharmaceuticals may benefit
current welfare of shelter dogs. However, not all shelters have access to anxiolytics due to the
requirement of veterinarian prescription. Future efforts to develop a reliable and validated
assessment for use by shelter staff to quantify coping behavior and communicate meaningful
information to veterinarians would facilitate communication between shelters and clinics. The
assessment should not only differentiate between dogs in need of medical intervention and
those not in need, but also allow veterinarians to make informed decisions on which
medications to prescribe by indicating whether the maladaptive coping behaviors are
manifesting as anxious-avoidance or excessive-arousal. By resulting in a total coping score, the
assessment could also be useful in monitoring efficacy of behavioral interventions through
ranked outcomes. Not only would this provide a standardized method to provide individualized
care to shelter dogs, but also further use of assessment outcomes could result in a national
database of shelter dog behavior available for studies on welfare.
18
CHAPTER 2: DEVELOPMENT OF A RELIABLE BEHAVIOR ASSESSMENT TO MEASURE COPING
BEHAVIOR OF SHELTER DOGS
2.1: Introduction
2.1.i: Stress in the Shelter Environment
Dogs that enter animal shelters experience a variety of well-documented environmental
stressors such as restricted space, novelty, sensory overload, and limited social interaction
(Hennessy, 2013; Marston & Bennett, 2003; Protopopova, 2016). Prolonged exposure to
stressors can result in states of chronic stress, in turn leading to dysregulation of the
hypothalamic-pituitary-adrenal (HPA) axis (Protopopova, 2016). Persistent activation of the HPA
axis is considered maladaptive and can lead to stress associated behavior changes and
compromised health through immunosuppression (Hennessy, 2013). Individuals respond
differently to environmental stressors based on genetics, past life experiences, and current
motivational state and health (for review see Marston & Bennett, 2003), but those that fail to
adjust to the shelter environment are at risk of developing maladaptive coping behaviors due to
chronic stress (Corson & O’Leary Corson, 1976; Hubrecht et al., 2017; Koolhaas et al., 1999).
Dogs that develop maladaptive coping may either exhibit anxious-avoidant behaviors
such as despondency, reflective of a proactive coping style, or excessive-aroused behaviors such
as lack of impulse control, correlated with reactive coping styles (Koolhaas et al., 1999). These
behaviors have been shown to decrease desirability to potential adopters, increase length of
stay, and indicate negative affective state, thus impacting current welfare as well as the
likelihood of exiting the stressful environment (Cohen & Todd, 2019; Protopopova et al., 2014).
2.1.ii: Indicators of coping ability and welfare state
The measurement of stress in animals is a crucial aspect of understanding their welfare
state and ability to cope with environmental stressors. Physiological markers of stress such as
increased cortisol levels and decreased heart rate variability may provide objective
measurements of the neuroendocrine stress response. While these measures have been
validated as indicators of physiological arousal, they do not indicate valence (Polgar et al., 2019).
Further complicating interpretation, collection methods can impact plasma cortisol levels by
acting as an acute stressor and baseline hormone levels vary among individuals. Beerda and
19
colleagues (1997) propose the use of ethological observations and behavioral assessments to
complement physiological measurements, as they provide valuable insights into an animal’s
subjective experience of stress. While dual collection of physiological and behavioral measures
may be plausible for experimental designs, application of physiological measures in shelters
would be impractical. Due to resource constraints, shelter staff evaluate the coping ability of
individual dogs primarily through behavioral observation. However, training amongst shelter
staff is highly variable and there is no standardization across shelters.
Several ethograms have been developed for measuring behavioral indicators of stress in
dogs (for reviews see Kartashova et al., 2021; Protopopova, 2016) with many authors agreeing
that stress behaviors include paw-lifting, self-grooming, lip-licking, panting, vocalizing,
scratching self, digging, yawning, trembling, body shake-off, averting gaze, and sniffing (Barnard
et al., 2018; Bauer et al., 2017; Gilbert-Gregory et al., 2016; Morris et al., 2020). However,
several of these behaviors manifest in diverse circumstances and can serve many functions
(Protopopova, 2016). Certain behaviors correlated with stress may also serve as appeasement
signals (for review see Pedretti et al., 2023). Additionally, there is some discrepancy between
labeling behaviors as indicative of stress or of fear (Barnard et al., 2018; Bauer et al., 2017;
Gilbert-Gregory et al., 2016). While stress is often synonymous with physiological arousal which
by itself does not reveal valence, fear is arguably negative (Jones & Boissy, 2011). Many authors
choose to differentiate stress induced physiological arousal as distress when accompanied with
a negative affective state or eustress when in positive affective state. Within this paper, stress is
used to connotate negative valence unless otherwise specified. Similar to stress, fear activates
the HPA-axis and is correlated with behaviors observed in maladaptive coping shelter dogs, e.g.,
trembling, cowering, wall bouncing, bar pawing, high tail position, lunging, biting, or non-
responsiveness (Barnard et al., 2018; Bauer et al., 2017; Gilbert-Gregory et al., 2016; Morris et
al., 2020; Tod et al., 2005). Other canine behavioral responses seen in shelters, e.g., crouching,
raised hackles, hiding, and growling, are considered freeze, flight, or fight behaviors, labels
traditionally used to categorize behavioral responses to threats (Bauer et al., 2017; Lindsay,
2005). Additionally, stereotypic and abnormal behaviors are frequently measured as indicative
of negative affective state despite a lack of definition and standardized methods across shelter
20
studies (Protopopova, 2016). However, several authors agree that circling, spinning, pacing, tail
chasing, wall bouncing, jumping, and continuous vocalizations are correlated with kennel stress
(Barnard et al, 2018; Bauer et al., 2017; Gilbert-Gregory et al., 2016). Considered in isolation,
meaning of certain behavior can be ambiguous but taken in context, correlation to affective
state is revealed. Therefore, it is important to consider the combination of multiple behaviors
and environmental context when interpreting behaviors associated with stress and fear.
To understand maladaptive coping behaviors, it is also essential to identify behaviors
indicative of positive affective state. Body posture indicative of good welfare includes relaxed
ears and tail with a loose to neutral body posture (Barnard et al., 2018; Morris et al., 2020).
Social, exploratory, and play behaviors have been used as good welfare indicators and can
indicate healthy adjustment to the shelter environment. Social behaviors, such as affiliative
behavior, solicitation of attention, willingness to approach people, and facing toward the front
of the kennel indicate positive affective state (Bauer et al., 2017) and are preferred by potential
adopters (Protopopova & Wynne, 2014; Wells & Hepper, 2000). Exploratory behavior includes
sniffing and engagement with an object by licking, nosing, pawing, chewing, and/or carrying
(Ley et al., 2007; Tod et al., 2005). Object play includes the interaction with toys by grabbing or
holding an object in the mouth (Barnard et al., 2018; Pullen et al., 2010).
2.1.iii: Assessment Development
A novel behavioral assessment must be both reliable and valid to be scientifically
supported (Harvey, 2021). Reliability refers to the proportion of measurement error or degree
of consistency across raters and time, while validity refers to the ability of the assessment to
measure the intended outcome or degree of applicability (McCall, 1984; Taylor & Mills, 2006).
To report reliability, inter- and intra-rater agreement is analyzed through the appropriate
statistical tests. Kappa values, which report on percent agreement corrected for chance, are
often the accepted standards for rater reliability. Inter-rater agreement requires a minimum of
two different raters assessing equivalent scenarios. This is to provide transparency when results
could be subject to reporting bias. Several research ethicists have documented observer effects
prevalent in subjective scoring systems (Burghardt et al., 2012). The degree to which
expectation bias affects observers has varied across studies (Marsh & Hanlon 2007; Tuyttens et
21
al., 2014), but the prevalence of observer preconceptions, vested interest, and ambiguity within
subjectivity, predisposes animal behavior research to interpretation errors (Tuyttens et al.,
2014). Anthropomorphism, which is the attribution of human mental characteristics to non-
human individuals, is particularly salient in the interpretation of companion animal behavior
(Serpell, 2019). Many behavior assessments attempt to mitigate the effects of observer bias by
relying on the establishment of categorical variables, such as ethograms, or by quantifying
results through creation of ranked scales, e.g., visual analog scales (for review see Kartashova et
al., 2021).
Consistency of repeat measures is established through intra-rater agreement, which can
be interpreted as the degree of which individuals are subject to observer drift. In ethological
studies, observer drift refers to the phenomena of observer performance variability over time
(McCall, 1984). For behavioral assessments, reporting on intra-rater reliability is accomplished
through repeat measures using video of the original observation interval after sufficient time
has passed. Inherent to human observation, bias and drift can lead to inter- and intra-rater
disagreement, respectively; however, acceptable standards for percent agreement function to
inform users of the magnitude of estimated measurement error within the assessment.
An assessment that measures coping behavior can further be considered a diagnostic
test, in which the results are applied to categorize dogs as either adaptive coping or
maladaptive coping. Content validity, the adequacy in sampling intended behavior (McCall,
1984), can be optimized by targeting species-specific stress behaviors and is critical during tool
development. To establish criterion validity, a reference standard, i.e., gold-standard, is required
for confirmatory diagnostics (Akobeng, 2007a; Akobeng, 2007b; Akobeng, 2007c; Patronek &
Bradley, 2016). Current accepted diagnostics for identifying dogs that could benefit from
anxiolytic treatment is veterinarian evaluation consisting of medical and behavioral review with
direct observation (Stelow, 2018). Criterion validity will be analyzed in Chapter 3.
2.1.iv: Establishing Content Validity: Existing Behavioral Tests
While a comprehensive evaluation of coping behavior in shelter dogs is not currently
available, behavioral tests have been used to measure welfare components by using behavioral
response as evaluation of affective state (Titulaer et al., 2013). Validated behavior tests and
22
welfare indicators can serve as a basis for content validity when developing a novel coping
behavior assessment. Behavior assessments should be standardized to minimize variability
between assessments and be feasible for most shelters to implement (Diederich & Giffroy,
2006; Jones & Gosling, 2005; Taylor & Mills, 2006), with the goal of providing a quantifiable
basis of behavior (Marston & Bennet, 2003). Human-approach, startle, and model dog tests
have been used to assess behavioral response of dogs to certain stimuli. Other behavioral
responses used as welfare indicators include play behavior, activity level, and eating behavior.
Human-approach tests evaluate fear of humans in dogs and have been used as a welfare
indicator in kenneling facilities (Bauer et al., 2017). Approaching the shelter dog in a non-
threatening way has been validated as an indirect measure of affect towards humans (Arhant &
Troxler, 2014) and found to have high inter- and intra-rater agreement among novice raters
(Mugenda et al., 2019). The human approach test formalizes a typical interaction between staff
and dogs, standardizing the protocol for observing presentation in-kennel.
Startle tests involve sudden exposure to visual or acoustic cues to observe the dog’s
reaction to unexpected stimuli. Variations include the opening of an umbrella (Bray et al., 2017;
Goddard & Beilharz, 1985; King et al., 2003; Netto & Planta, 1997; Sherman et al., 2015), the
movement of a remote control car (Goddard & Beilharz, 1985; Haverbeke et al., 2008; Sherman
et al., 2015), gunfire or air blast (Murphy, 1998; Haverbeke et al., 2008; Sherman et al., 2015),
loud noise made by metal objects (Foyer et al., 2016; Goddard & Beilharz, 1985; Sherman et al.,
2015; Wilsson & Sinn, 2012), and the sudden dropping or raising of an object (Foyer et al., 2016;
Wilsson & Sinn, 2012; Sherman et al., 2015). A startling stimulus triggers an acute stress
response, of which the behavioral consequences can provide insight into affective state. If the
brain perceives the stimuli as non-threatening, rapid return to system balance occurs. Initial
response may correlate with physiological state; dogs that are in a state of chronic stress may
respond more strongly to the stimulus or take longer to recover, as can be measured by initial
behavioral response and latency to resume baseline activity. Currently, it is routine for clinicians
to rely on visual assessment of behavior under acute stress, such as exposure to a startling
stimulus, when developing behavioral intervention strategies for individual dogs (for review see
Kartashova et al., 2021). However, while commonly used for temperament testing in shelters,
23
concerns exist about the content validity (for reviews see Patronek & Bradley, 2016; Taylor &
Mills, 2006). Several versions of the startle test have been shown to invoke fear, which may be a
context dependent response irrespective of welfare state, whereas a stimulus that invokes a
startle reaction below the threshold of fear may serve as a measure of coping ability.
Many shelters utilize a version of a ‘Model Dog’ test in which a fake dog is introduced to
the shelter dog to assess behavior toward conspecifics (Bennett et al., 2012; Assess-A-Petâ„¢
Bollen & Horowitz, 2008; Match-Up II Shelter Dog Rehoming Programâ„¢ Dowling-Guyer et al.,
2011; SAFER® Aggression Assessment Weiss, 2007; for review see Taylor & Mills, 2006). While
some disagreement exists over whether a model dog accurately elicits either affiliative or
aggressive behavior towards conspecifics, general census is favorable that behavior toward a
model dog correlates with conspecifics (Barnard et al., 2012; Goddard & Beilharz, 1984; Reid &
Collins, 2012; Shabelansky et al., 2015). In the context of a shelter, the model dog offers a way
to simulate encounters with unfamiliar conspecifics without exposing individuals to a potential
safety risk.
Play behavior in mammals has been considered an indicator of positive welfare
(Dawkins, 1998) primarily because it is self-rewarding and associated with health (Burghardt,
2005; Fraser & Duncan, 1998), although frequency and duration of play behavior is influenced
by a variety of factors, e.g., species, age, genetics, and environment (Martin & Caro, 1985; Rezac
et al., 2011; Smaldino et al., 2019; Sundman, et al., 2016; Svartberg et al., 2005). Evidence
suggests that play indicates a positive affective state (Bateson, 2014) by functioning to reinforce
social cohesion during initiation (Horowitz, 2009) and cessation, e.g., return to restful state
(Prato-Previde et al., 2003), while a lack of play may be demonstrative of a reduced behavioral
repertoire seen in instances of learned helplessness brought on by chronic stress (Maier, 1984).
Despite some debate on the reliability of play as a welfare indicator (Sommerville et al., 2017),
observing behavior at the beginning and end of dog-human play can provide insight into coping
behavior in shelter dogs while also contributing to improved welfare, as affiliative play between
humans and dogs has been shown to reduce stress (Horvath et al., 2008).
Current methods of assessing behavior, e.g., human approach and model dog, capture a
small snapshot of welfare state and do not quantify coping behavior in a clinically meaningful
24
way without further context. Identification of dogs in need of anxiolytic prescription is often
initiated by shelter staff despite considerable variation in training and experience. Furthermore,
shelters vary considerably in available resources, and many do not have access to veterinary
specialists in behavior (Russo et al., 2021). A reliable tool for staff that accurately communicates
shelter dogs’ behavioral needs to veterinary staff would facilitate communication and
partnerships between caregivers and clinicians, thereby providing the sheltering community a
means to better support shelter dog welfare.
2.2: Materials and Methods
2.2.i: Protocol Development
The novel assessment protocol and accompanying form were developed based on the
expertise of a board-certified veterinarian behaviorist and doctorate in ethology, extant
literature on species-specific stress behavior, existing validated behavior tests, and conversation
with shelter staff regarding practicality. From the humane society staff, typical human-dog
interactions that influence staff’s informal assessment of the dog were considered for ease of
implementation of the protocol to shelter routine. Human-animal interactions (HAI’s) were
designed to allow observation of behavior that indicates the dog’s coping state based on
previously validated behavioral tests, species-specific stress behavior, and expert discussion.
From January-June 2022, the assessment went through several iterations, i.e., 11 drafts
(see Appendix A), during which time the protocol was piloted at Capital Area Humane Society
(CAHS) in Lansing, Michigan on single-housed, healthy dogs aged 12 weeks and older (n = 107).
During the iteration process, components, i.e., prompts, within the assessment form were
revised, added, or eliminated to better reflect the range of coping behaviors exhibited by shelter
dogs (Figure 2.1). A focal group of pilot subjects (n = 17) were selected to represent a range of
coping ability and the assessment videos were reviewed by a board-certified veterinarian
behaviorist, doctorate in ethology, and MS candidate in animal behavior and welfare (the thesis
author). Videos were used to evaluate the clinical relevance of human-animal interactions and
accuracy of the assessment forms. Of the focal dogs, two subjects were chosen for in-depth
review based on severity of behaviors indicating either excessive-arousal (subject named Diesel)
or anxious-avoidance (subject named Tiger). Diesel consistently exhibited problematic behaviors
25
such as humping, hard mouthing, and difficulty disengaging across repeat assessments using
assessment form drafts 2-3 (see Appendix A.3 and A.4). Tiger was assessed once using
assessment form draft 10 (see Appendix A.10) and demonstrated reluctance to leave the
kennel, tense body posture, and avoidance throughout the assessment. The behaviors of Diesel
and Tiger partially informed prompt revisions to portray biologically relevant responses more
accurately, e.g., disengagement in play, activity during and after treat consumption, and
exit/return to kennel.
Figure 2.1. Decision tree for protocol development. Components of the assessment were
reviewed for reliability and clinical relevance during the iterative revision process.
Does the prompt have
acceptable inter-
observer reliability?
Yes
No
Revise
prompt
Is there clinical
relevance?
Yes
No
Prompt
remains
Yes
No
Prompt is
removed
The 11th iteration was approved for data collection (see Appendix A.11) and consisted of
eight HAI’s (Table 2.1.a). The HAI’s, explained in more detail during methods, included a human-
approach and startle tests (acoustic and visual) that were modified to simulate typical
environmental stimuli in shelters, namely a friendly human approach and exposure to a startling
stimulus intended to provoke a response below the threshold of fear. Furthermore, a play and
26
treat component served the dual purpose of social, object, and food enrichment while
observing welfare indicators, i.e., play behavior, eating behavior, and return to base activity
state. A variation of the model dog test was designed to mimic a typical shelter scenario, while
additionally simulating conspecific socialization to provide a potentially enriching experience. In
a reverse of the approach test, the manner of return to kennel was included to observe
behavioral response and a final food enrichment was provided in-kennel. Earlier versions of the
assessment were more extensive, including the additional elements, leash-walk (drafts 1-6),
treat dispersal (drafts 1-10), and anxiety score (see Appendix B.1) during acclimation (draft 9);
but upon evaluation for reliability and clinical relevance, they were deemed unnecessary and
therefore removed for practicality and feasibility of the tool. The approved draft had a total of
22 multiple-choice prompts across eight HAI’s. Fifteen of the prompts were selected to
contribute to the coping score, sc01-sc15 (Table 2.1.b). The remaining seven prompts (usc01-
usc07), eight instances of piloerection (pilo1-pilo8), and the single open-ended question
observed during HAI 6 (supp.01), did not contribute to coping score (hence described as
“unscored promptsâ€) but were collected for supplemental information on the individual’s
behavioral profile (Table 2.1.c).
To develop the scoring rubric (see Appendix D), each response for sc01-sc15 was
assigned a value from –3 to +3 by a board-certified veterinarian behaviorist. Positive values
were assigned to responses representative of behaviors associated with excessive-arousal,
negative values for behaviors associated with anxious-avoidance, and zero for behaviors
associated with adaptive coping. Responses that were coded as either positive or negative were
further ranked by severity. Maladaptive coping behaviors considered mild were valued as ±1,
moderate as ±2 and severe as ±3. Since not all prompts had equal number of response options,
a multiplication factor was applied to balance the minimum and maximum contribution to total
coping score among all 15 scored prompts, resulting in a total coping score range of –45 to +45.
27
Table 2.1. Abbreviations for assessment components. Abbreviations used for assessment
components. (a) The assessment consists of eight Human-animal interactions (HAI’s) which are
comprised of either scored prompts and/or unscored prompts. (b) Scored prompts (sc01-sc15)
consist of mutually exclusive multiple-choice questions. Each response abbreviated using letters
of the alphabet (e.g., a, b, c), contribute to the total coping score the amount designated by the
corresponding score value. (c) Unscored prompts (usc01-usc07, pilo1-pilo8, and supp.01)
consist of mutually exclusive multiple-choice questions (usc01 and usc03), nonexclusive
multiple-choice questions (usc02, usc04-usc07), binary present/absent (pilo1-pilo8) and open-
ended written responses (supp.01). Written responses (supp.01) were not analyzed. Each
response abbreviated using letters of the alphabet (e.g., a, b, c), were coded as present/absent
for reliability analysis. Image responses for ‘Overall Body Language’ (usc02 and usc04) were
grouped by association and responses for individual dogs were further analyzed for reliability
when groups consisted of more than one image (Groups 1-4 and Group 8).
(a) Components of the eight human-animal interactions (HAI’s)
Abbrev.
HAI
Scored Prompts
Unscored Prompts
HAI 1
HAI 2
HAI 3
HAI 4
HAI 5
HAI 6
HAI 7
HAI 8
In-kennel: Begin
sc01, sc02, sc03, sc04
usc01, usc02, usc03, pilo1
Play
Settle
Acoustic Startle
Visual Startle
Distraction
sc05, sc06, sc07
sc08, sc09
sc10
sc11
sc12, sc13
In-kennel: Return
sc14, sc15
Activity
usc04, pilo2
usc05, pilo3
pilo4
pilo5
supp.01, pilo6
pilo7
usc06, usc07, pilo8
(b) Scored prompts
Abbrev.
Scored Prompt
Response
sc01
Presentation with closed door
sc01(a)
sc01(b)
sc01(c)
sc01(d)
sc01(e)
sc01(f)
Friendly/attentive/neutral
Dissociated (fearful/stressed
Aware (fearful/stressed)
Hyper-active
Aggression
Hidden from view
Score Value
0
-3
-1.5
1.5
2
null
sc02
Ability to take treat
If sc01 =0 pts1, <0 pts2, >0 pts3
sc02(a)
sc02(b)
Consumes treat from hand
Takes treat from hand but does not consume
01, 02, 03
01, -12, 13
28
Table 2.1 (cont’d)
sc02(c)
sc02(d)
Consumes treat from floor
No interest in treat
sc03
Presentation with open door
sc03(a)
sc03(b)
sc03(c)
sc03(d)
sc03(e)
sc03(f)
Friendly/attentive/neutral
Dissociated (fearful/stressed
Aware (fearful/stressed)
Hyper-active
Aggression
Hidden from view
sc04
Latency to exit
sc04(a)
sc04(b)
sc04(c)
sc04(d)
sc04(e)
Bolts
Efficient
Delayed
Refuses to exit
Unsafe to allow exit
sc05
Initial reaction to play
sc05(a)
sc05(b)
sc05(c)
Engages in play
Approach - does not engage
No approach - does not engage
sc06
Ease of putting away toys
sc06(a)
sc06(b)
sc06(c)
Easy - not engaged with toy
Easy with trade for treat
Difficult - multiple trades required
sc07
Was play ended early?
sc07(a)
sc07(b)
sc07(c)
Yes - over-stimulated
Yes - fearful
No
sc08
Response to treat
sc08(a)
sc08(b)
sc08(c)
Consumes while standing or moving
Consumes while lying/sitting or plays with treat
Ignores treat/brief interest or holds treat in mouth
without consuming
29
01, -22, 23
01, -32, 33
0
-3
-1.5
1.5
2
null
3
0
-3
null
null
0
3
-3
-3
0
3
3
-3
0
3
0
-3
Table 2.1 (cont’d)
sc09
Activity (last 15 sec.)
sc09(a)
sc09(b)
sc09(c)
sc09(d)
Neutral - stationary
Neutral - active
Stressed - stationary
Stressed - active
sc10
Reaction to acoustic startle
sc10(a)
sc10(b)
sc10(c)
sc10(d)
sc10(e)
sc10(f)
sc10(g)
Retreat
Immediate approach
Freeze - upright
Freeze - cower
Flinches/startles
Stops briefly/orients
No reaction
sc11
Reaction to visual startle
sc11(a)
sc11(b)
sc11(c)
sc11(d)
sc11(e)
sc11(f)
sc11(g)
Retreat
Immediate approach
Freeze - upright
Freeze - cower
Flinches/startles
Stops briefly/orients
No reaction
sc12
Overall body language while stimulus is moving*
sc12(a)
sc12(b)
sc12(c)
sc12(d)
sc12(e)
Image A
Image B
Image C
Image D
Image E
30
0
0
-3
3
-3
1.5
3
-1.5
0
0
0
-3
1.5
3
-1.5
0
0
0
-2
-1
0
0
1
Table 2.1 (cont’d)
sc12(f)
sc12(g)
Image F
Image G
sc13
Can you obtain the dog’s attention?
sc13(a)
sc13(b)
sc13(c)
sc13(d)
No - oriented toward stimulus
No - retreat from stimulus
No - uninterested
Yes
sc14
Approach to kennel entry
sc14(a)
sc14(b)
sc14(c)
sc14(d)
sc14(e)
Actively pulling towards entry
Actively pulling away from entry
Actively pulling/darting in multiple directions
Requires encouragement
Cooperative
sc15
In-kennel behavior post-return
sc15(a)
sc15(b)
sc15(c)
Hyper-active
Fearful
Calm
-3
3
3
-3
0
0
-3
1.5
3
0
0
3
-3
0
*Images adapted from “Doggie Language†by Lili Chen (doggiedrawings.net/freeposters)
(c) Unscored Prompts
Abbrev.
Unscored Prompt
Characteristics
Response
usc01
Location after approach
Multiple-choice, mutually exclusive
usc01(a)
usc01(b)
usc01(c)
usc01(d)
Front
Back
Middle
Not stationary
usc02, usc04 Overall body language*
Multiple-choice, nonexclusive
usc02(a),
usc04(a)
usc02(b),
usc04(b)
Image 1
Image 2
Group 1 (Images 1-2)
31
Table 2.1 (cont’d)
usc02(c),
usc04(c)
usc02(d),
usc04(d)
usc02(e),
usc04(e)
usc02(f),
usc04(f)
usc02(g),
usc04(g)
usc02(h),
usc04(h)
usc02(i),
usc04(i)
usc02(j),
usc04(j)
usc02(k),
usc04(k)
usc02(l),
usc04(l)
usc02(m),
usc04(m)
usc02(n),
usc04(n)
usc02(o),
usc04(o)
usc02(p),
usc04(p)
usc02(q),
usc04(q)
usc02(r),
usc04(r)
Image 3
Image 4
Image 5
Image 6
Image 7
Image 8
Image 9
Image 10
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Group 2 (Images 3-6)
Group 3 (Images 7-10)
Group 4 (Images 11-13)
Group 5 (Image 14)
Group 6 (Image 15)
Group 7 (Image 16)
Group 8 (Images 17-20)
32
Table 2.1 (cont’d)
usc02(s),
usc04(s)
usc02(t),
usc04(t)
Image 19
Image 20
usc03
Location with open door
Multiple-choice, mutually exclusive
usc03(a)
usc03(b)
usc03(c)
usc03(d)
Front
Back
Middle
Not stationary
usc05
Misc. behavior
Multiple-choice, nonexclusive
usc05(a)
usc05(b)
usc05(c)
usc05(d)
usc05(e)
usc05(f)
Pawing at exit
Pacing
Whining
Barking
Other
None
usc06
In-kennel misc. behavior
Multiple-choice, nonexclusive
usc06(a)
usc06(b)
usc06(c)
usc06(d)
usc06(e)
usc06(f)
usc06(g)
usc06(h)
Spinning
Excessive jumping
Whining
Excessive barking
Pacing
Frantic pawing at door
Smeared feces in kennel
Other
usc07
Out-of-kennel misc. behavior
Multiple-choice, nonexclusive
usc07(a)
Dog directed reactivity (e.g.,
barrier aggression)
usc07(b)
Other
pilo1
pilo2
pilo3
Piloerection
Piloerection
Piloerection
Present/absent
Present/absent
Present/absent
33
Table 2.1 (cont’d)
pilo4
pilo5
pilo6
pilo7
pilo8
Piloerection
Piloerection
Piloerection
Piloerection
Piloerection
Present/absent
Present/absent
Present/absent
Present/absent
Present/absent
supp.01
Note behavior after dog is let off-leash
Open-ended written response
*Images adapted from “Doggie Language†by Lili Chen (doggiedrawings.net/freeposters)
During development, kennel cards (see Appendix B.2) were attached to the staff side of
the kennels and filled out twice daily by staff until the dog left the shelter for dogs assessed
using assessment form drafts 4-6 (n = 20). Kennel cards collected information on the dog’s
appetite, elimination, and in-kennel behavior. However, since shelter medical and behavioral
records were collected for the dogs assessed for reliability and validity testing, the use of kennel
cards was eliminated.
2.2.ii: Sample Population
From June 2022 through October 2022, all healthy dogs 12 weeks or older maintained in
solitary housing at Capital Area Humane Society (Lansing, MI) overnight for at least one night
prior to assessment were eligible for assessment. Of total intake (n = 344), 98 dogs were
assessed but seven were excluded before analysis due to health reasons (n = 1) and refusal to
exit kennel (n = 6) resulting in 91 assessments. Dogs were not eligible for assessment if they
were previously assessed during the protocol development (n = 5), group housed and 12 weeks
or older (n = 60), less than 12 weeks of age (n = 72), had medical difficulties (broken bones or
recent amputations n = 4, malignant mammary cancer n = 1), received for owner-requested
euthanasia services (n = 36), or were designated staff only handling by the shelter due to
extreme aggression, i.e. human safety risk (n = 4). Dogs that were initially considered for
inclusion but were no longer at the shelter at time of assessment because they were adopted,
transferred, went into foster, reclaimed by owner, or euthanized for medical reasons (n = 64)
were not assessed.
34
Per shelter policy, dogs without known breed history were designated “mixed breedâ€. As
such, there were 85 mixed breed dogs, one German Shepherd, one Siberian Husky, one Golden
Retriever, one Bichon Friese/mix, one French Bulldog, and one American Cocker Spaniel. The
mean age was 2.32 ± 2.41 years and ranged from 0.275 (0y 3m 9d) to 10.753 (10y 9m 1d) years;
however, it should be noted that age was often estimated by shelter clinicians. The mean weight
was 42.07 ± 18.08 lbs., ranging from 6.58 lbs. to 90 lbs., but several dogs (n = 30) were less than
12 months old, and weight may not be representative of adult size. Of total included dogs (n =
91), 44 were female (33 spayed, 11 intact) and 47 were male (29 neutered, 18 intact). Dogs that
were intact on intake were spayed/neutered according to shelter protocol. The shelter
performed all surgeries on select days of the week, therefore if a dog was eligible for
assessment prior to surgery, they were enrolled and assessed regardless of reproductive status.
Dogs that had received surgery on the day of designated assessment were not eligible until the
following assessment day.
2.2.iii: Shelter Housing and Care
Dogs were housed in kennels located in four different rooms in the shelter; kennel sizes
were either 4’-0†x 8’-0†or 4’-0†x 12’-0â€. All kennels were guillotine style with opaque panels
between dogs, an open metal door at the front, and an opaque door at the back of the kennels.
The guillotine door remained open with the exception of kennel cleaning. Items in the kennel
included a Kuranda bed, blankets, and toys. Fresh water was provided ad libitum and dogs were
fed per shelter standards twice a day. Dogs were exercised four times daily in an outdoor
fenced-in yard (lasting approximately five minutes each) and once with a leash walk around the
perimeter of the shelter.
2.2.iv: Protocol
Behavior assessments took place Monday-Friday between 1:00 pm and 5:00 pm to avoid
meal-time disruption. The exact days of the week depended on the available dogs, assessors’
schedules, and shelter events. Eligible dogs were determined through review of intake reports
based on age and arrival date. After enrollment, assessment order was randomized, and
subsequent enrollments were added to the end of the list in a randomized order. A maximum of
five dogs were assessed daily. If a dog was unavailable at the time of assessment, they were
35
moved to the bottom of the daily order. If they were unavailable for the entire shift, they were
moved to the following assessment day. Assessments either took place in an indoor room (594
S.F.) or an outdoor fenced area (545 S.F.). Location was based on room availability and weather
but kept consistent through the day and dogs were assigned to location on a rolling basis
according to the randomized enrollment order. Of the 91 dogs included for analysis, 43 were
assessed indoors and 48 outdoors.
Assessments occurred a minimum 20 hours after intake and were recorded using GoPro
Hero7 Black video recording devices from two perspectives. Assessment forms were completed
in real-time by two raters for inter-rater reliability. Assessment video was viewed a minimum of
4 months post-assessment for intra-rater reliability by one rater. Rater A, the first author, scored
all dogs and administered the behavior assessments. Depending on the schedule of the second
rater, Rater B, C, or E observed the assessments in-field and scored the dogs blinded from Rater
A’s scoring. This resulted in three rater pairs; all raters were female. Raters were trained using
ethograms and a written in-field protocol (see Appendix C). Except for the assessor (Rater A),
raters were asked not to interact with the dog during testing by remaining at a distance and
avoiding eye contact. Raters were also trained with practice assessments on dogs that were
ineligible for enrollment due to previous assessment during the pilot. Practice assessments
were carried out in-field and by watching video of pilot assessments. During training,
assessment forms were compared between the two raters and disagreements discussed using
the ethogram until both raters agreed on assessment response.
Medical and behavioral reports of the dogs while in-shelter were collected once the dog
had left the shelter via adoption (n = 79), foster (n = 4), return to owner (n = 2), or euthanasia (n
= 6). Shelter records were used for demographic information of participating dogs, discussed in
Chapter 3.
2.2.v: The Behavior Assessment
The coping behavior assessment consisted of the following human-animal interactions:
In-Kennel: Begin (HAI 1), Play (HAI 2), Settle (HAI 3), Acoustic Startle (HAI 4), Visual Startle (HAI
5), Distraction (HAI 6), and In-Kennel: Return (HAI 7). An additional category, Activity (HAI 8),
was included to gather supplementary information on the dog’s behavior while in-kennel and
36
during transition between the kennel and the assessment area, both between HAI 1-HAI 2 and
HAI 6-HAI 7. Each interaction has a series of prompted observations for record of the dog’s
response to the interaction. The order of interactions was intentionally designed so that
potential excitability of each category increased over time except for Play (HAI 2) which is meant
to provide the dog with a positive stimulating experience in the beginning to properly assess the
individual's ability to down-regulate afterward during Settle (HAI 3).
2.2.vi: Human-animal Interactions
In-kennel: Begin (HAI 1)
The assessor approached the front of the dog’s kennel and stood facing the kennel gate,
calling to the dog in a friendly manner and offering a treat (Zukes Mini Natural). If the dog
approached the human, she held the treat through the gate. If the dog did not take the treat,
she dropped it on the floor. If the dog does not remain near the gate, the assessor lightly tossed
the treat further into the enclosure. After allowing the dog the opportunity to consume the
treat, the assessor slowly opened the gate so that the dog could exit while the assessor secured
the dog using a slip lead. If the dog did not exit within 90 seconds of coaxing, the assessor shut
the gate and tried from the backside gate. If again the dog did not exit within 90 seconds, the
assessor recorded refusal of exit and ended assessment. For the dogs that did exit, they were
led outside for a brief walk (approximately 1 minute) before entering the assessment area and
let off lead. While the dog was given 1-2 minutes to acclimate, the assessor sat in a relaxed
manner on a low crate in the middle of the area. If the dog approached the assessor, she would
interact with the dog by talking calmly or petting the dog.
Play (HAI 2)
The assessor stood and called to the dog to get its attention. With the dog looking in the
direction of the human, the assessor would gently toss a tennis ball, a rope toy, and a dog toy
(stuffless stuffy) one at a time a few feet in front of her and attempt to engage the dog in play. If
the dog showed interest in interaction, the assessor would either play or pet the dog according
to its interest. If the dog showed no interest, the assessor would allow the dog to continue its
behavior for the 3-minute play duration while intermittingly attempting to capture the dog’s
interest in play by calling to the dog, bouncing the ball, squeaking the stuffy, or gently tossing
37
the toys. If the dog exhibited increasingly aroused or avoidant behaviors (e.g., constant whining,
pacing, and avoidance or uninhibited mouthing or mounting) then playtime ended early, and
toys were put away. At the end of playtime, the assessor would pick up toys if not in use or offer
a treat for a trade if the toys were in use, then place them out of reach of the dog.
Settle (HAI 3)
The assessor called the dog to get its attention and while the dog was looking in the
direction of the human, placed a large size Milkbone on a fabric mat. The assessor then sat
down approximately 8 feet from the mat to observe the dog for 2 minutes. During the 2
minutes, the assessor sat calmly and ignored solicitations for attention.
Acoustic startle (HAI 4)
The assessor remained seated and played a loud buzzer noise via a mobile app through a
Bluetooth speaker placed on the floor in the center of the assessment area. The assessor
observed the behavior of the dog through the duration of the sound and the following 3
seconds after the noise had ceased.
Visual startle (HAI 5)
The assessor attached a leash to the dog’s collar or harness and walked the dog over to a
tethered rope, where the assessor then secured the leash to the rope. As the assessor bent
down to pick up a separate rope attached to a metal folding chair, the assessor offered the dog
a treat (Zukes Mini Natural) to orient the dog to the chair. After standing, the assessor tugged
the rope so that the chair moved approximately 6 inches and observed the dog for the duration
of the chair movement and the following 3 seconds after movement ceased.
Distraction (HAI 6)
The assessor walked the dog to the end of the tethered leash and then continued
forward to where a model dog was hidden behind a barrier. The assessor brought the model
dog out from behind the visual shield and proceeded to walk the model dog by leash attached
to a harness for approximately 8 feet before stopping the model dog at a location out of reach
by the real dog. The assessor then walked toward the real dog, past the shoulder and behind, to
then call for the dog, offering a treat (Zukes Mini Natural). After 5-10 seconds, the assessor
unclipped the leash from the tethered rope and allowed the dog to approach the model dog.
38
Whether the dog approached the model dog or not, the assessor would then walk the model
dog an additional 4 feet, pausing to allow the dog to react while untethered. The assessor then
removed the model dog and placed it behind the barrier.
In-kennel: Return (HAI 7)
The assessor led the dog back to its kennel. When within approximately 5 feet of the
kennel entry, the assessor observed the manner of return. The assessor would continue walking
straight to the kennel and open the gate, verbally encouraging the dog if necessary. Once the
dog was inside the kennel, the assessor would place a medium size Milkbone in the kennel
while unclipping the leash. Finally, the assessor would close and secure the gate. If the dog
resisted entering the kennel, the assessor would use first the Milkbone as a lure and then wet
food.
Activity (HAI 8)
‘Activity’ represents miscellaneous behaviors that can correlate with maladaptive
coping. These behaviors are circled if observed either while the dog is in the kennel (during HAI
1 or HAI 7), or while the dog is transitioning to/from the kennel (after HAI 1 or before HAI 7).
2.2.vii: Analysis
Rater reliability was analyzed using R version 4.1.3; indoor and outdoor dogs were
analyzed separately. For inter-rater reliability, only dogs scored by rater pair A+B were included
in analysis (indoor = 35; outdoor = 37) because of the low number of dogs scored by rater pairs
A+C (indoor = 5; outdoor = 8) and A+D (indoor = 3; outdoor = 3). To test for intra-rater reliability,
Rater A watched assessment video a minimum of 4 months after the original assessment for
20% of the dogs assessed indoors (n = 10) and outdoors (n = 10). Dogs for intra-rater reliability
were randomly selected from the sampled population after excluding assessments that had ≥1
prompt with missing data from the original assessment performed by Rater A (n = 7). Prompt
responses from both scored (sc01-sc15) and unscored prompts (usc01-usc07, pilo1-pilo8) were
analyzed for agreement with percentage observed total agreement (Po) and Cohen’s kappa (k)
which adjusts for chance agreement using the proportion of agreement expected by chance
(Pe). Prompt responses from unscored prompts were also analyzed for proportion positive
agreement (Ppos) and proportion negative agreement (Pneg). Agreement analysis can result in
39
high percent agreement with low kappa values in instances where observed marginal totals are
symmetrically imbalanced by affecting the expected percent agreement, i.e., Pe (Feinstein &
Cicchetti, 1990). For this reason, it is suggested that Ppos and Pneg be reported in addition to Po
and k for a comprehensive understanding of the data (Cicchetti & Feinstein, 1990). See Table 2.2
for equations used for reliability analysis.
Unscored prompts are multiple choice (usc01-usc07) or binary present/absent (pilo1-
pilo8). One unscored prompt (supp.01) is open-ended, however written response was not
included in analysis. Of the seven multiple choice prompts, two are mutually exclusive (usc01
and usc03) and were analyzed using Po and k while each individual response was additionally
analyzed with Ppos and Pneg. Each response of all unscored prompts, including the two mutually
exclusive prompts, were analyzed separately as binary present or absent outcome variables to
test for Po, k, Ppos, and Pneg. Three prompts (usc05, usc06, usc07) include the response option
“Other (write)â€. The occurrence of response “Other†was included in reliability analysis;
however, written responses were not analyzed. Due to the relatedness between body language
images (responses for usc02 and usc04), images were grouped according to correlation and
analyzed as a group in addition to individually if there was more than one image in the body
language group. In total there was eight groups of correlated images, five of which consisted of
more than one image; Group 1 (images 1 and 2), Group 2 (images 3, 4, 5, and 6), Group 3
(images 7, 8, 9, and 10), Group 4 (images 11, 12, and 13), Group 5 (image 14), Group 6 (image
15), Group 7 (image 16) and Group 8 (image 17, 18, 19, and 20).
For the 15 scored prompts (sc01-sc15), both responses and scores were analyzed for
reliability. Several scored prompts (sc02, sc09, sc10, sc11, sc12, sc13, and sc14) had ≥2
responses that were scored 0 points, leading to the possibility of improved score agreement
over response agreement. Responses were analyzed for Po and k while prompt scores were
analyzed using the quadratic weighted kappa (QWK) for consideration of relatedness among
ranked integers (Vanbelle, 2016).
In summary, individual prompts were analyzed for reliability using the appropriate test
for unranked categorical data, totaling to 77 unscored binary response outcomes, five additional
unscored binary outcomes for body language groups with more than one correlated image, and
40
two mutually exclusive unscored prompts. Further, 15 mutually exclusive scored prompts were
analyzed for reliability both as unranked categorical data (response agreement) and as ranked
integers (score agreement). For interpretation of coefficient values, categorical descriptors
widely accepted as the reference standard (Cohen, 1960) will be referenced (Table 2.2(c)).
Table 2.2. Equations and symbols for reliability analysis. Assessment form data was analyzed
for inter- and intra-rater reliability using the appropriate measures of percent agreement,
proportion positive agreement, proportion negative agreement, and kappa value to correct for
chance. (a) Unscored assessment form data (responses for usc01-usc07 and pilo1-pilo8) was
rated by two raters for inter-rater reliability and again by a single rater using assessment video
for intra-rater reliability. The presence of the behavior in both instances was considered positive
agreement, count a, and the absence of the behavior in both instances was considered negative
agreement, count d. Total count included in analysis is designated with a capital N. (b) Test
coefficients used for rater reliability testing, the symbol used, and the equation. Po and k were
calculated for all prompt responses (usc01-usc07, sc01-sc15, and pilo1-pilo8). Ppos and Pneg were
calculated for nonexclusive unscored prompts (usc02, usc04-usc07, and pilo1-pilo8). QWK was
calculated for prompt scores (sc01-sc15). (c) Descriptive categories used for interpreting test
values, Po, Ppos, Pneg, k, and QWK (Cohen, 1960).
(a) Contingency key for binary data, agreement on presence or absence of behavior
Rater A (inter-rater)
or time 1 (intra-rater)
Present
Absent
Totals
Present
Absent
Totals
a
c
f1
b
d
f2
g1
g2
N
)
r
e
t
a
r
-
r
e
t
n
i
(
2
e
m
i
t
r
o
)
r
e
t
a
r
-
a
r
t
n
i
(
B
r
e
t
a
R
(b) Rater reliability test coefficients and equations
Test Coefficient Symbol & Equation Reporting
Description
Percent
agreement
Po = (a + d)
/ N
Pe = (f1g1 +f2g2)
/ N2
Percent
agreement
expected by
chance
Cohen’s kappa k = (Po – Pe)
/ (1 – Pe)
Unscored prompts (usc01-
usc07, pilo1-pilo8) and
scored prompt responses
(sc01(a), sc01(b), etc.)
Calculated for Cohen’s kappa
(unreported)
Unscored prompts (usc01-
usc07, pilo1-pilo8) and
scored prompt responses
(sc01(a), sc01(b), etc.)
where a is the positive agreement count, d is
negative agreement count, and N is total
observations
where f1 is total positive count by rater 1, g1 is
total positive count by rater 2, f2 is total
negative count by rater 1, and g2 is total
negative count by rater 2, and N is total
observations
Where Po is percent agreement and Pe is
percent agreement expected by chance
41
Table 2.2 (cont’d)
Proportion
positive
agreement
Proportion
negative
agreement
Quadratic
weighted kappa
Ppos = 2a
/ (N + (a - d))
Pneg = 2d
/ (N - (a – d))
QWK = 1 –
[(1 – k)
/ (1 – kmax)]
Nonexclusive unscored
prompts (usc02, usc04-usc07,
pilo1-pilo8)
Nonexclusive unscored
prompts (usc02, usc04-usc07,
pilo1-pilo8)
Prompt scores (sc01-sc15) where k is Cohen’s kappa and kmax is the
Where a is the positive agreement count, N is
total observations, and d is the negative
agreement count
Where d is the negative agreement count, N is
the total observations, and a is the positive
agreement count
maximum possible value of k which represents
the agreement expected by chance. In our
analysis, kmax was calculated in R through
library(irr), using observed scores
(c) Descriptive categories for test values
Test Value
Descriptive Category
Near perfect agreement
Substantial agreement
Moderate agreement
Fair agreement
Slight agreement
Worse than chance agreement
0.81-1.00
0.61-0.80
0.41-0.60
0.21-0.40
0.0-0.2
<0.0
2.3: Results
2.3.i: Inter-rater Reliability, Indoor Assessments
Unscored (Table 2.3) and scored (Table 2.4) responses showed a significant level of
agreement between raters, ranging from substantial (n = 8 unscored responses; n = 4 scored
responses) to near perfect (n = 73 unscored responses; n = 9 scored responses) percent
agreement (Po), with only three instances of moderate agreement (1 unscored response, 2
scored responses). However, when corrected for chance, agreement as measured by kappa
values ranged from worse than chance to near perfect agreement for both unscored responses
(k) and scored prompts (QWK), which is described in further detail below.
Unscored Prompts (indoor, inter-rater)
To assess agreement between raters A and B, I analyzed inter-rater reliability using
percent agreement (Po), proportion positive agreement (Ppos), proportion negative agreement
(Pneg), and Cohen’s kappa (k) to correct for chance agreement for all unscored responses. Both
mutually exclusive unscored prompts, location in kennel with closed door (usc01) and open
door (usc03), showed near perfect percent agreement (usc01(Po) = 0.971, usc03(Po) = 0.912).
42
However, when Cohen's kappa was calculated to correct for chance agreement, usc01 had
substantial agreement (k = 0.657), while usc03 had fair agreement (k = 0.227). To evaluate the
agreement for each possible response ('Front', 'Back', 'Middle', or 'Not Stationary' for both
prompts) responses were individually assessed for agreement.
For 32 unscored responses, the kappa value (k) could not be determined (k = NA)
because the proportion of agreement expected by chance (Pe) was equal to 1, resulting in a
denominator of 0 when calculating Cohen's kappa (Table 2.3). This occurred due to a low
observed occurrence, with 0 instances of positive agreement and a denominator of 0 for
proportion positive agreement (Ppos = NA), along with perfect occurrence of proportion negative
agreement (Pneg = 1) in all instances where k = NA.
Responses with unsatisfactory kappa values (k ≤ 0.60) had acceptable percent
agreement uncorrected for chance (Po > 0.60) in 35 instances and moderate percent agreement
in one instance (image 19, usc04(s)). Of the 36 responses that fall into this category, the
observed prevalence bias for 30 of them led to poor agreement when corrected for chance (k)
despite acceptable values for Po. There was no observed prevalence bias for the remaining six
responses (see “Moderate observed occurrence†in Table 2.3). Five of the responses with
moderate observed occurrence and unsatisfactory kappa values were body language images
observed during Play (HAI 2) while the remaining response was ‘None’ during miscellaneous
behavior in Settle (HAI 3). However, only one of these six responses (image 19, usc04(s)) had a
questionable degree of agreement (Po = 0.514). Image 19 is a dog sniffing with lowered head.
The ambiguous reliability results (Ppos = 0.370, Pneg = 0.605) could reflect confusion as to
whether the image pertains to sniffing the toys or the ground when not in proximity to the toys.
The remaining 12 responses had acceptable kappa values (k > 0.60) and acceptable
percent agreement uncorrected for chance (Po > 0.60). Nine of the responses had low observed
occurrence, one had high observed occurrence, and two had moderate observed occurrence
(Table 2.3).
43
Table 2.3. Inter-rater reliability of unscored prompts for indoor dogs (n = 35). Inter-rater reliability analysis (Po, Ppos, Pneg, k; see Table
2.2 for abbreviations) of unscored prompts for dogs assessed indoors by rater pair A + B (n = 35). Prompts with missing data were
excluded from analysis. Responses are organized by observed prevalence (low, high, and moderate) and listed in ascending order for
Cohen’s kappa (k). Low observed prevalence for a behavior, i.e., response, occurs when the number of dogs where the raters were in
negative agreement (d) exceeds the number of dogs for which either one or both raters recorded the behavior as present (a + b + c).
High observed prevalence for a behavior, i.e., response, occurs when the number of dogs where either one or both raters recorded
the behavior as present (a + b + c) exceeds the number of dogs where the raters were in negative agreement (d). Responses with low
or high observed prevalence indicate a prevalence bias that caused a discrepancy between Po and k in several instances, illuminated
by considering both Pneg and Ppos in context with Po and k. For instances where the number of dogs for which either one or both
raters recorded a behavior, i.e., response, as present (a + b + c) was within ±50% of the total number of dogs analyzed for that
prompt to the negative agreement count (0.50 x N* ± d). Moderately observed behaviors indicate that there was no prevalence bias
in the sample population for those responses.
Response
HAI
Disagree.
Abbrev.
Prompt
Neg.
Pos.
Pneg
Ppos
N*
Po
k
Agree.
Agree.
Count
Count (a)
Count (d)
(b + c)
Low observed occurrence (a + b + c < d) with high negative agreement (d)
1
1
1
1
1
1
1
1
Location after
usc01(c)
Middle
approach
Location with
usc03(c)
Middle
open door
Body language
usc02(d)
Image 4
Body language
usc02(e)
Image 5
Body language
usc02(g)
Image 7
Body language
usc02(i)
Image 9
Body language
usc02(m)
Image 13
Body language
usc02(o)
Image 15
35
34
35
35
35
35
35
35
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
NA
NA
NA
NA
NA
NA
NA
NA
1
1
1
1
1
1
1
1
NA
NA
NA
NA
NA
NA
NA
NA
35
34
35
35
35
35
35
35
0
0
0
0
0
0
0
0
44
Table 2.3 (cont’d)
1
1
1
1
2
2
2
2
2
2
2
2
3
3
3
4
5
7
8
8
8
Body language
usc02(p)
Image 16
Body language
usc02(q)
Image 17
Body language
usc02(s)
Image 19
Body language
usc02(t)
Image 20
Body language
usc04(a)
Image 1
Body language
usc04(b)
Image 2
Body language
usc04(a-b)
Group 1 (1-2)
Body language
usc04(c)
Image 3
Body language
usc04(m)
Image 13
Body language
usc04(o)
Image 15
Body language
usc04(p)
Image 16
Piloerection
pilo2
Present
Misc. behavior
usc05(a)
Pawing at exit
Misc. behavior
usc05(d)
Barking
Piloerection
Piloerection
Piloerection
Piloerection
pilo3
pilo4
pilo5
pilo7
Present
Present
Present
Present
In-kennel misc.
usc06(a)
Spinning
behavior
In-kennel misc.
usc06(b)
Excessive
behavior
jumping
In-kennel misc.
usc06(e)
Pacing
behavior
35
35
35
35
35
35
70
35
35
35
35
35
35
35
35
35
35
35
35
35
35
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
45
35
35
35
35
35
35
70
35
35
35
35
35
35
35
35
35
35
35
35
35
35
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Table 2.3 (cont’d)
8
8
8
1
1
1
1
1
1
1
1
1
1
1
2
2
2
2
In-kennel misc.
usc06(f)
Frantic pawing at
35
behavior
door
In-kennel misc.
usc06(g)
Smeared feces in
35
behavior
kennel
Out-of-kennel
usc07(a)
Dog directed
misc. behavior
reactivity
Body language
usc02(h)
Image 8
Body language
usc02(c)
Image 3
Location with
usc03(d)
Not stationary
35
35
35
34
open door
Body language
usc02(c-f)
Group 2 (3-6)
140
Body language
usc02(a-b)
Group 1 (1-2)
Location after
usc01(d)
Not stationary
approach
Location with
usc03(b)
Back
open door
Body language
usc02(a)
Image 1
Body language
usc02(b)
Image 2
Body language
usc02(f)
Image 6
Body language
usc02(l)
Image 12
Body language
usc04(d)
Image 4
Body language
usc04(e)
Image 5
Body language
usc04(g)
Image 7
Body language
usc04(l)
Image 12
70
35
34
35
35
35
35
35
35
35
35
35
35
35
26
31
32
135
68
34
32
34
34
34
33
34
34
30
34
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
46
0
0
0
9
4
2
5
2
1
2
1
1
1
2
1
1
5
1
1
1
1
0.743
0.886
0.941
0.964
0.971
0.971
0.941
0.971
0.971
0.971
0.943
0.971
0.971
0.857
0.971
NA
NA
NA
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
0.852
0.939
0.970
0.982
0.986
0.986
0.970
0.986
0.986
0.986
0.971
0.986
0.986
0.923
0.986
NA
NA
NA
-0.145
-0.061
-0.030
-0.017
-0.014
0
0
0
0
0
0
0
0
0
0
Table 2.3 (cont’d)
2
2
2
2
8
8
8
8
6
1
2
2
2
1
1
8
1
1
3
1
Body language
usc04(n)
Image 14
Body language
usc04(q)
Image 17
Body language
usc04(r)
Image 18
Body language
usc04(t)
Image 20
In-kennel misc.
usc06(d)
Excessive barking
behavior
In-kennel misc.
usc06(h)
Other
behavior
Out-of-kennel
usc07(b)
Other
misc. behavior
Piloerection
Piloerection
pilo8
pilo6
Present
Present
Body language
usc02(g-j)
Group 3 (7-10)
Body language
usc04(c-f)
Group 2 (3-6)
Body language
usc04(f)
Image 6
Body language
usc04(q-t)
Group 8 (17-20)
Body language
usc02(r)
Image 18
Piloerection
pilo1
Present
In-kennel misc.
usc06(c)
Whining
behavior
35
35
35
35
35
35
35
35
35
140
140
35
140
35
35
35
Body language
usc02(q-t)
Group 8 (17-20)
140
Body language
usc02(j)
Image 10
Misc. behavior
usc05(c)
Whining
Body language
usc02(n)
Image 14
35
35
35
0
0
0
0
0
0
0
0
2
2
2
2
5
1
1
1
1
2
5
6
47
34
34
34
34
34
34
33
31
24
128
128
25
115
33
33
33
138
32
29
28
1
1
1
1
1
1
2
4
9
10
10
8
20
1
1
1
1
1
1
1
0.971
0.971
0.971
0.971
0.971
0.971
0.943
0.886
0.743
0.929
0.929
0.771
0.857
0.971
0.971
0.971
0.993
0.971
0.971
0.971
0
0
0
0
0
0
0
0
0.308
0.286
0.286
0.334
0.334
0.667
0.667
0.667
0.667
0.80
0.909
0.923
0.986
0.986
0.986
0.986
0.986
0.986
0.971
0.939
0.842
0.962
0.962
0.862
0.920
0.985
0.985
0.985
0.996
0.985
0.983
0.982
0
0
0
0
0
0
0
0
0.203
0.248
0.261
0.263
0.263
0.653
0.653
0.653
0.663
0.785
0.892
0.906
Table 2.3 (cont’d)
1
3
Location after
usc01(b)
Back
approach
Misc. behavior
usc05(b)
Pacing
35
35
High observed occurrence (a + b + c > d) with high positive agreement (a)
1
2
3
1
Body language
usc02(k)
Image 11
Body language
usc04(k)
Image 11
Misc. Behavior
usc05(e)
Other
Location with
usc03(a)
Front
35
35
35
34
open door
1
2
29
30
19
31
1
Location after
usc01(a)
Front
35
33
approach
Moderate observed occurrence (a + b + c is ±50%(N) of d)
2
2
3
2
2
2
1
2
Body language
usc04(s)
Image 19
Body language
usc04(j)
Image 10
Misc. behavior
usc05(f)
None
Body language
usc04(h)
Image 8
Body Language
usc04(g-j)
Group 3 (7-10)
Body Language
usc04(i)
Image 9
Body language
usc02(k-m) Group 4 (11-13)
Body language
usc04(k-m)
Group 4 (11-13)
35
35
35
35
140
35
105
105
5
7
5
12
25
6
29
30
34
33
0
1
5
1
1
13
16
20
12
81
23
68
70
0
0
6
4
11
2
1
1
1
1
1
1
1
1
0.829
0.886
0.686
0.941
0.906
0.938
0.776
0.969
0
-0.050
0.333
0.476
0.50
0.278
0.306
0.477
1
0.971
0.985
0.667
0.653
17
12
10
11
34
6
8
5
0.514
0.657
0.714
0.686
0.757
0.829
0.924
0.952
0.370
0.538
0.50
0.686
0.595
0.667
0.879
0.923
0.605
0.727
0.80
0.686
0.827
0.885
0.944
0.966
0.105
0.266
0.342
0.428
0.429
0.553
0.823
0.889
N* = total dogs analyzed per response; prompts with missing data were excluded from analysis
48
Scored Prompts (indoor, inter-rater)
Scored prompts with unsatisfactory score agreement corrected for chance (QWK ≤ 0.60)
had unsatisfactory response agreement corrected for chance (k ≤ 0.60) in seven instances yet
acceptable response agreement (k > 0.60) in two instances (Table 2.4). When uncorrected for
chance, two of the nine instances were unacceptable (Po ≤ 0.60), the reaction to the acoustic
startle (sc10) and the reaction to the visual startle (sc11), both of which had differing response
agreement to score agreement due to three responses leading to an equivalent score of zero
(responses (e), (f), and (g) for both sc10 and sc11, see Table 2.1 for abbreviations).
Scored prompts with satisfactory score agreement corrected for chance (QWK > 0.60)
also had satisfactory response agreement corrected for chance (k > 0.60) in four instances, yet
unsatisfactory (k ≤ 0.60) in two instances, presentation with closed door (sc01) and presentation
with open door (sc03). However, when uncorrected for chance, all six scored prompts had
satisfactory agreement (Po > 0.60).
Table 2.4. Inter-rater reliability of scored prompts for indoor dogs (n = 35). Inter-rater
reliability analysis (Po, k, QWK; see Table 2.2 for abbreviations) of scored prompts for dogs
assessed indoors by rater pair A + B (n = 35), listed in ascending order for QWK. Prompts with
missing data were excluded from analysis. Scored prompts were analyzed for both response
agreement (Po and k) and score agreement (QWK) because seven of the prompts had more than
one response that was scored zero. Prompts with differing agreement count (a) between
responses and scores are highlighted with bold red font in columns ‘a’ for both response and
score agreement.
HAI
Abbrev
Prompt
1
7
3
6
4
5
2
7
1
2
6
sc02
sc14
sc09
sc12
sc10
sc11
sc05
sc15
sc04
sc06
sc13
Ability to take treat
Approach to kennel entry
Activity
Overall body language
Reaction to acoustic startle
Reaction to visual startle
Initial reaction to play
In-kennel behavior post-return
Latency to exit
Ease of putting away toys
Can you obtain the dog's
attention
N*
32
34
34
34
35
33
34
35
34
34
28
49
Response Agreement
Score Agreement
a
28
24
25
24
16
15
30
31
26
30
24
Po
0.875
0.706
0.735
0.706
0.457
0.455
0.882
0.886
0.765
0.882
0.857
k
0.739
0.473
0.578
0.518
0.278
0.321
0.783
0.551
0.487
0.770
0.697
a
30
26
28
24
19
23
30
31
26
30
24
k
-0.026
0.077
0.288
0.467
0.488
0.552
0.573
0.598
0.60
0.636
0.753
Table 2.4 (cont’d)
3
1
1
2
sc08
sc01
sc03
sc07
Response to treat
Presentation with closed door
Presentation with open door
Was play ended early
32
34
35
33
31
29
30
33
0.969
0.853
0.857
1
0.951
0.320
0.481
1
31
29
30
33
0.979
1
1
1
N* = total dogs analyzed per response; prompts with missing data were excluded from analysis
2.3.ii: Inter-rater Reliability, Outdoor Assessments
Unscored (Table 2.5) and scored (Table 2.6) responses showed a significant level of
agreement between raters, ranging from substantial (n = 10 unscored responses; n = 4 scored
responses) to near perfect (n = 72 unscored responses; n = 9 scored responses) percent
agreement (Po), with only three instances of moderate agreement (1 unscored response, 2
scored responses). However, when corrected for chance, agreement as measured by kappa
values ranged from worse than chance to near perfect agreement for both unscored responses
(k) and scored prompts (QWK), described in further detail below.
Unscored Prompts (outdoor, inter-rater)
To assess agreement between raters A and B, I analyzed inter-rater reliability using
percent agreement (Po), proportion positive agreement (Ppos), proportion negative agreement
(Pneg), and Cohen’s kappa (k) to correct for chance agreement for all unscored responses. Both
mutually exclusive unscored prompts, location in kennel with closed door (usc01) and open
door (usc03), showed near perfect percent agreement (usc01(Po) = 0.946, usc03(Po) = 0.970).
However, when Cohen's kappa was calculated to correct for chance agreement, usc01 had
moderate agreement (k = 0.486), while usc03 had slight agreement (k = 0). To evaluate the
agreement for each possible response ('Front', 'Back', 'Middle', or 'Not Stationary' for both
prompts) responses were individually assessed for agreement.
For 35 unscored responses, the kappa value (k) could not be determined (k = NA)
because the proportion of agreement expected by chance (Pe) was equal to 1, resulting in a
denominator of 0 when calculating Cohen's kappa (Table 2.5). This occurred due to a low
observed occurrence, with 0 instances of positive agreement and a denominator of 0 for
50
proportion positive agreement (Ppos = NA), along with perfect occurrence of proportion negative
agreement (Pneg = 1) in all instances where k = NA.
Responses with unsatisfactory kappa values (k ≤ 0.60) had acceptable percent
agreement uncorrected for chance (Po > 0.60) in 42 instances. Of the 42 responses, the
observed prevalence bias for 35 of them led to poor agreement when corrected for chance (k)
despite acceptable values for Po. There was no observed prevalence bias for the remaining
seven responses (see “Moderate observed occurrence†in Table 2.5).
The remaining five responses had acceptable kappa values (k > 0.60) and acceptable
percent agreement uncorrected for chance (Po > 0.60). Three of the responses had low
observed occurrence and two had moderate observed occurrence (Table 2.5).
51
Table 2.5. Inter-rater reliability of unscored prompts for outdoor dogs (n = 37). Inter-rater reliability analysis (Po, Ppos, Pneg, k; see
Table 2.2 for abbreviations) of unscored prompts for dogs assessed outdoors by rater pair A + B (n = 37). Prompts with missing data
were excluded from analysis. Responses are organized by observed prevalence (low, high, and moderate) and listed in ascending
order for Cohen’s kappa (k). Low observed prevalence for a behavior, i.e., response, occurs when the number of dogs where the
raters were in negative agreement (d) exceeds the number of dogs for which either one or both raters recorded the behavior as
present (a + b + c). High observed prevalence for a behavior, i.e., response, occurs when the number of dogs where either one or
both raters recorded the behavior as present (a + b + c) exceeds the number of dogs where the raters were in negative agreement
(d). Responses with low or high observed prevalence indicate a prevalence bias that caused a discrepancy between Po and k in
several instances, illuminated by considering both Pneg and Ppos in context with Po and k. For instances where the number of dogs for
which either one or both raters recorded a behavior, i.e., response, as present (a + b + c) was within ±50% of the total number of
dogs analyzed for that prompt to the negative agreement count (0.50 x N* ± d). Moderately observed behaviors indicate that there
was no prevalence bias in the sample population for those responses.
HAI
Response
Abbrev.
Prompt
Pneg
Ppos
N*
Po
k
Pos.
Agree.
(a)
Neg.
Agree.
(d)
Disagree.
(b + c)
Low observed occurrence (a + b + c < d) with high negative agreement (d)
1
1
1
1
1
1
1
1
1
1
Location with
open door
Location with
open door
Body language
usc03(b)
Back
usc03(c)
Middle
usc02(b)
Image 2
Body language
usc02(d)
Image 4
Body language
usc02(e)
Image 5
Body language
usc02(i)
Image 9
Body language
usc02(m)
Image 13
Body language
usc02(o)
Image 15
Body language
usc02(p)
Image 16
Body language
usc02(q)
Image 17
33
33
37
37
37
37
37
37
37
37
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
1
1
1
1
1
1
1
1
1
1
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
33
33
37
37
37
37
37
37
37
37
0
0
0
0
0
0
0
0
0
0
52
Table 2.5 (cont’d)
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
4
5
7
8
8
Body language
usc02(s)
Image 19
Body language
usc02(t)
Image 20
Piloerection
during In-kennel:
Begin
Body language
pilo1
Present
usc04(b)
Image 2
Body language
usc04(c)
Image 3
Body language
usc04(e)
Image 5
Body language
usc04(l)
Image 12
Body language
usc04(m)
Image 13
Body language
usc04(o)
Image 15
Body language
usc04(p)
Image 16
Body language
usc04(q)
Image 17
Body language
usc04(r)
Image 18
Piloerection
pilo2
Present
Misc. behavior
usc05(b)
Pacing
Misc. behavior
usc05(d)
Barking
pilo3
pilo4
pilo5
pilo7
Present
Present
Present
Present
usc06(a)
Spinning
Piloerection
Piloerection
Piloerection
Piloerection
In-kennel misc.
behavior
In-kennel misc.
behavior
usc06(b)
Excessive jumping
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
37
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
53
Table 2.5 (cont’d)
In-kennel misc.
behavior
In-kennel misc.
behavior
In-kennel misc.
behavior
In-kennel misc.
behavior
Body language
usc06(d)
Excessive barking
37
usc06(e)
Pacing
usc06(f)
usc06(g)
usc02(f)
Frantic pawing at
door
Smeared feces in
kennel
Image 6
Body language
usc02(j)
Image 10
Body language
usc04(d)
Image 4
Body language
usc04(t)
Image 20
Body language
usc02(c-f)
Group 2 (3-6)
Location after
approach
Location after
approach
Location with
open door
Body language
usc01(b)
Back
usc01(c)
Middle
usc03(d)
Not Stationary
usc02(a)
Image 1
Body language
usc02(a-b) Group 1 (1-2)
Body language
usc02(c)
Image 3
Body language
usc02(g)
Image 7
Body language
usc04(a)
Image 1
Body language
usc04(a-b) Group 1 (1-2)
Body language
usc04(g)
Image 7
Body language
usc04(n)
Image 14
Misc. behavior
usc05(a)
Pawing at exit
8
8
8
8
1
1
2
2
1
1
1
1
1
1
1
1
2
2
2
2
3
37
37
37
37
37
37
37
148
37
37
33
37
74
37
37
37
74
37
37
37
0
0
0
0
6
2
2
2
8
1
1
1
1
1
2
2
1
1
9
2
1
1
1
1
1
0.838
0.946
0.946
0.946
0.946
0.973
0.973
0.970
0.973
0.986
0.946
0.946
0.973
0.986
0.757
0.946
0.973
NA
NA
NA
NA
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
0.912
0.972
0.972
0.972
0.972
0.986
0.986
0.985
0.986
0.993
0.972
0.972
0.986
0.993
0.862
0.972
0.986
NA
NA
NA
NA
-0.047
-0.028
-0.028
-0.028
-0.012
0
0
0
0
0
0
0
0
0
0
0
0
37
37
37
37
31
35
35
35
140
36
36
32
36
73
35
35
36
73
28
35
36
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
54
Table 2.5 (cont’d)
6
8
8
8
2
1
1
8
1
1
1
1
8
2
3
2
1
Piloerection
pilo6
Present
In-kennel misc.
behavior
Out-of-kennel
misc. behavior
Piloerection
usc06(h)
Other
usc07(b)
Other
pilo8
Present
Body language
usc04(c-f)
Group 2 (3-6)
Body language
usc02(l)
Image 12
Body language
usc02(g-j)
Group 3 (7-10)
In-kennel misc.
behavior
Body language
usc06(c)
Whining
usc02(r)
Image 18
Body language
usc02(q-t)
Group 8 (17-20)
Body language
usc02(h)
Image 8
Body language
usc02(n)
Image 14
Out-of-kennel
misc. behavior
Body language
usc07(a)
usc04(q-t)
Dog directed
reactivity
Group 8 (17-20)
37
37
37
37
148
37
148
37
37
148
37
37
37
0
0
0
0
1
1
1
1
1
1
1
4
2
30
36
36
33
135
31
141
32
33
144
34
28
32
148
11
122
Misc. behavior
usc05(c)
Whining
Body language
usc04(i)
Image 9
Location after
approach
usc01(d)
Not Stationary
37
37
37
High observed occurrence (a + b + c > d) with high positive agreement (a)
2
1
Body language
usc04(k)
Image 11
Location with
open door
usc03(a)
Front
37
33
34
29
36
0
0
2
6
1
29
32
55
7
1
1
4
12
5
6
4
3
3
2
5
3
15
1
2
0
8
1
0.811
0.973
0.973
0.892
0.919
0.865
0.959
0.892
0.919
0.980
0.946
0.865
0.919
0.899
0.973
0.946
1
0
0
0
0
0.143
0.286
0.250
0.333
0.400
0.400
0.500
0.615
0.571
0.595
0.800
0.857
1
0.784
0.970
0.879
0.985
0.896
0.986
0.986
0.943
0.957
0.925
0.979
0.941
0.957
0.990
0.971
0.918
0.955
0.942
0.986
0.967
1
0
0
0
0
0
0
0.123
0.213
0.231
0.275
0.373
0.393
0.471
0.534
0.536
0.541
0.786
0.825
1
-0.121
0
Table 2.5 (cont’d)
1
1
Body language
usc02(k)
Image 11
Location after
approach
usc01(a)
Front
Moderate observed occurrence (a + b + c is ±50%(N) of d)
2
2
3
3
2
2
2
1
2
Body language
usc04(f)
Image 6
Body language
usc04(s)
Image 19
Misc. behavior
usc05(e)
Other
Misc. behavior
usc05(f)
None
Body language
usc04(j)
Image 10
Body language
usc04(h)
Image 8
Body language
usc04(g-j)
Group 3 (7-10)
Body language
usc02(k-m) Group 4 (11-13)
Body language
usc04(k-m) Group 4 (11-13)
37
37
37
37
37
37
37
37
148
111
111
29
34
1
11
15
10
8
7
21
30
29
0
1
26
13
10
15
18
20
95
68
74
8
2
10
13
12
12
11
10
32
13
8
0.784
0.946
0.879
0.971
0
0
0.500
0.479
0.730
0.649
0.676
0.676
0.703
0.730
0.784
0.883
0.928
0.167
0.629
0.714
0.625
0.593
0.583
0.568
0.822
0.879
0.839
0.667
0.625
0.714
0.766
0.800
0.856
0.913
0.949
0.123
0.335
0.347
0.347
0.359
0.415
0.430
0.736
0.828
N* = total dogs analyzed per response; prompts with missing data were excluded from analysis
56
Scored Prompts (outdoor, inter-rater)
Scored prompts with unsatisfactory score agreement corrected for chance (QWK ≤ 0.60)
had unsatisfactory response agreement corrected for chance (k ≤ 0.60) in seven instances yet
acceptable response agreement (k > 0.60) in one instance (Table 2.6). When uncorrected for
chance, one of the eight instances were unacceptable (Po ≤ 0.60), the reaction to the acoustic
startle (sc10), which had differing response agreement to score agreement due to three
responses leading to an equivalent score of zero (responses (e), (f), and (g), see Table 2.1 for
abbreviations).
Scored prompts with satisfactory score agreement corrected for chance (QWK > 0.60)
also had satisfactory response agreement corrected for chance (k > 0.60) in three instances, yet
unsatisfactory (k ≤ 0.60) in two instances, presentation with open door (sc03), and reaction to
visual startle (sc11). However, when uncorrected for chance, only reaction to visual startle
(sc11) had unsatisfactory agreement (Po < 0.60). Two prompts, presentation with closed door
(sc01) and whether play was ended early (sc07) had incalculable score agreement corrected for
chance (QWK = NA). Response agreement was incalculable for sc07 (k = NA) due to perfect
agreement yet unacceptable for sc01 (k = 0.358) despite near perfect agreement uncorrected
for chance (Po = 0.919).
Table 2.6. Inter-rater reliability of scored prompts for outdoor dogs (n = 37). Inter-rater
reliability analysis (Po, k, QWK; see Table 2.2 for abbreviations) of scored prompts for dogs
assessed outdoors by rater pair A + B (n = 37), listed in ascending order for QWK. Prompts with
missing data were excluded from analysis. Scored prompts were analyzed for both response
agreement (Po and k) and score agreement (QWK) because seven of the prompts had more than
one response that was scored zero. Prompts with differing agreement count (a) between
responses and scores are highlighted with bold red font in columns ‘a’ for both response and
score agreement.
HAI
Abbrev
Prompt
7
7
sc14
sc15
4
sc10
Approach to
kennel entry
In-kennel
behavior post-
return
Reaction to
acoustic startle
Response Agreement
Score Agreement
N*
36
35
a
25
30
Po
k
0.694
0.441
0.857
0.316
a
30
30
QWK
0
0
37
17
0.459
0.295
20
0.156
57
Table 2.6 (cont’d)
1
3
6
2
6
1
5
2
3
1
1
2
sc04
sc09
sc13
sc05
sc12
sc02
sc11
sc06
sc08
sc03
sc01
sc07
Latency to exit
Activity
Can you obtain
the dog’s
attention
Initial reaction to
play
Overall body
language
Ability to take
treat
Reaction to visual
startle
Ease of putting
away toys
Response to treat
Presentation with
open door
Presentation with
closed door
Was play ended
early
36
36
34
36
37
31
35
33
33
36
37
37
23
25
28
30
28
30
12
31
32
33
34
37
0.639
0.694
0.824
0.088
0.435
0.598
0.833
0.732
0.757
0.543
0.968
0.938
0.343
0.159
0.939
0.836
0.970
0.917
0.947
0.538
0.919
0.358
1
NA
23
31
28
30
28
29
32
31
32
33
34
37
0.235
0.262
0.427
0.441
0.529
0.622
0.682
0.895
0.977
1
NA
NA
N* = total dogs analyzed per response; prompts with missing data were excluded from analysis
2.3.iii: Intra-rater Reliability, Indoor Assessments
Unscored (Table 2.7) and scored (Table 2.8) responses showed a significant level of
agreement between repeat assessment by rater A, using assessment video, ranging from
substantial (n = 7 unscored responses; n = 4 scored responses) to near perfect (n = 62 unscored
responses; n = 10 scored responses) percent agreement (Po), with only three instances of
moderate agreement (n = 2 unscored responses, n = 1 scored response) and one instance of fair
agreement (n = 1 unscored response). However, when corrected for chance, agreement as
measured by kappa values ranged from worse than chance to near perfect agreement for both
unscored responses (k) and scored prompts (QWK), described in further detail below.
Unscored Prompts (indoor, intra-rater)
To assess agreement within rater A, I analyzed intra-rater reliability using percent
agreement (Po), proportion positive agreement (Ppos), proportion negative agreement (Pneg), and
58
Cohen’s kappa (k) to correct for chance agreement for all unscored responses. Both mutually
exclusive unscored prompts, location in kennel with closed door (usc01) and open door (usc03),
had perfect percent agreement (usc01(Po) = 1.0, usc03(Po) = 1.0). However, when Cohen's kappa
was calculated to correct for chance agreement, usc01 was incalculable (k = NA), while usc03
was perfect (k = 1.0). To evaluate the agreement for each response ('Front', 'Back', 'Middle', or
'Not Stationary' for both prompts) responses were individually assessed for agreement.
For 46 unscored responses, the kappa value (k) could not be determined (k = NA)
because the proportion of agreement expected by chance (Pe) was equal to 1, resulting in a
denominator of 0 when calculating Cohen's kappa (Table 2.7). This occurred due to a low
observed occurrence, with 0 instances of positive agreement and a denominator of 0 for
proportion positive agreement (Ppos = NA), along with perfect occurrence of proportion negative
agreement (Pneg = 1) for 45 of the instances where k = NA. The remaining response, conversely,
had perfect positive agreement (Ppos = 1.0) and zero instances of negative agreement (Pneg =
NA).
Responses with unsatisfactory kappa values (k ≤ 0.60) had acceptable percent
agreement uncorrected for chance (Po > 0.60) in 18 instances but unacceptable percent
agreement in three instances, image 18 (usc02(r)) and image 11 (usc04(k)) which had moderate
percent agreement (Po = 0.60) and image 19 (usc04(s)) which had fair agreement (Po = 0.30). Of
the 18 instances of acceptable percent agreement, the observed prevalence bias for 15 of them
led to poor agreement when corrected for chance (k) despite acceptable values for Po. There
was no observed prevalence bias for the remaining three which had substantial to near perfect
percent agreement (Po ≥ 0.80) (see “Moderate observed occurrence†in Table 2.7).
The remaining 12 responses had acceptable kappa values (k > 0.60) and acceptable
percent agreement uncorrected for chance (Po > 0.60). Seven of the responses had low
observed occurrence, three had high observed occurrence, and two had moderate observed
occurrence (Table 2.7).
59
Table 2.7. Intra-rater reliability of unscored prompts for indoor dogs (n = 10). Intra-rater reliability analysis (Po, Ppos, Pneg, k; see Table
2.2 for abbreviations) of unscored prompts for dogs assessed indoors by rater A (n = 10). Prompts with missing data were excluded
from analysis. Responses are organized by observed prevalence (low, high, and moderate) and listed in ascending order for Cohen’s
kappa (k). Low observed prevalence for a behavior, i.e., response, occurs when the number of dogs where the raters were in negative
agreement (d) exceeds the number of dogs for which either one or both raters recorded the behavior as present (a + b + c). High
observed prevalence for a behavior, i.e., response, occurs when the number of dogs where either one or both raters recorded the
behavior as present (a + b + c) exceeds the number of dogs where the raters were in negative agreement (d). Responses with low or
high observed prevalence indicate a prevalence bias that caused a discrepancy between Po and k in several instances, illuminated by
considering both Pneg and Ppos in context with Po and k. For instances where the number of dogs for which either one or both raters
recorded a behavior, i.e., response, as present (a + b + c) was within ±50% of the total number of dogs analyzed for that prompt to
the negative agreement count (0.50 x N* ± d). Moderately observed behaviors indicate that there was no prevalence bias in the
sample population for those responses.
Abbrev.
HAI
Response
Prompt
Pneg
Ppos
Po
k
Pos.
Agree.
(a)
Neg.
Agree.
(d)
Disagree.
(b + c)
Low observed occurrence (a + b + c < d) with high negative agreement (d)
1
1
1
1
1
1
1
1
1
1
1
Location with open door
usc01(b)
Back
Location with open door
usc01(c)
Middle
Location with open door
usc01(d)
Not Stationary
Location with open door
usc03(b)
Back
Location with open door
usc03(c)
Middle
Body language
Body language
Body language
Body language
Body language
Body language
usc02(b)
Image 2
usc02(d)
Image 4
usc02(e)
Image 5
usc02(f)
Image 6
usc02(g)
Image 7
usc02(i)
Image 9
0
0
0
0
0
0
0
0
0
0
0
60
10
10
10
10
10
10
10
10
10
10
10
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
1
1
1
1
1
1
1
1
1
1
1
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Table 2.7 (cont’d)
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
2
2
2
3
3
3
Body language
Body language
Body language
Body language
Body language
Body language
Body language
Piloerection
Body language
Body language
Body language
Body language
Body language
Body language
Body language
Body language
Body language
Body language
Body language
Body language
Piloerection
Misc. behavior
Misc. behavior
Misc. behavior
usc02(l)
Image 12
usc02(m)
Image 13
usc02(o)
Image 15
usc02(p)
Image 16
usc02(q)
Image 17
usc02(s)
Image 19
usc02(t)
Image 20
pilo1
Present
usc04(a)
Image 1
usc04(b)
Image 2
usc04(a-b) Group 1 (1-2)
usc04(c)
Image 3
usc04(e)
Image 5
usc04(g)
Image 7
usc04(l)
Image 12
usc04(m)
Image 13
usc04(n)
Image 14
usc04(o)
Image 15
usc04(p)
Image 16
usc04(q)
Image 17
pilo2
Present
usc05(a)
Pawing at exit
usc05(e)
Pacing
usc05(d)
Barking
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
61
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Table 2.7 (cont’d)
3
4
5
8
8
8
8
8
8
8
1
1
1
2
2
2
3
7
8
8
8
8
Piloerection
Piloerection
Piloerection
pilo3
pilo4
pilo5
Present
Present
Present
In-kennel misc. behavior
pilo8
Spinning
In-kennel misc. behavior
usc06(b)
Excessive jumping
In-kennel misc. behavior
usc06(d)
Excessive barking
In-kennel misc. behavior
usc06(e)
Pacing
In-kennel misc. behavior
usc06(f)
In-kennel misc. behavior
usc06(g)
Out-of-kennel misc.
behavior
Body language
usc07(a)
usc02(h)
Frantic pawing at
door
Smeared feces in
kennel
Dog directed
reactivity
Image 8
Body language
Body language
Body language
Body language
Body language
Misc. behavior
Piloerection
usc02(a)
Image 1
usc02(a-b) Group 1 (1-2)
usc04(d)
Image 4
usc04(r)
Image 18
usc04(t)
Image 20
usc05(c)
Whining
pilo7
Present
In-kennel misc. behavior
usc06(c)
Whining
In-kennel misc. behavior
usc06(h)
Other
Out-of-kennel misc.
behavior
Piloerection
usc07(b)
Other
pilo8
Present
10
10
10
10
10
10
10
10
10
10
8
9
19
9
9
9
8
9
9
9
9
8
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
62
0
0
0
0
0
0
0
0
0
0
2
1
1
1
1
1
2
1
1
1
1
2
1
1
1
1
1
1
1
1
1
1
0.80
0.90
0.950
0.90
0.90
0.90
0.80
0.90
0.90
0.90
0.90
0.80
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
0.889
0.947
0.974
0.947
0.947
0.947
0.889
0.947
0.947
0.947
0.947
0.889
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
-0.111
0
0
0
0
0
0
0
0
0
0
0
Table 2.7 (cont’d)
1
2
1
1
1
2
3
2
1
1
1
Body language
Body language
Body language
Body language
Body language
Body language
Misc. behavior
Body language
usc02(r)
Image 18
usc04(q-t)
Group 8 (17-20)
usc02(q-t)
Group 8 (17-20)
usc02(g-j)
Group 3 (7-10)
usc02(j)
Image 10
usc04(j)
Image 10
usc05(f)
None
usc04(c-f)
Group 2 (3-6)
Location with open door
usc03(d)
Not Stationary
Body language
Body language
usc02(c)
Image 3
usc02(c-f)
Group 2 (3-6)
0
3
1
1
1
1
1
2
1
1
1
5
28
35
36
8
8
8
36
9
9
39
4
9
4
3
1
1
1
2
0
0
0
High observed occurrence (a + b + c > d) with moderate to perfect proportion positive agreement (Ppos)
Location after approach
usc01(a)
Front
10
1
2
2
3
1
1
Body language
Body language
Misc. behavior
Body language
usc04(s)
Image 19
usc04(k)
Image 11
usc05(e)
Other
usc02(k)
Image 11
Location with open door
usc03(a)
Front
Moderate observed occurrence (a + b + c is ±50%(N) of d)
2
6
2
Body language
usc04(h)
Image 8
Piloerection
pilo6
Present
Body language
usc04(g-j)
Group 3 (7-10)
0
0
0
1
2
1
7
7
30
0
7
4
1
1
0
2
2
5
3
6
8
7
9
1
1
5
63
0.60
0.775
0.90
0.925
0.90
0.90
0.90
0.950
1
1
1
1
0.30
0.60
0.90
0.90
1
0.80
0.80
0.333
0.40
0.333
0.40
0.667
0.667
0.667
0.667
1
1
1
1
0.462
0.750
0.941
0.933
1
0.50
0.50
0.875
0.667
0.714
0.862
0.946
0.960
0.941
0.941
0.941
0.973
1
1
1
NA
0
0
0.667
0.80
1
0.875
0.875
0.923
0.20
0.262
0.304
0.362
0.615
0.615
0.615
0.643
1
1
1
NA
-0.40
-0.176
0.615
0.737
1
0.412
0.412
0.590
Table 2.7 (cont’d)
2
1
Body language
Body language
usc04(i)
Image 9
usc02(k-m) Group 4 (11-13)
3
7
5
22
2
1
0.800
0.967
0.750
0.933
0.833
0.978
0.60
0.911
N* = total dogs analyzed per response; prompts with missing data were excluded from analysis
64
Scored Prompts (indoor, intra-rater)
Scored prompts with unsatisfactory score agreement corrected for chance (QWK ≤ 0.60)
had unsatisfactory response agreement corrected for chance (k ≤ 0.60) in two instances yet
acceptable response agreement (k > 0.60) in one instance (Table 2.8). When uncorrected for
chance, all three instances had acceptable response agreement (Po > 0.60).
Scored prompts with satisfactory score agreement corrected for chance (QWK > 0.60)
also had satisfactory response agreement corrected for chance (k > 0.60) in six instances, yet
unsatisfactory (k ≤ 0.60) in one instance, ‘Activity’ during HAI 3 (sc09), which had unacceptable
agreement when uncorrected for chance (Po = 0.60).
Five prompts had incalculable score agreement corrected for chance (QWK = NA).
Response agreement was incalculable (k = NA) for whether play was ended early (sc07) due to
perfect agreement yet unacceptable (k ≤ 0.60) for ‘Presentation with closed door’ (sc01) and
‘Ability to take treat’ (sc02). ‘Presentation with open door’ (sc03) and ‘Approach to kennel entry’
(sc14) had acceptable response agreement (k > 0.60). However, when uncorrected for chance,
all five prompts had substantial to near perfect percent agreement (Po ≥ 0.80).
Table 2.8. Intra-rater reliability of scored prompts for indoor dogs (n = 10). Intra-rater
reliability analysis (Po, k, QWK; see Table 2.2 for abbreviations) of scored prompts for dogs
assessed indoors by rater A (n = 10), listed in ascending order for QWK. Prompts with missing
data were excluded from analysis. Scored prompts were analyzed for both response agreement
(Po and k) and score agreement (QWK) because seven of the prompts had more than one
response that was scored zero. Prompts with differing agreement count (a) between responses
and scores are highlighted with bold red font in columns ‘a’ for both response and score
agreement.
Response Agreement
Score Agreement
HAI
Abbrev
Prompt
7
5
4
6
6
1
3
sc15
sc11
sc10
sc13
sc12
sc04
sc09
In-kennel behavior post-
return
Reaction to visual startle
Reaction to acoustic
startle
Can you obtain the dog's
attention
Overall body language
Latency to exit
Activity
Po
0.9
0.70
0.80
k
0
0.50
0.697
0.80
0.636
0.90
0.90
0.60
0.737
0.80
0.452
a
9
7
8
8
9
9
9
QWK
0
0.455
0.524
0.714
0.737
0.80
0.857
a
9
7
8
8
9
9
6
65
Table 2.8 (cont’d)
2
2
3
1
1
1
2
7
sc05
sc06
sc08
sc01
sc02
sc03
sc07
sc14
Initial reaction to play
Ease of putting away toys
Response to treat
Presentation with closed
door
Ability to take treat
Presentation with open
door
Was play ended early
Approach to kennel entry
10
10
10
9
8
9
10
9
1
1
1
0.90
0.80
0.90
1
0.90
1
1
1
0
0.524
0.615
NA
0.804
10
10
10
9
10
9
10
10
1
1
1
NA
NA
NA
NA
NA
2.3.iv: Intra-rater Reliability, Outdoor Assessments
Unscored (Table 2.9) and scored (Table 2.10) responses showed a significant level of
agreement between repeat assessment by rater A, using assessment video, ranging from
substantial (n = 6 unscored responses; n = 2 scored responses) to near perfect (n = 63 unscored
responses; n = 10 scored responses) percent agreement (Po), with only six instances of
moderate agreement (n = 3 unscored responses, n = 3 scored response). However, when
corrected for chance, agreement as measured by kappa values ranged from worse than chance
to near perfect agreement for both unscored responses (k) and scored prompts (QWK),
described in further detail below.
Unscored Prompts (outdoor, intra-rater)
To assess agreement within rater A, I analyzed inter-rater reliability using percent
agreement (Po), proportion positive agreement (Ppos), proportion negative agreement (Pneg), and
Cohen’s kappa (k) to correct for chance agreement for all unscored responses. Both mutually
exclusive unscored prompts, location in kennel with closed door (usc01) and open door (usc03),
had near perfect percent agreement (usc01(Po) = 0.90, usc03(Po) = 1.0). However, when Cohen's
kappa was calculated to correct for chance agreement, usc01 had substantial agreement (k =
0.615), while usc03 was incalculable (k = NA). To evaluate the agreement for each possible
response ('Front', 'Back', 'Middle', or 'Not Stationary' for both prompts) responses were
individually assessed for agreement.
66
For 50 unscored responses, the kappa value (k) could not be determined (k = NA)
because the proportion of agreement expected by chance (Pe) was equal to 1, resulting in a
denominator of 0 when calculating Cohen's kappa (Table 2.9). This occurred due to a low
observed occurrence, with 0 instances of positive agreement and a denominator of 0 for
proportion positive agreement (Ppos = NA), along with perfect occurrence of proportion negative
agreement (Pneg = 1).
Responses with unsatisfactory kappa values (k ≤ 0.60) had acceptable percent
agreement uncorrected for chance (Po > 0.60) in 16 instances but unacceptable percent
agreement in three instances, image 11 (usc04(k)), image 9 (usc04(i)), and ‘Other’ (usc05(e))
which all had moderate percent agreement (Po = 0.50, 0.60, and 0.60 respectively). and image
19 (usc04(s)) which had fair agreement (Po = 0.30). For usc04(k), the observed prevalence bias
led to poor agreement when corrected for chance (k); however, there was no observed
prevalence bias for the latter two responses (see “Moderate observed occurrence†in Table 2.9).
The remaining 12 responses had acceptable kappa values (k > 0.60) and acceptable
percent agreement uncorrected for chance (Po > 0.60). Seven of the responses had low
observed occurrence, one had high observed occurrence, and four had moderate observed
occurrence (Table 2.9).
67
Table 2.9. Intra-rater reliability of unscored prompts for outdoor dogs (n=10). Intra-rater reliability (Po, Ppos, Pneg, k; see Table 2.2 for
abbreviations) of unscored prompts for dogs assessed outdoors by rater A (n = 10). Prompts with missing data were excluded from
analysis. Responses are organized by observed prevalence (low, high, and moderate) and listed in ascending order for Cohen’s kappa
(k). Low observed prevalence for a behavior, i.e., response, occurs when the number of dogs where the raters were in negative
agreement (d) exceeds the number of dogs for which either one or both raters recorded the behavior as present (a + b + c). High
observed prevalence for a behavior, i.e., response, occurs when the number of dogs where either one or both raters recorded the
behavior as present (a + b + c) exceeds the number of dogs where the raters were in negative agreement (d). Responses with low or
high observed prevalence indicate a prevalence bias that caused a discrepancy between Po and k in several instances, illuminated by
considering both Pneg and Ppos in context with Po and k. For instances where the number of dogs for which either one or both raters
recorded a behavior, i.e., response, as present (a + b + c) was within ±50% of the total number of dogs analyzed for that prompt to
the negative agreement count (0.50 x N* ± d). Moderately observed behaviors indicate that there was no prevalence bias in the
sample population for those responses.
Abbrev.
HAI
Response
Prompt
Pneg
Ppos
Po
k
Pos.
Agree.
(a)
Neg.
Agree.
(d)
Disagree.
(b + c)
Low observed occurrence (a + b + c < d) with high negative agreement (d)
1
1
1
1
1
1
1
1
1
1
1
Location with open door
usc01(c)
Middle
Location with open door
usc01(d)
Not Stationary
Location with open door
usc03(b)
Back
Location with open door
usc03(c)
Middle
Location with open door
usc03(d)
Not Stationary
Body language
Body language
Body language
Body language
Body language
Body language
usc02(a)
Image 1
usc02(b)
Image 2
usc02(a-b) Group 1 (1-2)
usc02(c)
Image 3
usc02(d)
Image 4
usc02(e)
Image 5
0
0
0
0
0
0
0
0
0
0
0
68
10
10
10
10
10
10
10
20
10
10
10
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
1
1
1
1
1
1
1
1
1
1
1
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Table 2.9 (cont’d)
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
2
2
2
2
2
Body language
Body language
Body language
Body language
Body language
Body language
Body language
Body language
Piloerection
Body language
Body language
Body language
Body language
Body language
Body language
Body language
Body language
Body language
Body language
Body language
Body language
Body language
Body language
Piloerection
usc02(g)
Image 7
usc02(i)
Image 9
usc02(j)
Image 10
usc02(m)
Image 13
usc02(o)
Image 15
usc02(p)
Image 16
usc02(s)
Image 19
usc02(t)
Image 20
pilo1
Present
usc04(a)
Image 1
usc04(b)
Image 2
usc04(a-b) Group 1 (1-2)
usc04(c)
Image 3
usc04(d)
Image 4
usc04(e)
Image 5
usc04(g)
Image 7
usc04(m)
Image 13
usc04(n)
Image 14
usc04(o)
Image 15
usc04(p)
Image 16
usc04(q)
Image 17
usc04(r)
Image 18
usc04(t)
Image 20
pilo2
Present
10
10
10
10
10
10
10
10
10
10
10
20
10
10
10
10
10
10
10
10
10
10
10
10
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
69
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Table 2.9 (cont’d)
3
3
3
3
4
5
7
8
8
8
8
8
8
8
8
1
1
1
1
2
3
8
Misc. behavior
Misc. behavior
Misc. behavior
Piloerection
Piloerection
Piloerection
Piloerection
usc05(a)
Pawing at exit
usc05(b)
Pacing
usc05(c)
Whining
pilo3
pilo4
pilo5
pilo7
Present
Present
Present
Present
In-kennel misc. behavior
usc06(a)
Spinning
In-kennel misc. behavior
usc06(b)
Excessive jumping
In-kennel misc. behavior
usc06(e)
Pacing
In-kennel misc. behavior
usc06(f)
In-kennel misc. behavior
usc06(g)
Frantic pawing at
door
Smeared feces in
kennel
Dog directed
reactivity
Other
usc07(a)
usc07(b)
pilo8
Present
usc02(h)
Image 8
usc02(g-j)
Group 3 (7-10)
usc02(l)
Image 12
usc02(q)
Image 17
usc04(l)
Image 12
usc05(d)
Barking
Out-of-kennel misc.
behavior
Out-of-kennel misc.
behavior
Piloerection
Body language
Body language
Body language
Body language
Body language
Misc. behavior
In-kennel misc. behavior
usc06(c)
Whining
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
2
2
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
0.80
0.950
0.90
0.90
0.90
0.90
0.90
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
0.889
0.974
0.947
0.947
0.947
0.947
0.947
-0.111
-0.026
0
0
0
0
0
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
8
38
9
9
9
9
9
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
70
Table 2.9 (cont’d)
8
8
2
2
1
1
1
1
1
2
2
2
In-kennel misc. behavior
usc06(d)
Excessive barking
In-kennel misc. behavior
usc06(h)
Other
Body language
Body language
Body language
usc04(g-j)
Group 3 (7-10)
usc04(k-m) Group 4 (11-13)
usc02(q-t)
Group 8 (17-20)
Location after approach
usc01(b)
Back
Body language
Body language
Body language
Body language
Body language
Body language
usc02(f)
Image 6
usc02(r)
Image 18
usc02(c-f)
Group 2 (3-6)
usc04(q-t)
Group 8 (17-20)
usc04(f)
Image 6
usc04(c-f)
Group 2 (3-6)
0
0
3
4
1
1
1
1
1
6
1
1
9
9
30
20
37
8
8
8
38
33
9
39
1
1
7
6
2
1
1
1
1
1
0
0
High observed occurrence (a + b + c > d) with substantial to near perfect proportion positive agreement (Ppos)
1
2
1
Body language
Body language
usc02(k)
Image 11
usc04(k)
Image 11
Location with open door
usc01(a)
Front
Moderate observed occurrence (a + b + c is ±50%(N) of d)
2
6
3
3
2
Body language
usc04(i)
Image 9
Piloerection
Misc. behavior
Misc. behavior
Body language
pilo6
Present
usc05(e)
Other
usc05(f)
None
usc04(h)
Image 8
0
1
1
6
7
3
4
7
2
5
1
4
3
4
3
2
8
4
8
0
0
3
3
1
71
0.90
0.90
0.825
0.80
0.950
0.90
0.90
0.90
0.975
0.975
1
1
0.80
0.50
0.90
0.60
0.70
0.60
0.70
0.80
0
0
0.462
0.571
0.50
0.667
0.667
0.667
0.667
0.923
1
1
0.889
0.615
0.941
0
0
0.60
0.667
0.50
0.947
0.947
0.896
0.870
0.974
0.941
0.941
0.941
0.987
0.985
1
1
0
0.286
0.667
0.750
0.824
0.60
0.727
0.875
0
0
0.358
0.444
0.481
0.615
0.615
0.615
0.655
0.908
1
1
-0.111
0
0.615
-0.250
-0.154
0.231
0.40
0.412
Table 2.9 (cont’d)
2
1
2
1
Body language
Body language
Body language
Body language
usc04(j)
Image 10
usc02(k-m) Group 4 (11-13)
usc04(s)
Image 19
usc02(n)
Image 14
2
8
6
3
7
19
3
7
1
3
1
0
0.90
0.90
0.90
1
0.80
0.842
0.923
1
0.933
0.927
0.857
1
0.737
0.769
0.783
1
72
Scored Prompts (outdoor, intra-rater)
Scored prompts with unsatisfactory score agreement corrected for chance (QWK ≤ 0.60)
had unsatisfactory response agreement corrected for chance (k ≤ 0.60) in three instances (Table
2.10). When uncorrected for chance, one instance had acceptable response agreement (Po >
0.60) while the remaining two, ‘Activity’ during HAI 3 (sc09) and ‘Overall body language’ during
HAI 6 (sc12) was unacceptable (Po = 0.60).
Scored prompts with satisfactory score agreement corrected for chance (QWK > 0.60)
also had satisfactory response agreement corrected for chance (k > 0.60) in all five instances.
The remaining seven prompts had incalculable score agreement corrected for chance (QWK =
NA). Response agreement was incalculable (k = NA) for three of the seven instances due to
perfect response agreement (Po = 1.0). For the other four instances, corrected for chance,
response agreement was acceptable in one instance, ‘Presentation with closed door’ (sc01) but
unacceptable for ‘Reaction to acoustic startle’ (sc10), ‘Reaction to visual startle’ (sc11), and
‘Approach to kennel entry’ (sc14). However, for two of these (sc11 and sc14), percent
agreement was acceptable when uncorrected for chance (Po = 0.80).
Table 2.10. Intra-rater reliability of scored prompts for outdoor dogs (n = 10). Intra-rater
reliability analysis (Po, k, QWK; see Table 2.2 for abbreviations) of scored prompts for dogs
assessed outdoors by rater A (n = 10), listed in ascending order for QWK. Prompts with missing
data were excluded from analysis. Scored prompts were analyzed for both response agreement
(Po and k) and score agreement (QWK) because seven of the prompts had more than one
response that was scored zero. Prompts with differing agreement count (a) between responses
and scores are highlighted with bold red font in columns ‘a’ for both response and score
agreement.
Response Agreement
Score Agreement
HAI
Abbrev
Prompt
3
7
6
2
1
2
3
sc09
sc15
sc12
sc06
sc02
sc05
sc08
Activity
In-kennel behavior post-return
Overall body language
Ease of putting away toys
Ability to take treat
Initial reaction to play
Response to treat
a
6
8
6
9
10
10
10
Po
0.60
0.80
0.60
0.90
1
1
1
k
0.333
0.412
0.310
0.615
1
1
1
a
8
8
6
9
10
10
10
QWK
0
0.118
0.375
0.615
1
1
1
73
Table 2.10 (cont’d)
6
1
1
1
2
4
5
7
sc13
sc01
sc03
sc04
sc07
sc10
sc11
sc14
Can you obtain the dog's attention
Presentation with closed door
Presentation with open door
Latency to exit
Was play ended early
Reaction to acoustic startle
Reaction to visual startle
Approach to kennel entry
2.4: Discussion
10
10
10
10
10
6
8
8
1
1
1
1
1
1
1
NA
NA
NA
0.60
0.80
0.80
0.50
0.583
0
10
10
10
10
10
6
10
10
1
NA
NA
NA
NA
NA
NA
NA
To support shelter dog welfare, a practical and reliable behavior assessment that
quantifies coping behavior would provide shelter staff with a tool that can communicate clinical
need for anxiolytic medication to veterinarians, streamlining individualized behavioral
intervention plans for dogs that are struggling to cope in the shelter environment. This novel
assessment can be performed in its entirety by a single handler with low-cost materials that
shelters are likely to have on hand. The most expensive item used for the assessment was the
model, stuffed dog which retails around $75.00 (US). The average length of time to complete
the assessment was 20 ± 4 minutes with a range of 10-42 minutes. Of the assessments that took
an excess of one standard deviation, 13 assessments lasted between 25 to 30 minutes. One
single assessment took 42 minutes because the dog was highly resistant to entering the kennel.
The assessment ended early before ‘In-kennel behavior post-return’ (sc15) could be scored.
Shelter dog behavior varies across individuals and therefore variation in assessment duration
can be expected. However, considering cooperative dogs as the baseline, an excess of 10
minutes may inhibit staff’s ability to routinely perform the assessment when time is a concern.
However, with further refinement based on validity analysis in Chapter 3, the assessment
duration could decrease, leading to improved practicality for staff use.
2.4.i: Inter-rater Reliability
For both indoor and outdoor dogs, unscored responses had acceptable levels of inter-
rater reliability demonstrated by near perfect negative agreement for responses with low
74
prevalence, substantial to near perfect positive agreement for responses with high prevalence,
and moderate to near perfect percent agreement for responses with moderate observed
occurrence. Kappa values were highly variable, reflecting the observed prevalence bias of
several of the responses. Prevalence may be specific to the sample population and Chapter 3
validity analysis will provide insight on which unscored prompts contribute clinically relevant
information on coping ability.
For both indoor and outdoor dogs, two scored prompts (sc10 and sc11) had fair to
moderate percent agreement. Acoustic Startle, HAI 4 and Visual Startle, HAI 5 consisted of sc10
and sc11, respectively. Additional training may improve inter-rater reliability; however, because
startle tests are contrary to the enriching purpose of the assessment, these prompts are
candidates for removal pending validity analysis in Chapter 3. For the remaining 13 scored
prompts, percent agreement on response was acceptable (substantial to near perfect) although
kappa values on score agreement varied significantly.
Discrepancies between response agreement and score agreement could occur for prompts that
had multiple responses that would result in zero (n = 7). Two of these, sc10 and sc11, are
candidates for removal as discussed. For the remaining five, one (sc02) had improved score
agreement from response agreement for dogs assessed indoors but worse score agreement for
dogs assessed outdoors, two (sc09 and sc14) had improved score agreement in both
assessment locations, and two (sc12 and sc13) had equivalent response to score agreement in
both locations. When score agreement improves, responses within a single prompt that are
scored zero could be combined while maintaining score integrity if the behaviors are also found
to be clinically equivalent. Alternatively, additional training may improve response agreement.
However, when score agreement worsens, as happened for sc02 during outdoor assessments,
reliability performance is unsatisfactory. Further refinement or elimination of sc02 could be
warranted pending validity analysis.
2.4.ii: Intra-rater Reliability
For both indoor and outdoor dogs, unscored responses had acceptable levels of intra-
rater reliability demonstrated by substantial to near perfect negative agreement for responses
with low prevalence, moderate to near perfect positive agreement for responses with high
75
prevalence, and moderate to near perfect percent agreement for responses with moderate
prevalence. Here kappa values reflected prevalence biases and varied significantly as was the
case for inter-rater reliability.
All fifteen scored prompts had acceptable intra-rater reliability ranging from moderate to
near perfect response agreement, although kappa values reflected prevalence biases and varied
significantly. Twelve (sc01, sc02, sc03, sc04, sc05, sc06, sc07, sc08, sc11, sc13, sc14, and sc15)
had substantial to near perfect response agreement in both assessment locations while three
(sc09, sc10, and sc12) had moderate response agreement in at least one of the locations. For
the three prompts with the lowest response agreement (moderate), the disagreements are
characterized by a shift to a more neutral response. For ‘Activity, HAI 3’ (sc09), the rating
changed from ‘Stressed - active’ to either ‘Neutral - stationary’ or ‘Neutral - active’ (n = 3) or
there was a shift between ‘Neutral - active’ and ‘Neutral - stationary’ (n = 4). For ‘Reaction to
acoustic startle, HAI 4’ (sc10), the rating changed from either ‘Flinches/startles’ or ‘Stops
briefly/orients’ to ‘Immediate approach’ (n = 4). Finally, for ‘Overall body language, HAI 6’
(sc12), the rating shifted from either Image 2 (lowered, forward lean) or Image 5 (alert, forward
motion) to Image 4 (neutral) (n = 7). The disagreements could in part be due to perception since
video recordings were used for the second rating and subtle behavioral cues may not translate
from live observation to observation of recordings.
Improved score agreement over response agreement occurred for four scored prompts
(sc02, sc09, sc11, and sc14). For sc02 and sc14, response disagreements did not result in a score
change, implying that the disagreement occurred between responses that were scored zero. For
sc09 and sc11, some disagreements resulted in a score change while others did not. Thus, for
sc02 and sc14, score integrity is maintained during disagreements, but this was not always the
case for the response disagreements observed for sc09 and sc11. As previously discussed,
response disagreements for sc09 could be due to the differences between real-life and video
observation and sc11 has been identified as a candidate for removal.
2.4.iii: Limitations
Assessment scores on the extreme ends of total coping score (scores near –45 or +45)
were not observed. The observed score range was relatively moderate (indoor range –18 to
76
+10; outdoor range –9 to +11.5) and is reflective of the prevalence imbalance for several
behavioral responses. Consequently, several behaviors were never observed and therefore
conclusions regarding rater agreement are limited. Distribution of total score is discussed
further in Chapter 3. Four dogs were excluded from enrollment because they were designated
‘Staff only’ due to high levels of aggression. It is possible that had these dogs been assessed,
they would have had high positive assessment scores. Six dogs were excluded for refusal to exit
the kennel. It is likely that these dogs were on the extreme end of the anxious-avoidant
spectrum but were not included in analysis due to the truncated assessment score.
Several assessments had at least one prompt that was left blank (rater A = 7 of 91
assessments; rater B = 36 of 72 assessments; rater C = 2 of 13 assessments; rater D = 0 of 6
assessments). Some instances may be attributed to perspective, as raters were instructed to
leave prompts blank if they were unable to see the dog’s response. However, most of the
missing data is likely due to human error. There has been extensive research on survey design to
maximize data quality with emphasis on tailored design methods and fitness of use (Biemer,
2010; Dillman, 2011). While the tool is not a survey in the traditional sense, it does require
citizen participation for completion. Since shelter staff would be motivated to comply, fitness of
use must consider ease of response. Design based on accessibility principles may improve
compliance by reducing participant burden (Lazar et al., 2004). Future integration into an app
could also encourage compliance by requiring input for each prompt before opening the next
question. Web survey research suggests that careful design consideration can decrease missing
data (Couper et al., 2001). The format of the tool used to record assessment results should be
further evaluated to maximize data quality.
The tool is intended for use throughout shelters in the United States. However, the
ability to extrapolate results may be circumscribed by the sample population. All assessments
took place at a single non-profit humane society in the Midwest; intake population may not be
representative of the national shelter dog population discussed in Chapter 3. Furthermore, dogs
that were unhealthy were excluded from enrollment as health influence on behavior would be
confounded with coping behavior. Further methodological and environmental limitations are
discussed in Chapter 3 during discussion on validity.
77
2.5: Conclusion
Several behavioral responses had low or high prevalence which created a wide range of
variability among agreement when corrected for chance (kappa values). For this reason, percent
agreement, proportion positive agreement, and proportion negative agreement were reported
for a comprehensive investigation into rater reliability. Overall, the tool was reliable for dogs
assessed either indoors or outdoors with acceptable measures of agreement other than two
prompts that had unsatisfactory measures of inter-rater reliability, ‘Reaction to acoustic startle,
HAI 4’ (sc10) and ‘Reaction to visual startle, HAI 5’ (sc11). Sc10 and sc11 are candidates for
removal pending validity analysis. Subsequent analysis on criterion validity will provide insight
on clinical value by comparing assessment results to blinded veterinarian diagnoses and
comparison between total scores and prompt scores of dogs assessed indoors to those assessed
outdoors, to be discussed in Chapter 3. Based on analysis of reliability and validity, the tool will
be further refined to improve practicality while maintaining validity. The refined tool should
undergo additional reliability and validity studies across shelters and users to ensure accuracy
and replicability.
78
CHAPTER 3: VALIDITY OF A BEHAVIOR ASSESSMENT TO MEASURE COPING BEHAVIOR OF
SHELTER DOGS
3.1: Introduction
An estimated 3.1 to 5 million dogs enter shelters nationwide each year (Pet statistics,
ASPCA; Rowan & Kartel, 2018; Woodruff & Smith, 2020). Shelters play a crucial role in providing
health, safety, and welfare services for homeless animals. However, environmental stressors like
novelty, sensory overload, limited opportunity to socialize, and spatial restriction negatively
impact shelter dog welfare (Hennessy, 2013; Marston & Bennett, 2003; Protopopova, 2016).
Extensive research on this topic has informed management practices including environmental
modifications, enrichment strategies, and behavioral intervention programs (for review see
Taylor & Mills, 2007; Wells, 2004). Behavioral intervention programs may include a combination
of any one or more of the following strategies: operant conditioning, environmental
enrichment, or anxiolytic medication prescribed by veterinarians (Abrams et al., 2020; Corsetti
et al., 2021; Protopopova & Wynne, 2015; Protopopova et al., 2018; Tod et al., 2005).
Once behavior issues are identified, appropriate intervention plans, e.g., positive
reinforcement training, desensitization, and counterconditioning techniques, can be
implemented (for review see Protopopova & Gunter, 2017). However, some dogs continue to
struggle and could benefit from pharmaceutical treatment that would manage underlying
anxiety or physiological factors (Abrams et al., 2020; Ballantyne, 2018; Corsetti et al., 2021;
Gilbert-Gregory et al., 2016; Gruen & Sherman, 2008; Tod et al., 2005).
Shelter veterinarians typically rely on reports from staff and volunteers to assess and
monitor the behavioral health of shelter animals. While communication from caregivers to
shelter veterinarians is crucial, a validated behavior assessment specifically designed to identify
dogs struggling to cope in the shelter environment does not currently exist. While diagnosis and
prescription are within the professional capacity of all practicing veterinarians, the level of
comfort in addressing behavioral issues is variable (Stelow, 2018), further demonstrating the
need for a diagnostic tool that would quantify coping state. Scientific validation of behavior
assessments requires reporting on the tool’s ability to measure the intended outcome (McCall,
1984). Principle components of establishing criterion validity include sensitivity (percentage of
population that is positive and identified as positive), specificity (percentage of the population
79
that is negative and identified as negative), and a clear criterion of the identified behavior
(Akobeng, 2007a; Akobeng, 2007b; Akobeng, 2007c; Patronek & Bradley, 2016). To characterize
sensitivity and specificity, a reference standard, i.e., gold-standard, is required for confirmatory
diagnostics. Current accepted diagnostics for identifying dogs that could benefit from anxiolytic
treatment is veterinarian evaluation. A complete plan for diagnosis includes review of medical
and behavioral history and direct observation (Stelow, 2018). However, shelters frequently do
not have the medical nor behavioral history of the dog, or it is very sparse. Therefore, shelter
veterinarians rely on physical examinations performed upon intake and direct observation of
shelter dogs to diagnose behavior problems. Identification of true positive dogs for maladaptive
coping can thus be established through professional diagnosis by a veterinarian behaviorist.
To define diagnoses, maladaptive coping is descriptive of dogs whose behavioral
symptoms would warrant intervention, whereas adaptive coping categorizes dogs that are at a
low risk for mental deterioration while in-shelter. Adaptive responses to stress encompass a
variety of behavioral responses evolved to effectively increase fitness of the individual while in
the presence of a real or perceived threat. These behaviors have been studied in several species
under the dichotomy of reactive and proactive coping styles which are characterized by
physiological and behavioral differences in response to stressors (Koolhaas et al., 1999). An
individual’s coping style is dependent on several factors such as genetics, past experiences, and
situational context. Dogs that acclimate to the shelter environment habituate to stressful stimuli
and adjust behavior accordingly. However, for the shelter dogs that remain in a state of stress,
behavioral responses fail to eliminate the perception of danger, resulting in a maladaptive
coping response. Within maladaptive coping, dogs may either be diagnosed as anxious-avoidant
or excessive-aroused, which are defined in the following paragraph.
Anxious-avoidant diagnosis encompasses behaviors that correlate with a reactive coping
style often recognized by decreased physical activity. Anxious-avoidant dogs may benefit from
anti-depressant pharmaceutical therapies such as buspirone which attenuates fear and worry
through enhancing serotonin activity in areas of the brain (Stahl, 2021). On the other end of the
spectrum, increased physical activity is correlated with proactive coping styles, which manifests
as excessive arousal under chronic stress. Excessive-aroused dogs may benefit from serotonin
80
antagonist/reuptake inhibitor (SARI) pharmaceutical therapies such as trazodone hydrochloride
by decreasing arousal through antihistamine activity and antagonism of alpha-1 adrenergic
receptors (Abrams et al., 2020; Gilbert-Gregory et al., 2016; Gruen & Sherman, 2008; Gruen et
al., 2014; Kim et al., 2022).
Shelters need to rely on staff to communicate maladaptive coping and behavioral
intervention to administration and veterinarians for timely treatment. However, a validated
behavior assessment specifically designed to identify dogs struggling to cope in the shelter
environment does not currently exist. Therefore, the aim of this study was to evaluate the
criterion validity of a novel diagnostic tool developed to quantify shelter dog coping behavior.
3.1.i: Development of the behavioral assessment tool
The tool consists of a battery of behavioral assessments and was previously analyzed for
inter- and intra-rater reliability in Chapter 2. Criterion validity was based on expert observation
and assessment by a board-certified veterinary behaviorist, which is the industry standard for
behavioral diagnostics in dogs. I hypothesized that low negative scores would correlate with
maladaptive coping-anxious-avoidant dogs (MC-AA), high positive scores with maladaptive
coping-excessive-aroused dogs (MC-EA) and near 0 scores with adaptive coping dogs (AC).
Furthermore, analysis of individual tool components for clinical value will inform refinement of
the assessment for further reliability and validity testing.
3.2: Materials and Methods
3.2.i: Ethical Approval
This study was approved by the Institutional Animal Care and Use Committee of
Michigan State University (IACUC AMEND202100602 / PROTO202100191). Permission for
subject participation was granted by the shelter President/CEO as the legal owner of the dogs
while in shelter using the organization consent form provided by IACUC.
3.2.ii: Inclusion Criteria
Healthy dogs, 12 weeks or older, housed in solitary kennels at Capital Area Humane
Society (Lansing, MI), were eligible for assessment. Dogs were excluded for health reasons
(obesity, cancer, broken bones), if they were younger than 12 weeks of age, if they were group-
housed, if they refused to exit the kennel, or if they were designated to be handled only by staff
81
due to severe aggression. Shelter housing and care was previously described in Chapter 2. A
total of 91 dogs were assessed, 44 of which were female (33 spayed, 11 intact) and 47 of which
were male (29 neutered, 18 intact). There is no evidence that sex is correlated with increased
risk of relinquishment (Patronek et al., 1996) and our sample population generally aligns with
the sex distribution of previous domestic (Arhant & Troxler, 2014; Brown et al., 2013; Salman et
al., 1998) and international shelter studies (Diesel et al., 2010; Shih et al., 2021). It was the
shelter’s policy to spay/neuter all adoptable dogs; however, surgeries were performed on select
days of the week which resulted in 31.87% of subjects intact at time of assessment.
Breed history was largely unknown (described in detail in Chapter 2) with a mean weight
of 42.07 ± 18.08 lbs. Age of the dogs, as estimated by shelter clinicians, ranged from 0y 3m 9d
to 10y 9m 1d with a mean age of 2.32 ± 2.41 years.
3.2.iii: Behavior Assessment
Behavioral assessments took place from June 2022 through October 2022, a minimum of
20 hours after intake. Eligible dogs were assessed once (n = 91). Enrollment protocol was
previously described in Chapter 2. Assessments were recorded using GoPro Hero7 Black video
recording devices from two perspectives for intra-rater reliability and validity testing. In brief,
the behavior assessment consisted of eight categories, i.e., human-animal interactions (HAI’s).
All assessments were performed by the same handler in the following order, In-Kennel: Begin
(HAI 1), Play (HAI 2), Settle (HAI 3), Acoustic Startle (HAI 4), Visual Startle (HAI 5), Distraction
(HAI 6), and In-Kennel: Return (HAI 7). The final category, Activity (HAI 8), was assessed during
transitions between the kennel and assessment area. Assessment protocol was described in
detail in Chapter 2.
To determine whether the area for assessment had an impact on behavioral scores,
eligible dogs were randomly assigned to either an indoor (n = 43) or outdoor (n = 48)
assessment. Indoor assessments took place in the shelter’s educational room (594 S.F.) while
outdoor assessments occurred in a fenced enclosure (545 S.F.) on the back area of the shelter
property. Assessment location was kept consistent through the day but varied per day of the
week based on room availability and inclement weather. The diagnostic criterion standard was
established by identification of the dogs as either adaptive coping (AC), maladaptive coping
82
anxious-avoidant (MC-AA) or maladaptive coping excessive-aroused (MC-EA) by a board-
certified veterinarian behaviorist. This individual was blinded to assessment results and
performed diagnoses retrospectively using the assessment video recordings. In addition, dogs
that were diagnosed as maladaptive coping were further rated by the board-certified
veterinarian behaviorist as mild, moderate, or severe based on the observed severity of
behavioral symptoms.
3.2.ii: Sample Population
Power analysis was performed for each goal: a) assessing the ability of coping score to
differentiate between diagnoses and b) determining whether assessment area impacted score.
Target enrollment was established from the higher required subject enrollment for a minimum
of 80% power with an estimated standard deviation of 0.50. Based on the expected ratio of
observed diagnoses by a board-certified veterinarian behaviorist knowledgeable in shelter
medicine of 40% AC / 40% MC-EA / 20% MC-AA, to assess score validity across diagnoses, nine
subjects were required (SAS PROC POWER for one-way ANOVA with three sample populations,
i.e., diagnoses). To determine if the assessment area impacts score, 46 dogs per assessment
area would be required (SAS GLIMMIX using hypothetical data for two treatments, i.e., indoor
or outdoor assessment). Therefore, the target enrollment was set at 92 dogs and duration of
the study was influenced by number of dogs assessed in addition to the researchers’ schedules.
Age, sex, and size were categorized for reporting on the characteristics of the sample
population (Table 3.1 and Figure 3.1). Age categories were adapted from the AAHA guidelines
for canine life stages based on the approximation of when dogs experience cessation of rapid
growth (9 months), social and physical maturation (3 years), and the last 25% of estimated
lifespan (8 years) (Canine Life Stage Definitions, AAHA). Dogs younger than 3 months were
excluded from the study so therefore dogs less than 9 months were labeled as juveniles, dogs
from 9 months up to 3 years as adolescents, dogs from 3 years up to 8 years as adults, and dogs
8 years or older as geriatric. Size categories, small (less than 24.0 lbs.), medium (24.0 lbs. to 44.0
lbs.), and large (more than 44.0 lbs.) were based on the shelter’s standards; however, juveniles
were categorized by their current weight and not that of adult size.
83
Table 3.1. Characteristics of sample population (n = 91). Sample population by characteristics
age, sex, and size (n = 91). Dogs were categorized based on their age as juvenile, adolescent,
adult, or geriatric. Dogs less than 3 months were ineligible for assessment. The sex of the dogs
was either female-spayed (F-S), female-intact (F-I), male-neutered (M-N), or male-intact (M-I).
Size was based on current weight and categorized as small, medium, or large.
Characteristic
Number
%
Age
Juvenile (< 9 mos.)
Adolescent (≥ 9 mos. and < 3 yrs.)
Adult (≥ 3 yrs. and < 8 yrs.)
Geriatric (≥ 8 yrs.)
Sex
Female-Spayed
Female-Intact
Total Female
Male-Neutered
Male-Intact
Total Male
Size
Small (< 24.0 lbs.)
Medium (≥ 24.0 lbs. and ≤ 44.0 lbs.)
Large (> 44.0 lbs.)
23
45
18
5
33
11
44
29
18
47
17
34
40
25.27
49.45
19.78
5.49
36.26
12.09
48.35
31.87
19.78
51.65
18.68
37.36
43.96
84
Figure 3.1. Sample population characteristics clustered by age and size (n = 91). Sample
population characteristics for factors age and size (n = 91). Bars show the number of dogs in
each size category, small, medium, and large, clustered by age, juvenile, adolescent, adult, and
geriatric. The number of dogs per descriptive combination of age and size (n) is shown above
bars.
Juvenile (< 9 mos.)
Adolescent (≥ 9 mos. and < 3 yrs.)
Adult (≥ 3 yrs. and < 8 yrs.)
Geriatric (≥ 8 yrs.)
Small (< 24.0 lbs.)
Medium (≥ 24.0 lbs. and ≤ 44.0 lbs.)
Large (> 44.0 lbs.)
3.2.iii: Validity Analysis
For validity testing, assessment results from Rater A (first author) were compared to
blinded veterinarian diagnoses and statistical analyses were calculated in R version 4.1.3. Total
coping score and prompt scores (sc01-sc15) were individually analyzed for validity. Dogs with
missing prompt scores were excluded from total score analyses. For analysis of each prompt
score, all assessments with complete data were included for that prompt score (Table 3.2). Total
coping score was assumed normally distributed by comparison of the histogram, residual plot,
and Q-Q plot of standardized residuals. However, individual prompt scores appeared non-
normally distributed and validity tests for prompt scores were selected based on robustness to
non-normality.
85
Table 3.2. Total dogs used for validity analysis. Number of dogs included in validity analysis for
indoor, outdoor, and pooled indoor + outdoor models for each factor (individual scored
prompts, unscored prompts, and total coping score). See Table 2.1 for abbreviations. Total
coping score was analyzed for four models (Full Model, Reduced Model, Refined Model A, and
Refined Model B; described in results), of which only the Full Model was analyzed for both
indoor and outdoor assessments in addition to pooled assessments (indoor + outdoor). All
assessments used for validity analysis were from Rater A and dogs with one or more missing
scored prompt were excluded from total coping score analysis. Therefore, of the 91 dogs that
were assessed, 84 were analyzed for the Full Model and Reduced Model while 86 were included
in the Refined Model A and Refined Model B. Indoor and outdoor assessments were pooled for
validity analysis on unscored prompts.
Factor
Outdoor (n)
Indoor (n)
Indoor +
Scored prompts
Outdoor (n)
sc01
sc02
sc03
sc04
sc05
sc06
sc07
sc08
sc09
sc10
sc11
sc12
sc13
sc14
sc15
Total coping score
Full Model (sc01-sc15)
Reduced Model (sc01-sc09, sc11-sc15)
Refined Model A (sc04-sc06, sc08, sc09, sc13, sc14)
Refined Model B (sc01, sc03-sc06, sc08, sc09, sc13, sc14)
Unscored prompts
usc01 – usc07
pilo1 – pilo8
86
43
43
43
43
43
43
43
42
43
43
42
43
43
43
43
41
.
.
.
.
.
48
48
47
47
48
47
48
47
48
48
48
48
48
47
46
43
.
.
.
.
.
91
91
90
90
91
90
91
89
91
91
90
91
91
90
89
84
84
86
86
91
91
To test our prediction that score correlates with diagnosis, we used a linear mixed model
(ANOVA) with a Tukey adjustment for multiple comparisons, 95% confidence interval, and
significance of α ≤ 0.05, to consider the total score as predicted by diagnosis using all 15 prompt
scores as the full model. Significant p-values indicate the tool can identify a difference in total
score between the diagnostic categories (R code: emmeans). Significant p-values on contrasts
AC:MC-AA, MC-EA:MC-AA, or MC-EA:AC would indicate the tool is able to identify a difference
in total score range between the respective diagnostic categories. The decision to pool indoor
and outdoor assessments was based on either linear regression (R code: lm) if the factor was
assumed normally distributed (total coping score) or the Wilcoxon rank sum test with a
continuity correction (R code: wilcox.test) if the factor was non-normally distributed (sc01-
sc15). Assessments were pooled for additional statistical testing (indoor + outdoor) if p > 0.05,
indicating that assessment area had no effect on factor results.
To propose ways in which the tool could be refined to improve practicality while
maintaining validity, I performed cumulative regression analysis using each prompt score as a
predictor of diagnosis for indoor and outdoor dogs until all coefficients had a p-value ≤ 0.25 (R
code: clm + stepAIC). The refined model was then analyzed for validity using the same strategy
as with the full model. Additionally, to analyze the clinical value of unscored prompts, we
performed a Chi-squared test on unscored responses (usc01-usc05 and pilo1-pilo8) with
diagnostic category (AC, MC-AA, or MC-EA) as the outcome variable (R code: chisq.test). To
correct for multiple comparisons, a Bonferroni correction for three variables was applied (α =
0.05 / 3) so that p-values ≤ 0.0167 indicate a statistically significant correlation between the
behavior and diagnostic outcome. However, because the tool is intended to capture behavioral
nuances, to avoid becoming overly constrictive, prompts with p-values ≤ 0.10 are considered
clinically relevant. For two unscored prompts (usc01 and usc03), the assumption of reasonably
large, expected cell count (≥ 5) was violated which would cause inaccurate chi square
approximations. Consequently, for ‘Location after approach, HAI 1’ and ‘Location with open
door, HAI 1’, clinical relevance was analyzed using Fisher’s exact test. Unscored prompts that
were not clinically relevant were candidates for removal.
87
3.3: Results
3.3.i: Diagnosis Distribution
Of all dogs (n = 91), 40.66% were diagnosed MC-AA (n = 37), 21.98% MC-EA (n = 20), and
37.36% AC (n = 34). Further diagnostics of mild, moderate, or severe for maladaptive coping are
delineated in Table 3.3 and Figure 3.2. Assessments with more than one missing prompt score
were excluded from total score analyses (n = 7). Those excluded from total score analyses
consisted of the following diagnoses: MC-AA mild (outdoors = 1), MC-AA moderate (outdoors =
1), MC-AA severe (outdoors = 1), AC (indoors = 1; outdoors = 1), MC-EA mild (outdoors = 1), and
MC-EA moderate (indoors = 1).
Table 3.3. Diagnosis distribution. Dogs were diagnosed by a veterinarian behaviorist blinded to
assessment scores as one of three possible outcomes, Maladaptive Coping-Anxious-Avoidant
(MC-AA), Maladaptive Coping-Excessive-Aroused (MC-EA) or Adaptive Coping (AC). (a)
Distribution of diagnoses for total sample population (n = 91). MC-AA and MC-EA dogs were
further categorized by severity of behavioral symptoms (mild, moderate, severe). (b) Severity
distribution of dogs diagnosed as maladaptive coping (MC-AA + MC-EA) (n = 57).
(a) Diagnosis distribution of sample population (n = 91)
Diagnosis
MC-AA
Total MC-AA
MC-EA
Total MC-EA
mild
moderate
severe
mild
moderate
severe
n
18
13
6
37
11
7
2
20
%
19.78
14.29
06.59
40.66
12.09
7.69
2.20
21.98
Total AC
34
(b) Distribution of severity across maladaptive coping diagnoses (n = 57)
37.36
Diagnosis
MC (AA + EA)
Total MC (AA+EA)
mild
moderate
severe
n
29
20
8
57
%
50.877
35.088
14.035
100.0
88
Figure 3.2. Diagnosis distribution across all dogs (n = 91). Diagnoses distribution across sample
population (n = 91) by diagnosis (inner ring), severity (middle ring) and assessment area (outer
ring). Dogs were diagnosed by a board-certified veterinarian behaviorist blinded to assessment
scores using assessment videos as adaptive coping (AC) or maladaptive coping (MC) anxious-
avoidant (AA) or excessive-aroused (EA). One full circle along the perimeter of each ring
represents the sample population (n = 91) with factors shown proportionate to number of dogs.
For example, three dogs assessed indoors (outer ring), had severe symptoms (middle ring) for
MC-AA (inner ring). AC dogs were not rated for severity. Number of dogs (n) in each category is
shown in the outer ring. 37.36% were AC, 40.66% were MC-AA, 21.98% were MC-EA, with
50.877% of MC dogs exhibiting mild symptoms, 35.088% moderate, and 14.035% severe. MC-
EA/severe was not diagnosed for any of the dogs assessed outdoors.
89
To evaluate if canine characteristics affect diagnosis, the variables sex, size, and age,
were fitted to a multivariate ANOVA model (R code: clm). Since not all factor combinations were
represented in our sample population, neither 3-way nor 2-way interactions could be
calculated. Therefore, the model does not account for interactions. Diagnosis outcome (MC-AA,
AC, MC-EA) and individual factors (sex, size, age) were categorical. Sex consisted of female-
spayed (F-S), female-intact (F-I), male-neutered (M-N), male-intact (M-I); size consisted of small
(< 24.0 lbs.), medium (≥ 24.0 lbs. & ≤ 44.0 lbs.), large (> 44.0 lbs.); and age consisted of juvenile
(< 9 mos.), adolescent (≥ 9 mos. & < 3 yrs.), adult (≥ 3 yrs. & < 8 yrs.), geriatric (≥ 8 yrs.). No
significant differences were found in diagnoses across groups (sex p = 0.129; size p = 0.178; age
p = 0.408), indicating individual characteristics had no effect on diagnosis. However, there was a
significant difference found in the individual sex coefficient for F-S (p = 0.039) and a near
significant difference for the individual sex coefficient M-N (p = 0.054). Further pairwise
comparison on sex failed to find evidence of a significant difference between contrasts averaged
over the levels of age and size with a C.I. = 0.95 and a Tukey adjustment for comparing a family
of four estimates (F-S, F-I, M-N, and M-I), leading to the conclusion that dog characteristics did
not impact diagnoses.
3.3.ii: Total Score Distribution
Observed total scores (Figure 3.3) ranged from –18 to +10 for dogs assessed indoors (n =
41) and –9 to +11.5 for dogs assessed outdoors (n = 43). The median score was zero for indoor
assessments and –0.5 for outdoor assessments. The scores shared by most dogs (mode) were –
3, 0, and +1 for indoor assessments (5 dogs each) and +4 for outdoor assessments (5 dogs).
90
Figure 3.3. Histogram of total coping score for sample population. The histogram of total
coping score for the sample population excluding dogs with at least one missing prompt score (n
= 84). Total coping score was assumed normally distributed. (a) The frequency (number of dogs)
for total coping score for dogs assessed indoors (n = 41). The observed score range was -18 to
+10, median score was 0, and mode was shared between -3, 0, and +1 with five dogs each. (b)
The frequency (number of dogs) of total coping score for dogs assessed outdoors (n = 43). The
observed score range was -9 to +11.5, median score was -0.5, and mode was +4 with five dogs.
(a) Frequency of total scores for dogs assessed indoors (n = 41)
(b) Frequency of total scores for dogs assessed outdoors (n = 43)
91
3.3.iii: Impact of Assessment Area on Total Score
There was no significant difference in total scores when comparing indoor assessments
versus outdoors (p = 0.842). However, there was a notable distinction for sc10 (Reaction to
acoustic startle, HAI 4), which was significantly different between indoor and outdoor dogs (W =
1243, p = 0.025) (Table 3.4). Therefore, total score was analyzed both with and without sc10 for
pooled data (indoor + outdoor), as well as analyzing indoor and outdoor datasets separately
using all 15 scored prompts, allowing us to compare validity across scenarios.
Table 3.4. Prompt score comparison of indoor to outdoor assessments. Scored prompts were
analyzed to compare indoor to outdoor assessments using the Wilcoxon-rank sum test.
‘Reaction to acoustic startle’ (sc10) was statistically significant, indicating that assessment area
may impact dog behavior for sc10. There was no evidence that any of the remaining prompt
scores were affected by the area.
Abbrev.
Scored prompt
W
p
1091
1069
10004.5
898
946
1126
984
1111.5
981.5
1243
930
942.5
1023.5
991
969
0.411
0.410
0.935
0.260
0.442
0.252
0.137
0.258
0.558
0.025*
0.431
0.431
0.937
0.843
0.780
sc01
sc02
sc03
sc04
sc05
sc06
sc07
sc08
sc09
sc10
sc11
sc12
sc13
sc14
sc15
Presentation with closed door, HAI 1
Ability to take treat, HAI 1
Presentation with open door, HAI 1
Latency to exit, HAI 1
Initial reaction to play, HAI 2
Ease of putting away toys, HAI 2
Was play ended early, HAI 2
Response to treat, HAI 3
Activity, HAI 3
Reaction to acoustic startle, HAI 4
Reaction to visual startle, HAI 5
Overall body language, HAI 6
Can you obtain the dog's attention, HAI 6
Approach to kennel entry, HAI 7
In-kennel behavior post-return, HAI 7
*Significant at α ≤ 0.05
92
3.3.iv: Validity
Indoor assessments
For the full model consisting of all 15 scored prompts (Figure 3.4), the estimated
marginal mean total score for dogs assessed indoors (n = 41) was -3.00 ± 1.280 for MC-AA dogs
(n = 17), 1.367 ± 1.362 for AC dogs (n = 15), and 1.444 ± 1.759 for MC-EA dogs (n = 9). There was
a numeric difference between groups with negative scores correlated with MC-AA and positive
scores correlated with AC and MC-EA; however, there was no statistically significant difference
between marginal means. In contrasting estimated range of diagnostic groups, there was a
trend suggesting total score range of MC-AA dogs differed from AC dogs (p = 0.063) but no
significant difference from MC-EA dogs (p = 0.116) nor AC from MC-EA (p = 0.999). Three dogs
(MC-AA/severe = 1, AC = 1, MC-EA/mild = 1) were identified as potential influential outliers
based on a Cook’s distance greater than 0.098 (4/n). Further, the only two MC-EA/severe dogs
assessed were observationally close to the 0.098 cutoff, each rated with a negative total coping
score (-6 and –3.5). Despite being statistical outliers, from a biological perspective each
diagnosis and score is relevant since variability in shelter dog behavior is expected. Therefore,
the outliers were not excluded from the model and results reflect their inclusion.
93
Figure 3.4. Full Model (15 scored prompts) for indoor assessments (n = 41). Estimated marginal
mean of total coping score ±SE for each diagnosis using the Full Model (sum of sc01 through
sc15) for dogs assessed indoors (n = 41). At statistical significance (p ≤ 0.05), there was no
evidence of a difference between total coping score of diagnoses. (a) There is a numerical
difference of total coping score between diagnoses, with -3.00 ± 1.280 estimated for MC-AA
dogs, 1.367 ± 1.362 for AC dogs, and 1.444 ± 1.759 for MC-EA dogs. (b) By severity, total coping
score increases as expected from MC-AA/severe on the left to MC-EA/moderate on the right.
However, the negative score estimated for MC-EA/severe was unexpected. The dogs that were
diagnosed MC-EA/severe (n = 2), were statistical outliers (Cook’s distance greater than 4/n) but
remained in the model as biologically relevant since variability in shelter dog behavior is
expected.
(a) Estimated marginal mean total coping score per diagnosis
94
Figure 3.4 (cont’d)
(b) Estimated marginal mean total coping score per severity of symptoms
Outdoor assessments
For the full model consisting of all 15 scored prompts (Figure 3.5), the estimated
marginal mean total score for dogs assessed outdoors (n = 43) was -4.176 ± 0.898 for MC-AA
dogs (n = 17), 1.647 ± 0.898 for AC dogs (n = 17), and 3.833 ± 1.234 for MC-EA dogs (n = 9). Here
there is a statistically significant difference of MC-AA marginal mean total score from AC and
MC-EA (p ≤ 0.05). Statistical results on marginal mean differences are reflected in the significant
difference in total score range of MC-AA dogs to AC (p = 0.0001) and MC-EA (p < 0.0001). There
was no significant difference in range from AC dogs to MC-EA (p = 0.334), but numerically MC-
EA dogs have a higher marginal mean total score compared to AC dogs.
95
Figure 3.5. Full Model (15 scored prompts) for outdoor assessments (n = 43). Estimated
marginal mean of total coping score ±SE for each diagnosis using the Full Model (sum of sc01
through sc15) for dogs assessed outdoors (n = 43). Pairwise significant differences (p ≤ 0.05)
between the groups are shown by color. (a) The est. mean for MC-AA dogs is different than that
of AC and MC-EA dogs at statistical significance. There is a numerical difference of total coping
score between diagnoses, with -4.176 ± 0.898 estimated for MC-AA dogs, 1.647 ± 0.898 for AC
dogs, and 3.833 ± 1.234 for MC-EA dogs. (b) By severity, total coping score increases as
expected from MC-AA/severe on the left to MC-EA/moderate on the right. While est. mean
drops slightly from MC-AA/moderate to MC-AA/mild and from MC-EA/mild to MC-
EA/moderate, variability in shelter dog behavior is expected and results are biologically
relevant. There were no dogs assessed outdoors that were diagnosed MC-EA/severe.
(a) Estimated marginal mean total coping score per diagnosis
96
Figure 3.5 (cont’d)
(b) Estimated marginal mean total coping score per severity of symptoms
Indoor + outdoor assessments (pooled)
When pooling both indoor and outdoor assessments (n = 84), for the full model
consisting of all 15 scored prompts (Figure 3.6), the estimated marginal mean total score was -
3.588 ± 0.773 for MC-AA dogs (n = 34), 1.516 ± 0.796 for AC dogs (n = 32) and 2.639 ± 1.062 for
MC-EA dogs (n = 18). In agreement with the results from the outdoor assessments, there was a
significant difference in marginal mean total score for MC-AA dogs from AC and MC-EA (p ≤
0.05) and a significant difference in score range of MC-AA dogs to AC (p < 0.0001) and MC-EA (p
< 0.0001).
97
Figure 3.6. Full Model (15 scored prompts) for indoor + outdoor assessments (n = 84).
Estimated marginal mean of total coping score ±SE for each diagnosis using the Full Model (sum
of sc01 through sc15) for dogs assessed indoors and outdoors (n = 84). Pairwise significant
differences (p ≤ 0.05) between the groups are shown by color. (a) The est. mean for MC-AA dogs
is different than that of AC and MC-EA dogs at statistical significance. There is a numerical
difference of total coping score between diagnoses, with -3.588 ± 0.773 estimated for MC-AA
dogs, 1.516 ± 0.796 for AC dogs, and 2.639 ± 1.062 for MC-EA dogs. (b) By severity, total coping
score increases as expected from MC-AA/severe on the left to MC-EA/moderate on the right,
with a slight drop from MC-EA/mild to MC-EA/severe; variability in shelter dog behavior is
expected and results are biologically relevant. Estimated mean for MC-EA/severe is
unexpectedly low, which is reflective of the statistical outliers diagnosed MC-EA/severe (n = 2).
(a) Estimated marginal mean total coping score per diagnosis
98
Figure 3.6 (cont’d)
(b) Estimated marginal mean total coping score per severity of symptoms
To account for the difference in scores across assessment areas for sc10, a reduced
model consisting of 14 scored prompts with the elimination of the acoustic startle was analyzed
(Figure 3.7). Assuming normal based on the histogram of total score, residual plot, and Q-Q plot
of standardized residuals, the reduced model’s ability to predict diagnosis was analyzed using a
linear mixed model (ANOVA) with a Tukey adjustment. Statistical results agreed with the
findings from the full model, corroborating validity. MC-AA dogs had a significantly different
marginal mean total score (-4.43 ± 0.751) from AC (1.23 ± 0.774) and MC-EA (1.97 ± 1.032). A
significant range difference was seen in MC-AA dogs from AC (p < 0.0001) and MC-EA (p <
0.001) with no significant difference between AC and MC-EA (p = 0.8354).
99
Figure 3.7. Reduced Model (14 scored prompts) for indoor + outdoor assessments (n = 84).
Estimated marginal mean of total coping score ±SE for each diagnosis using the Reduced Model
(sum of sc01 through sc15, excluding sc10) for dogs assessed indoors and outdoors (n = 84).
Pairwise significant differences (p ≤ 0.05) between the groups are shown by color. (a) The est.
mean for MC-AA dogs is different than that of AC and MC-EA dogs at statistical significance.
There is a numerical difference of total coping score between diagnoses, with -4.43 ± 0.751
estimated for MC-AA dogs, 1.23 ± 0.774 for AC dogs, and 1.97 ± 1.032 for MC-EA dogs. (b) By
severity, total coping score increases as expected from MC-AA/severe on the left to MC-
EA/moderate on the right. Estimated mean for MC-EA/severe is unexpectedly low, which is
reflective of the statistical outliers diagnosed MC-EA/severe (n = 2).
(a) Estimated marginal mean total coping score per diagnosis
100
Figure 3.7 (cont’d)
(b) Estimated marginal mean total coping score per severity of symptoms
The statistical difference between estimated marginal mean total score of MC-AA dogs
to AC and MC-EA across the full model (outdoor + indoor) as well as the reduced model
corroborates validity results for total coping score.
3.3.v: Refinement
During the refinement process, our goal was to reduce the number of components
within the tool while also maintaining validity. Candidates for removal were identified through
cumulative regression analysis on total score for all dogs (indoor + outdoor assessments) until
all scored prompts that remained in the model had a p-value ≤ 0.25. This resulted in a drastically
reduced number of scored prompts from 15 to three (sc04, sc08, and sc14). However, a tool
used for behavioral diagnostics based on three scores would not provide enough context for
clinical interpretation of the behaviors, rendering the model clinically irrelevant. Therefore,
analysis was performed for indoor and outdoor assessments separately. This reduced indoor
101
scores to sc04, sc05, sc08, sc13, and sc14 while outdoor scores were reduced to sc04, sc06,
sc08, sc09, and sc14. After accounting for duplicates, the refined version of the tool based on
statistical significance (Model A) consisted of ‘Latency to exit, HAI 1’ (sc04), ‘Initial reaction to
play, HAI 2’ (sc05), ‘Ease of putting away toys, HAI 2’ (sc06), ‘Response to treat, HAI 3’ (sc08),
‘Activity, HAI 3’ (sc09), ‘Ability to obtain the dog’s attention, HAI 6’ (sc13) and ‘Approach to
kennel entry, HAI 7’ (sc14).
Upon discussion with the veterinarian behaviorist that blindly diagnosed the dogs, two
scored prompts that were eliminated from Model A were clinically valuable, ‘Presentation with
closed door, HAI 1’ (sc01) and ‘Presentation with open door, HAI 1’ (sc03). Therefore, two
refined models were analyzed for validity, Model A and Model B, which in addition to sc04,
sc05, sc06, sc08, sc09, sc13, and sc14 included sc01 and sc03, and results were compared to the
full and reduced models. Both Model A and Model B were assumed normal based on
histograms of total score, residual plots, and Q-Q plots of standardized residuals.
Results for Model A (Figure 3.8) and Model B (Figure 3.9) were in concordance with the
Full and Reduced Models (indoor + outdoor), indicating that validity was maintained during
proposed refinement (Table 3.5). There was a statistically significant difference of MC-AA
marginal mean total score from AC and MC-EA (p ≤ 0.05) in Model A and Model B with a
significant difference in total score range of MC-AA dogs to AC (Model A, p = 0.0015; Model B, p
= 0.004) and MC-EA (Model A, p = 0.0001; Model B, p < 0.0001). However, there was no
significant difference in range from AC dogs to MC-EA (Model A, p = 0.3341; Model B, p <
0.325), but numerically there was a higher marginal mean total score for MC-EA dogs compared
to AC dogs. Validity remained consistent across Model A and B, justifying the professional
decision to include sc01 and sc03 in the assessment in addition to the statistically significant
scores.
102
Figure 3.8. Refined Model A (7 scored prompts) for indoor + outdoor assessments (n = 86).
Estimated marginal mean of total coping score ±SE for each diagnosis using the Refined Model A
(sum of sc04, sc05, sc06, sc08, sc09, sc13, sc14) for dogs assessed indoors and outdoors (n =
86). Pairwise significant differences (p ≤ 0.05) between the groups are shown by color. (a) The
est. mean for MC-AA dogs is different than that of AC and MC-EA dogs at statistical significance.
There is a numerical difference of total coping score between diagnoses, with -2.89 ± 0.685
estimated for MC-AA dogs, 0.727 ± 0.706 for AC dogs, and 2.417 ± 0.955 for MC-EA dogs. (b) By
severity, total coping score increases as expected from MC-AA/severe on the left to MC-
EA/moderate on the right. Estimated mean for MC-EA/severe is unexpectedly low, which is
reflective of the statistical outliers diagnosed MC-EA/severe (n = 2).
(a) Estimated marginal mean total coping score per diagnosis
103
Figure 3.8 (cont’d)
(b) Estimated marginal mean total coping score per severity of symptoms
104
Figure 3.9. Refined Model B (9 scored prompts) for indoor + outdoor assessments. Estimated
marginal mean of total coping score ±SE for each diagnosis using the Refined Model B (sum of
sc01, sc03, sc04, sc05, sc06, sc08, sc09, sc13, sc14) for dogs assessed indoors and outdoors (n =
86). Pairwise significant differences (p ≤ 0.05) between the groups are shown by color. (a) The
est. mean for MC-AA dogs is different than that of AC and MC-EA dogs at statistical significance.
There is a numerical difference of total coping score between diagnoses, with -3.34 ± 0.710
estimated for MC-AA dogs, 0.727 ± 0.731 for AC dogs, and 2.500 ± 0.990 for MC-EA dogs. (b) By
severity, total coping score increases as expected from MC-AA/severe on the left to MC-
EA/moderate on the right. Estimated mean for MC-EA/severe is unexpectedly low, which is
reflective of the statistical outliers diagnosed MC-EA/severe (n = 2).
(a) Estimated marginal mean total coping score per diagnoses (n = 86)
105
Figure 3.9 (cont’d)
(b) Estimated marginal mean total coping score per severity of symptoms (n = 86)
106
Table 3.5. Total score as predicted by diagnosis, linear mixed model. Total coping score was
analyzed for validity using a linear mixed model to determine the total score as predicted by
diagnosis (diag.) for four models (Full Model, Reduced Model, Refined Model A, Refined Model
B) for pooled indoor + outdoor assessments. Additionally, the Full Model was analyzed for
indoor and outdoor assessments separately. Pairwise significant differences (p ≤ 0.05) between
estimated marginal means (emmeans) are designated within each model by superscripts. For all
models, the emmean for MC-AA is significantly different than those for AC and MC-EA dogs
except for the indoor assessments using the Full Model. Across all models, there is a numeric
difference between negative emmeans for MC-AA dogs and positive emmeans for MC-EA dogs
with the emmeans for AC dogs falling between the two. There is evidence that the score range
for MC-AA dogs is significantly different than that of MC-EA and AC dogs (contrast p ≤ 0.05) for
all models except indoor assessments using the Full Model.
Full Model (15 scored prompts)
Outdoor
(n=43)
Indoor
(n=41)
Pooled
(n=84)
Reduced
Model
(14 scored
prompts)
Pooled
(n=84)
Refined
Model A
(7 scored
prompts)
Pooled
(n=86)
Refined
Model B
(9 scored
prompts)
Pooled
(n=86)
Diag.
MC-AA
AC
MC-EA
Estimated marginal mean ±SE
-3.00a ±1.280
1.367a ±1.362
1.444a ±1.759
-4.176a ±0.898
1.647b ±0.898
3.833b ±1.234
p-value for contrasts
-3.588a ±0.773
1.516b ±0.796
2.639b ±1.062
-4.43a ±0.751
1.23b ±0.774
1.97b ±1.032
-2.89a ±0.685
0.727b ±0.706
2.417b ±0.955
-3.34a ±0.710
0.727b ±0.731
2.500b ±0.990
0.116
0.063
AC :
MC-AA
MC-EA :
MC-AA
MC-EA :
AC
*Significant at α = 0.05
0.999
0.0001*
<0.0001*
<0.0001*
0.0015*
0.004*
<0.0001*
<0.0001*
<0.0001*
0.0001*
<0.0001*
0.334
0.676
0.8354
0.3341
0.325
Superscripts denote statistically significant differences between marginal means within each model
Among the scored prompts included in the Refined Model B, several responses were not
observed (n = 9) (Table 3.6). Of the nine responses that were never observed, two would have
caused the analysis to end (sc04(d) and sc04€) and therefore would have zero dogs included in
the analysis. The observed mode matched the predicted response for 20 responses, while 10
responses had the most dogs diagnosed in a category other than expected.
107
Table 3.6. Contingency table for Refined Model B (n = 86). The response contingency table for
all scored prompts included in the Refined Model B shows the number of dogs per diagnosis
that received each score. The highest proportion of dogs per response (row) is illustrated with
bold font. Positive values are predicted to have the largest proportion of MC-EA dogs, negative
values are predicted to have the highest proportion of MC-AA dogs, and zero values are
predicted to have the highest proportion of AC dogs. Responses with the highest proportion of
dogs other than what was expected are shown by bold red font for diagnoses columns (n).
Prompt
Abbrev. Response
MC-AA
Score
(n)
AC
(n)
MC- EA
(n)
Presentation
sc01(a)
Friendly/Attentive/Neutral
with closed
sc01(b)
Dissociated (fearful/stressed)
door,
HAI 1
sc01(c)
Aware (fearful/stressed)
sc01(d)
Hyper-active
sc01(e)
Aggression
sc01(f)
Hidden from view
Presentation
sc03(a)
Friendly/Attentive/Neutral
with open
sc03(b)
Dissociated (fearful/stressed)
door,
HAI 1
sc03(c)
Aware (fearful/stressed)
sc03(d)
Hyper-active
sc03(e)
Aggression
sc03(f)
Hidden from view
Latency to
sc04(a)
Bolts
exit,
HAI 1
sc04(b)
Efficient
sc04(c)
Delayed
sc04(d)
Refuses to exit
sc04(e)
Unsafe to allow exit
Initial
sc05(a)
Engages in play
reaction to
sc05(b)
Approach - does not engage
play,
HAI 2
sc05(c)
No approach - does not engage
108
0
-3
-1.5
1.5
3
NULL
0
-3
-1.5
1.5
3
NULL
3
0
-3
END
END
0
3
-3
29
34
17
0
7
0
1
0
0
0
0
0
0
0
1
2
0
0
29
31
20
0
8
0
0
0
0
22
15
0
0
5
28
4
0
1
1
0
0
2
26
5
0
0
21
12
1
0
0
0
0
0
4
15
1
0
0
12
8
0
Table 3.6 (cont’d)
Ease of
sc06(a)
Easy - not engaged with toy
putting away
sc06(b)
Easy with trade for treat
toys,
HAI 2
sc06(c)
Difficult - multiple trades required
Response to
sc08(a)
Consumes while standing or moving
treat,
HAI 3
sc08(b)
Consumes while lying/sitting or plays
with treat
sc08(c)
Ignores treat/brief interest or holds
treat in mouth without consuming
Activity,
sc09(a)
Neutral - stationary
HAI 3
sc09(b)
Neutral - active
sc09(c)
Stressed - stationary
sc09(d)
Stressed - active
Ability to
sc03(a)
No - oriented toward stimulus
obtain dog’s
sc03(b)
No - retreat from stimulus
attention,
sc03(c)
No - uninterested
HAI 6
sc13(d)
Yes
Approach to
sc14(a)
Actively pulling towards entry
kennel entry,
sc14(b)
Actively pulling away from entry
HAI 7
sc14(c)
Actively pulling/darting in multiple
directions
sc14(d)
Requires encouragement
sc14(e)
Cooperative
-3
0
3
3
0
-3
0
0
-3
3
3
-3
0
0
-3
1.5
3
0
0
31
5
0
3
5
20
13
1
12
12
10
8
2
5
3
29
10
10
13
16
3
5
9
1
0
27
0
4
3
11
18
6
24
1
3
10
1
0
23
0
0
4
4
26
5
10
1
4
5
2
0
13
0
5
8
4
3
Several prompts were unscored (usc01-usc07, pilo1-pilo8) and used to gather
exploratory data in anticipation that the information may prove beneficial in maladaptive coping
diagnosis (Table 3.7). Many of the behavioral responses (n = 28) were never observed and
excluded from analysis (X2 = N/A). Responses from three of the five non-mutually exclusive
prompts (Overall body language, HAI 1 (usc02), Overall body language, HAI 2 (usc04), and
109
Miscellaneous behavior, HAI 3 (usc05)) were found to be correlated with at least one of the
three diagnoses at a clinically relevant level (p ≤ 0.10). Three of the responses for these prompts
were statistically significant: Image 14, Overall body language, HAI 1 (usc02(n) p = 9.16x10-5),
Image 6, Overall body language, HAI 2 (usc04(f) p = 8.92x10-3), and Image 8, Overall body
language, HAI 2 (usc04(h) p = 4.68x10-3). An additional five unscored responses were above
statistical significance but within clinical value (0.0167 < p ≤ 0.10): Image 11, Overall body
language, HAI 1 (usc02(k) p = 0.467), Image 18, Overall body language, HAI 1 (usc02(r) p =
0.033), Image 10, Overall body language, HAI 2 (usc04(j) p = 0.020), Image 11, Overall body
language, HAI 2 (usc04(k) p = 0.089), and None, Miscellaneous behavior, HAI 3 (usc05(f) p =
0.066). There was no evidence at either a statistical significance or clinical relevance the
following prompts are biologically relevant: In-kennel miscellaneous behavior, HAI 8 (usc06),
Out-of-kennel miscellaneous behavior (usc07), Location after approach, HAI 1 (usc01), Location
with open door, HAI 1 (usc03), and Piloerection, HAI 1-8 (pilo1-pilo8).
Table 3.7. Unscored response correlation to diagnosis. Unscored prompts were analyzed for
correlation to diagnosis using a Chi-squared (X2) test or Fisher’s exact test. A clinically
meaningful p-value (p ≤ 0.10) indicates a relationship between diagnosis and behavior, i.e.,
response. Unscored prompts where there is no evidence of a relationship between all possible
responses and diagnosis were candidates for removal during refinement. (a) Responses for
unscored prompts usc02, usc04-usc07, and pilo1-pilo8 that are correlated to diagnosis at a
clinically meaningful level (p ≤ 0.10) are considered biologically relevant to coping ability. (b)
Because the assumption of large cell count was violated for usc01 and usc03, correlation to
diagnosis was analyzed using the Fisher’s exact test (R code: fisher.test). There was no evidence
of a relationship between in-kennel location (usc01 and usc03) and diagnosis.
(a) Responses tested with the Chi-squared test
Abbrev.
Unscored Prompt
Response
usc02(a) Overall body language, HAI 1
usc02(b) Overall body language, HAI 1
usc02(c)
Overall body language, HAI 1
usc02(d) Overall body language, HAI 1
usc02(e) Overall body language, HAI 1
usc02(f)
Overall body language, HAI 1
usc02(g) Overall body language, HAI 1
usc02(h) Overall body language, HAI 1
Image 1
Image 2
Image 3
Image 4
Image 5
Image 6
Image 7
Image 8
110
X2
2.985
N/A
4.000
N/A
N/A
1.738
1.695
1.561
p
0.225
.
0.135
.
.
0.419
0.429
0.458
Table 3.7 (cont’d)
usc02(i)
Overall body language, HAI 1
usc02(j)
Overall body language, HAI 1
usc02(k) Overall body language, HAI 1
usc02(l)
Overall body language, HAI 1
usc02(m) Overall body language, HAI 1
usc02(n) Overall body language, HAI 1
usc02(o) Overall body language, HAI 1
usc02(p) Overall body language, HAI 1
usc02(q) Overall body language, HAI 1
usc02(r)
Overall body language, HAI 1
usc02(s)
Overall body language, HAI 1
usc02(t)
Overall body language, HAI 1
pilo1
Piloerection, HAI 1
usc04(a) Overall body language, HAI 2
usc04(b) Overall body language, HAI 2
usc04(c)
Overall body language, HAI 2
usc04(d) Overall body language, HAI 2
usc04(e) Overall body language, HAI 2
usc04(f)
Overall body language, HAI 2
usc04(g) Overall body language, HAI 2
usc04(h) Overall body language, HAI 2
usc04(i)
Overall body language, HAI 2
usc04(j)
Overall body language, HAI 2
usc04(k) Overall body language, HAI 2
usc04(l)
Overall body language, HAI 2
usc04(m) Overall body language, HAI 2
usc04(n) Overall body language, HAI 2
usc04(o) Overall body language, HAI 2
usc04(p) Overall body language, HAI 2
usc04(q) Overall body language, HAI 2
usc04(r)
Overall body language, HAI 2
usc04(s)
Overall body language, HAI 2
usc04(t)
Overall body language, HAI 2
pilo2
Piloerection, HAI 2
Image 9
Image 10
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Present
Image 1
Image 2
Image 3
Image 4
Image 5
Image 6
Image 7
Image 8
Image 9
Image 10
Image 11
Image 12
Image 13
Image 14
Image 15
Image 16
Image 17
Image 18
Image 19
Image 20
Present
111
N/A
0.449
6.129
2.235
N/A
.
0.799
0.0467*
0.327
.
18.596
9.16x10-05**
N/A
N/A
N/A
.
.
.
6.824
0.033*
N/A
N/A
2.985
1.476
N/A
N/A
2.985
N/A
9.438
3.589
.
.
0.225
0.478
.
.
0.225
.
0.0089**
0.166
10.730
0.0047**
2.746
7.790
4.831
1.695
N/A
3.589
N/A
N/A
N/A
N/A
0.652
3.589
N/A
0.253
0.020*
0.089*
0.429
.
0.166
.
.
.
.
0.721
0.166
.
Table 3.7 (cont’d)
usc05(a) Miscellaneous behavior, HAI 3
Pawing at exit
usc05(b) Miscellaneous behavior, HAI 3
usc05(c) Miscellaneous behavior, HAI 3
usc05(d) Miscellaneous behavior, HAI 3
usc05(e) Miscellaneous behavior, HAI 3
usc05(f) Miscellaneous behavior, HAI 3
pilo3
pilo4
pilo5
pilo6
pilo7
Piloerection, HAI 3
Piloerection, HAI 4
Piloerection, HAI 5
Piloerection, HAI 6
Piloerection, HAI 7
Pacing
Whining
Barking
Other
None
Present
Present
Present
Present
Present
usc06(a)
In-kennel misc. behavior, HAI 8
Spinning
usc06(b)
In-kennel misc. behavior, HAI 8
Excessive jumping
usc06(c)
In-kennel misc. behavior, HAI 8
Whining
usc06(d)
In-kennel misc. behavior, HAI 8
Excessive barking
usc06(e)
In-kennel misc. behavior, HAI 8
Pacing
usc06(f)
In-kennel misc. behavior, HAI 8
Frantic pawing at door
usc06(g)
In-kennel misc. behavior, HAI 8
Smeared feces in kennel
usc06(h)
In-kennel misc. behavior, HAI 8
Other
usc07(a) Out-of-kennel misc. behavior, HAI 8
Dog directed reactivity
usc07(b) Out-of-kennel misc. behavior, HAI 8
Other
pilo8
Piloerection, HAI 8
Present
**Statistically significant at α ≤ 0.0167
*Clinically meaningful at α ≤ 0.10
(b) Unscored prompts tested with Fisher’s exact test
Abbrev.
Unscored Prompt
usc01
usc03
Location after approach, HAI 1
Location with open door, HAI 1
1.476
2.985
3.504
N/A
3.754
5.446
3.589
N/A
N/A
0.478
0.225
0.173
.
0.153
0.066*
0.166
.
.
1.308
0.520
N/A
N/A
N/A
0.020
1.539
N/A
N/A
N/A
4.000
4.022
2.304
3.038
.
.
.
0.990
0.463
.
.
.
0.135
0.134
0.316
0.219
p
0.810
0.280
3.4: Discussion
The novel tool developed for measuring coping behavior can differentiate MC-AA dogs
from AC and MC-EA dogs in shelters at a statistically significant level. It was piloted at a single
112
humane society located in the Midwest (USA); however, based on previous shelter studies, the
sample population was similarly sized to that of dogs observed in shelters (Brown et al., 2013;
Shih et al., 2021). Furthermore, the age distribution was representative of the typical US shelter.
According to the US national shelter database, 30% of incoming shelter dogs in 2022 were less
than 5 months of age, with the remainder 70% of incoming dogs 5 months or older (Data Detail
2022, Shelter Animals Count). Previous studies on shelter dogs have reported a range of ages;
however, we see a trend across studies demonstrating most shelter dogs are juveniles to adults,
with proportionately fewer geriatric residents (Arhant & Troxler, 2014; Brown et al., 2013;
Patronek et al., 1996; Salman et al., 1998).
While the tool was unable to differentiate MC-EA dogs from AC dogs at a statistically
significant level, numerically MC-EA dogs had higher marginal mean total scores than AC dogs
across all models apart from the Full Model for Indoor Assessments, in which there was a
numerical difference between means with no evidence of statistical significance. The failure to
detect a statistical difference in total scores from indoor assessments is likely due to the
influential impact of the two MC-EA/severe dogs that both scored negative total coping scores.
No dogs had extreme negative or positive coping scores. Therefore, our hypotheses were
partially supported in that we saw negative scores for MC-AA dogs and positive scores for MC-
EA dogs while scores for AC dogs fell between the two, although the range was more moderate
than predicted. There was an unexpectedly low proportion of MC-EA dogs compared to AC and
MC-AA dogs in our sample population in comparison to the focal sample from the pilot study.
When considering severity, there were only two dogs diagnosed with MC-EA severe. It is
possible that our study did not have enough power to detect a statistically significant difference
between MC-EA and AC dogs. Additionally, the median total score was not as expected
considering the diagnosis distribution of the sample population. This suggests that the scale of
the scoring rubric may have been too narrow to capture the variation among dogs. The score
values assigned to each prompt should be reassessed in future studies.
The effectiveness of a behavior assessment for shelter use depends not only on
reliability and validity, but also on practicality. The tool must be easy to implement and efficient.
Therefore, several components of the tool are candidates for removal based on validity analysis.
113
Removal of clinically irrelevant components will improve the efficiency of the behavior
assessment for practical use. While initial refinement was based on statistical methods (Model
A), clinical expertise informed the additional inclusion of sc01 and sc03 (Model B). Within the
industry, kennel presentation is weighed considerably in staff’s evaluation of shelter dogs. The
removal of kennel presentation may decrease user support, rendering the tool impractical. The
final iteration of the tool, therefore, consists of nine scored prompts (sc01, sc03-sc06, sc08,
sc09, sc13, sc14) and three unscored prompts (usc02, usc04, usc05) observed across five HAI’s
(see Appendix E).
Areas of concern that need further evaluation and critique include the statistically
insignificant findings of sc01 and sc03, and the incongruent distribution of diagnoses for sc01,
sc05, sc09, and sc13. There was no statistical evidence that sc01 and sc03, which score in-kennel
presentation, were significantly correlated with diagnosis. However, indirect evidence indicates
that in-kennel presentation may be clinically relevant; body language images 11, 14, and 18
(usc02(k), usc02(n), usc02(r)) were correlated with diagnosis. Within the shelter industry, in-
kennel presentation is heavily considered during formal and informal assessments. Disregard for
in-kennel presentation may decrease professional support in implementing the tool. It is
possible that the response options for in-kennel presentation (‘Friendly/attentive/neutral’,
‘Dissociated (fearful/stressed)’, ‘Aware (fearful/stressed)’, ‘Hyper-active’, ‘Aggression’, or ‘Hidden
from view’) are not fully representative of the behavioral responses of shelter dogs. Assessment
videos should be coded for in-kennel behavior to further evaluate the format and terminology
of in-kennel presentation.
Of the statistically significant scored prompts, there were four prompts (sc01, sc05, sc09,
sc13) with incongruent proportion of MC-AA to MC-EA dogs receiving either positive or negative
scores for that prompt. For scored prompts that correlate with diagnosis, it was expected that
MC-AA dogs would have negative scores and MC-EA dogs would have positive scores with some
variation. Therefore, the proportion of dogs receiving positive scores should predominately be
MC-EA dogs while negative scores should predominately be MC-AA dogs. ‘Activity, HAI 3’ (sc09)
had five MC-AA dogs score +3 while four MC-EA dogs scored +3. ‘Ability to obtain the dog’s
attention, HAI 6’ (sc13) had two MC-EA dogs score –3 while one MC-AA dog scored –3. These
114
results could be explained by low occurrence and consequently low power. Further validity
testing using a larger sample population could better reveal the relationship of these two
prompts to diagnosis. However, the remaining prompt, ‘Initial reaction to play, HAI 2’ (sc05) had
28 MC-AA dogs score +3 while only eight MC-EA dogs scored +3. This is likely due to an error in
score value assignment. ‘Approach - does not engage’ (sc05(b) was assigned a score value of +3.
Upon re-evaluation by a board-certified veterinarian behaviorist, we believe this value should
be negative. There is evidence that sc05(b) is significantly correlated with diagnosis which is
reflected in the variation across diagnostic groups; however, we would propose restructuring
the prompt to better represent MC-AA and MC-EA behaviors. Allotting –3 points to sc05(b)
would better represent validity results, but video analysis and behavior coding of the
assessment videos can further inform appropriate revisions.
The scored prompt ‘Presentation with closed door, HAI 1’ (sc01) which was included in
Refined Model B, had one MC-AA dog score +3 for ‘Aggression’ (sc01(e)). This was the only
instance of ‘Aggression’ for in-kennel presentation. ‘Aggression’ was defined in the ethogram as
the presence of any one of the following behaviors: growling, snarling, snapping, lunging with
lips drawn away from teeth, hard stare with freeze. Dogs may exhibit aggression due to various
motivational states including both fear and excitement (Leuscher & Reisner, 2008). By ascribing
+3 points to ‘Aggression’, valence is assumed to be excessive arousal, which may not always be
the case. Furthermore, dogs that posed a safety risk due to aggression (i.e., bite history) were
excluded from enrollment. ‘Aggression’ may be an inappropriate category for evaluating in-
kennel behavior and instead more nuanced behaviors such as specific body language signals
should be considered. Tool revision should include re-evaluation of in-kennel presentation
response categories based on the behavioral coding of assessment videos.
There was evidence that three of the unscored prompts were clinically relevant (usc02,
usc04, usc05). Overall Body Language, HAI 1 (usc02) and Overall Body Language, HAI 2 (usc04)
prompts the assessor to circle all that apply from a collection of 20 images. While only six
images were found to be clinically or statistically significant, results could be influenced by the
presence of all 20 images and therefore none of the responses should be eliminated during
refinement. However, for Miscellaneous behavior, HAI 3 (usc05), the response options should be
115
revised to better reflect the clinical findings. Assessors were instructed to circle all that apply
from a selection of six responses; however, the only response that was clinically significant was
‘None’ (usc05(f)). Despite the nonexclusive nature of the prompt, if ‘None’ was selected, the
assessor would not have selected any of the other options. This implies that it is the absence of
behaviors that correlates with at least one diagnosis. To improve usability, usc05 was revised to
capture the presence or absence of each behavior for the refined assessment (Appendix E).
Responses were restructured based on relatedness and integration of the predominant
behaviors written for response ‘Other (write)’ (usc05(e)). Therefore, the revised prompt became
‘Are the following behaviors observed (Yes or No), HAI 3’ with the responses ‘Attention seeking
through close proximity/touch’, ‘Vigilance to environment’, ‘Vocalizing’, and ‘Pacing and/or
pawing at exit’ for inclusion in the refined assessment form (Appendix E).
Upon evaluation post-analysis, the tool was found to lack the ability to quantify
solicitation of human touch. Anecdotally, there was variation among individuals during
acclimation and Play (HAI 2) in desire to interact with the human through proximity. Several
dogs that were not interested in play were content to receive petting and scratching from the
assessor instead. Additionally, during Settle (HAI 3), assessors were instructed to ignore the dog
and observe behavior independent from human cues. However, several dogs would solicit
attention during this time, illustrated by the inclusion of ‘Attention seeking through close
proximity/touch’ in the refined assessment form. HAI 3 was predominately inspired by the
collection of studies that investigate activity as it correlates with shelter dog welfare (for review
see Protopopova, 2016). Monitoring activities like eating, resting, or drinking in the presence of
a human is common practice in shelters as a measure of positive welfare (Bauer et al., 2017).
Therefore, HAI 3 included the opportunity for independent activity with the availability of a food
reward. A water dish located in the assessment area was available throughout the assessment.
Incorporating a measurement for solicitation of touch in either HAI 2 or HAI 3 may improve the
tool’s ability to differentiate between diagnostic groups.
During HAI 6, the dog is allowed to interact with the model dog off-leash, of which the
behavior of the dog was clinically relevant to diagnostic outcome. The version of the tool
analyzed for reliability and validity included the unscored, open-response prompt ‘Note
116
behavior after the dog is let off-leash' (supp.01). Written responses were not analyzed in this
study; however, future evaluation of behavior using video coding can inform the addition of a
score for interaction with the model dog while off-leash.
3.4.i: Limitations
The histories of shelter dogs are largely unknown, confounding the ability to comment
on behavioral influences from past experiences. There is the potential that prior experiences
may influence behavior during certain human-animal interactions, as familiar scenarios may
cause a conditioned response in lieu of a coping response. However, it is unlikely that the
assessed dogs would have experienced similar interactions unless they have previously been
housed at a shelter that performs temperament testing. Future goals include validation of the
refined tool to track behavioral changes in shelter over time. Habituation to the assessment can
be expected, however sensitization may indicate poor coping ability. Therefore, it is possible
that prior experience with behavior testing does not limit the ability of the tool to quantify
coping ability. Longitudinal studies that follow individuals across time are needed to validate
repeat performance.
Factors that may have affected behavior during assessment include the presence of an
additional human (the second rater) and the presence of the GoPro cameras. For reliability
testing (discussed in Chapter 2), there were two raters present during assessments. One rater
handled the dog and performed the assessment while the second rater observed without
interacting with the dog. Initial reaction to the observer and cameras ranged from avoidance,
absence of acknowledgement, or exploratory approach. Most dogs reacted with approach, and
it is unlikely behavior was significantly affected as all dogs assumed neutral behavior toward the
observer and cameras after acclimation.
Criterion validity was established by a single veterinarian. While the gold-standard for
psychological diagnosis in dogs is the evaluation by a veterinarian behaviorist, it is possible that
factors such as video perspective and unknown medical history may have affected diagnostics.
Evaluation of the assessment videos by additional veterinarian behaviorists and subsequent
reliability analysis could reduce the margin of error. Validation of the tool is statistically limited
to differentiating MC-AA dogs from AC and MC-EA dogs. The inability of the tool to identify MC-
117
EA may have been a consequence of low power or the influence of the two identified influential
outliers that were diagnosed MC-EA/severe, both of which scored a negative value for total
coping score.
3.4.ii: Future goals
Additional studies examining reliability and validity in various shelters would provide a
more comprehensive understanding of coping behavior in relation to tool use. Future goals
should be to a) validate the refined tool’s ability to differentiate between MC-EA and AC dogs
and b) establish score values that communicate pharmaceutical need using the receiver
operating characteristic (ROC) curve to establish optimal cut-off points (Akobeng, 2007c). Once
threshold values are established for diagnosis of MC-AA and MC-EA dogs, specificity (proportion
of true AC dogs that score AC) and sensitivity (proportion of true MC dogs that score MC) can be
determined to calculate likelihood ratios (the ratio between the probability of observing the
score in MC dogs to the probability of observing the same result in AC dogs), commonly
reported as a measure for the usefulness of diagnostic tests (Akobeng, 2007a; Akobeng, 2007b).
Finally, the refined tool can be used in future studies to track behavioral changes over time and
in response to behavioral interventions. Long-term, the assessment could be distributed
externally and formatted for a mobile app that automatically logs behavior into a shared
database accessible by shelter staff and veterinarians to facilitate fluid partnerships. Such a
database could be utilized to deepen our understanding of species-specific needs and coping
strategies on a multicentric level and inform the care and welfare of shelter dogs nationally or
even internationally.
3.5: Conclusion
A novel tool was developed to assess and quantify coping behavior of shelter dogs. The
tool was able to statistically differentiate MC-AA dogs and numerically differentiate MC-EA dogs
from AC dogs based on total coping score. Six of the original 15 scored prompts and four of the
original seven unscored prompts can be eliminated while maintaining validity, resulting in a
refined version of the tool that consists of five HAI’s: Approach (formerly In-Kennel: Begin, HAI
1), Play (formerly Play, HAI 2), Settle (formerly Settle, HAI 3), Model Dog (formerly Distraction,
HAI 6), and Return (formerly In-Kennel: Return, HAI 7). The refined tool will be further edited
118
based on behavioral coding of assessment videos before use in future studies to validate the
refined version of the assessment and to track behavioral changes in shelter dogs over time and
in response to pharmaceutical interventions.
119
BIBLIOGRAPHY
Abrams, J., Brennen, R., & Byosiere, S.E. (2020). Trazodone as a mediator of transitional stress in
a shelter: Effects on illness, length of stay, and outcome. Journal of Veterinary Behavior,
36, 13–18.
Ader, R., & Cohen, N. (1993). Psychoneuroimmunology: Conditioning and stress. Annual Review
of Psychology, 44, 53–85.
Akobeng, A. K. (2007a). Understanding diagnostic tests 1: Sensitivity, specificity and predictive
values. Acta Paediatrica, 96(3), 338–341.
Akobeng, A. K. (2007b). Understanding diagnostic tests 2: Likelihood ratios, pre- and post-test
probabilities and their use in clinical practice. Acta Paediatrica, 96(4), 487–491.
Akobeng, A. K. (2007c). Understanding diagnostic tests 3: Receiver operating characteristic
curves. Acta Paediatrica, 96(5), 644–647.
Albuquerque, N., & Resende, B. (2023). Dogs functionally respond to and use emotional
information from human expressions. Evolutionary Human Sciences, 5, e2.
Appleby, D. L., Bradshaw, J. W., & Casey, R. A. (2002). Relationship between aggressive and
avoidance behaviour by dogs and their experience in the first six months of life. The
Veterinary Record, 150(14), 434–438.
Arena, L., Berteselli, G. V., Lombardo, F., Candeloro, L., Dalla Villa, P., & De Massis, F. (2019a).
Application of a welfare assessment tool (Shelter Quality Protocol) in 64 Italian long-
term dogs’ shelters: Welfare hazard analysis. Animal Welfare , 28(3), 353–363.
Arena, L., Wemelsfelder, F., Messori, S., Ferri, N., & Barnard, S. (2017). Application of Free
Choice Profiling to assess the emotional state of dogs housed in shelter environments.
Applied Animal Behaviour Science, 195, 72–79.
Arena, L., Wemelsfelder, F., Messori, S., Ferri, N., & Barnard, S. (2019b). Development of a fixed
list of terms for the Qualitative Behavioural Assessment of shelter dogs. PloS One,
14(10), e0212652.
Arhant, C., & Troxler, J. (2014). Approach behaviour of shelter dogs and its relationships with the
attitudes of shelter staff to dogs. Applied Animal Behaviour Science, 160, 116–126.
Ballantyne, K. C. (2018). Separation, confinement, or noises: What is scaring that dog? The
Veterinary Clinics of North America. Small Animal Practice, 48(3), 367–386.
120
Barnard, S., Pedernera, C., Candeloro, L., Ferri, N., Velarde, A., & Dalla Villa, P. (2016).
Development of a new welfare assessment protocol for practical application in long-term
dog shelters. The Veterinary Record, 178(1), 18.
Barnard, S., Siracusa, C., Reisner, I., Valsecchi, P., & Serpell, J. A. (2012). Validity of model devices
used to assess canine temperament in behavioral tests. Applied Animal Behaviour
Science, 138(1), 79–87.
Barnard, S., Wells, D. L., & Hepper, P. G. (2018). Laterality as a Predictor of Coping Strategies in
Dogs Entering a Rescue Shelter. Symmetry, 10(11), 538.
Bartlett, P. C., Bartlett, A., Walshaw, S., & Halstead, S. (2005). Rates of euthanasia and adoption
for dogs and cats in Michigan animal shelters. Journal of Applied Animal Welfare Science:
JAAWS, 8(2), 97–104.
Bateson, P. (2014). Play, Playfulness, Creativity and Innovation. Animal Behavior and Cognition,
1(2), 99–112.
Bauer, A. E., Jordan, M., Colon, M., Shreyer, T., & Croney, C. C. (2017). Evaluating FIDO:
Developing and pilot testing the Field Instantaneous Dog Observation tool. Pet
Behaviour Science, 4, 1–15.
Beerda, B., Schilder, M. B., Bernadina, W., van Hooff, J. A., de Vries, H. W., & Mol, J. A. (1999b).
Chronic stress in dogs subjected to social and spatial restriction. II. Hormonal and
immunological responses. Physiology & Behavior, 66(2), 243–254.
Beerda, B., Schilder, M. B., van Hooff, J. A., & de Vries, H. W. (1997). Manifestations of chronic
and acute stress in dogs. Applied Animal Behaviour Science, 52(3), 307–319.
Beerda, B., Schilder, M. B., van Hooff, J. A., de Vries, H. W., & Mol, J. A. (1999a). Chronic stress in
dogs subjected to social and spatial restriction. I. Behavioral responses. Physiology &
Behavior, 66(2), 233–242.
Bennett, S. L., Litster, A., Weng, H.Y., Walker, S. L., & Luescher, A. U. (2012). Investigating
behavior assessment instruments to predict aggression in dogs. Applied Animal
Behaviour Science, 141(3), 139–148.
Benus, R. F., Bohus, B., Koolhaas, J. M., & van Oortmerssen, G. A. (1989). Behavioural strategies
of aggressive and non-aggressive male mice in active shock avoidance. Behavioural
Processes, 20(1-3), 1–12.
Berteselli, G. V., Arena, L., Candeloro, L., Dalla Villa, P., & De Massis, F. (2019). Interobserver
agreement and sensitivity to climatic conditions in sheltered dogs’ welfare evaluation
121
performed with welfare assessment protocol (Shelter Quality protocol). Journal of
Veterinary Behavior, 29, 45–52.
Biemer, P. P. (2010). Total survey error: Design, implementation, and evaluation. Public Opinion
Quarterly, 74(5), 817–848.
Bollen, K. S., & Horowitz, J. (2008). Behavioral evaluation and demographic information in the
assessment of aggressiveness in shelter dogs. Applied Animal Behaviour Science, 112(1),
120–135.
Boxall, J., Heath, S., Bate, S., & Brautigam, J. (2004). Modern concepts of socialisation for dogs:
Implications for their behaviour, welfare and use in scientific procedures. Alternatives to
Laboratory Animals: ATLA, 32 (sup2), 81–93.
Bradley, J., & Rajendran, S. (2021). Increasing adoption rates at animal shelters: A two-phase
approach to predict length of stay and optimal shelter allocation. BMC Veterinary
Research, 17, 70.
Bray, E. E., Sammel, M. D., Seyfarth, R. M., Serpell, J. A., & Cheney, D. L. (2017). Temperament
and problem solving in a population of adolescent guide dogs. Animal Cognition, 20(5),
923–939.
Broom, D. M. (1991). Animal welfare: Concepts and measurement. Journal of Animal Science,
69(10), 4167–4175.
Broom, D. M. (2019). Welfare concepts. In J.C. Choe (Ed.), Encyclopedia of animal behavior
(second edition) (Vol. 1, pp. 80–83). Academic Press.
Brown, W. P., Davidson, J. P., & Zuefle, M. E. (2013). Effects of phenotypic characteristics on the
length of stay of dogs at two no kill animal shelters. Journal of Applied Animal Welfare
Science: JAAWS, 16(1), 2–18.
Buff, P. R., Carter, R. A., Bauer, J. E., & Kersey, J. H. (2014). Natural pet food: a review of natural
diets and their impact on canine and feline physiology. Journal of Animal Science, 92(9),
3781–3791.
Burghardt, G. M. (2005). The genesis of animal play: Testing the limits. MIT Press.
Burghardt, G. M., Bartmess-LeVasseur, J. N., Browning, S. A., Morrison, K. E., Stec, C. L., Zachau,
C. E., & Freeberg, T. M. (2012). Perspectives - minimizing observer bias in behavioral
studies: A review and recommendations. Ethology: Formerly Zeitschrift Fur
Tierpsychologie, 118(6), 511–517.
122
Canine Life Stage Definitions. (n.d.). Retrieved August 17, 2023, from
https://www.aaha.org/aaha-guidelines/life-stage-canine-2019/canine-life-stage-
definitions/
Cannon, W. B. (1915). Bodily changes in pain, hunger, fear, and rage: An account of recent
researches into the function of emotional excitement. New York: D. Appleton.
Carenzi, C., & Verga, M. (2009). Animal welfare: Review of the scientific concept and definition.
Italian Journal of Animal Science, 8(sup1), 21–30.
Cicchetti, D. V., & Feinstein, A. R. (1990). High agreement but low kappa: II. Resolving the
paradoxes. Journal of Clinical Epidemiology, 43(6), 551–558.
Clay, L., Paterson, M. B., Bennett, P., Perry, G., & Phillips, C. C. (2020). Do behaviour assessments
in a shelter predict the behaviour of dogs post-adoption? Animals, 10(7), 1225.
Cohen, J. (1960). Kappa: Coefficient of concordance. Educational and Psychological
Measurement.
Cohen, S. E., & Todd, P. M. (2019). Stated and revealed preferences in companion animal choice.
Behavior Research Methods, 51(4), 1498–1509.
Coppola, C. L., Enns, R. M., & Grandin, T. (2006a). Noise in the animal shelter environment:
Building design and the effects of daily noise exposure. Journal of Applied Animal
Welfare Science: JAAWS, 9(1), 1–7.
Coppola, C. L., Grandin, T., & Enns, R. M. (2006b). Human interaction and cortisol: Can human
contact reduce stress for shelter dogs? Physiology & Behavior, 87(3), 537–541.
Corsetti, S., Borruso, S., Malandrucco, L., Spallucci, V., Maragliano, L., Perino, R., D’Agostino, P., &
Natoli, E. (2021). Cannabis sativa L. may reduce aggressive behaviour towards humans in
shelter dogs. Scientific Reports, 11(1), 2773.
Corson, S. A., & O’Leary Corson, E. (1976). Constitutional differences in physiologic adaptation
to stress and distress. In G. Serban (Ed.), Psychopathology of Human Adaptation (pp. 77–
94). Springer US.
Couper, M. P. (2001). Web survey design and administration. Public Opinion Quarterly, 65(2),
230–253.
Data Detail 2022. (n.d.). Shelter Animals Count. Retrieved August, 17 2023, from
https://www.shelteranimalscount.org/intake-and-outcome-database-iod/
123
Dawkins, M. S. (1998). Evolution and animal welfare. The Quarterly Review of Biology, 73(3),
305–328.
De Boer, S. F., Slangen, J. L., & Van der Gugten, J. (1990). Plasma catecholamine and
corticosterone levels during active and passive shock-prod avoidance behavior in rats:
Effects of chlordiazepoxide. Physiology & Behavior, 47(6), 1089–1098.
De Palma, C., Viggiano, E., Barillari, E., Palme, R., Dufour, A. B., Fantin, C., & Natoli, E. (2005).
Evaluating the temperament in shelter dogs. Behaviour, 142(9), 1307–1328.
Diederich, C., & Giffroy, J.-M. (2006). Behavioural testing in dogs: A review of methodology in
search for standardisation. Applied Animal Behaviour Science, 97(1), 51–72.
Diesel, G., Brodbelt, D., & Pfeiffer, D. U. (2008). Reliability of assessment of dogs’ behavioural
responses by staff working at a welfare charity in the UK. Applied Animal Behaviour
Science, 115(3-4), 171–181.
Diesel, G., Brodbelt, D., & Pfeiffer, D. U. (2010). Characteristics of relinquished dogs and their
owners at 14 rehoming centers in the United Kingdom. Journal of Applied Animal
Welfare Science: JAAWS, 13(1), 15–30.
Dillman, D. A. (2011). Mail and Internet Surveys: The Tailored Design Method – 2007 Update
with New Internet, Visual, and Mixed-Mode Guide. John Wiley & Sons.
Dowling-Guyer, S., Marder, A., & D’Arpino, S. (2011). Behavioral traits detected in shelter dogs
by a behavior evaluation. Applied Animal Behaviour Science, 130(3), 107–114.
Downes, M. J., Dean, R. S., Stavisky, J. H., Adams, V. J., Grindlay, D. J. C., & Brennan, M. L. (2013).
Methods used to estimate the size of the owned cat and dog population: A systematic
review. BMC Veterinary Research, 9, 121.
Engel, G. L., & Schmale, A. H. (1972). Conservation-withdrawal: A primary regulatory process for
organismic homeostasis. In R. Porter & J. Knight (Eds.), Physiology, emotion, &
psychosomatic illness (pp. 57–85). Associated Scientific Publishers.
Feigenson, K. A., Kusnecov, A. W., & Silverstein, S. M. (2014). Inflammation and the two-hit
hypothesis of schizophrenia. Neuroscience and Biobehavioral Reviews, 38, 72–93.
Feinstein, A. R., & Cicchetti, D. V. (1990). High agreement but low kappa: I. The problems of two
paradoxes. Journal of Clinical Epidemiology, 43(6), 543–549.
Fidler, M., Light, P., & Costall, A. (1996). Describing dog behavior psychologically: Pet owners
versus non-owners. Anthrozoös, 9(4), 196–200.
124
Foyer, P., Svedberg, A.M., Nilsson, E., Wilsson, E., Faresjö, Å., & Jensen, P. (2016). Behavior and
cortisol responses of dogs evaluated in a standardized temperament test for military
working dogs. Journal of Veterinary Behavior, 11, 7–12.
Fraser, D., & Duncan, I. J. (1998). ‘Pleasures’, ’pains' and animal welfare: Toward a natural history
of affect. Animal Welfare, 7(4), 383–396.
Gee, N. R., Rodriguez, K. E., Fine, A. H., & Trammell, J. P. (2021). Dogs supporting human health
and well-being: A biopsychosocial approach. Frontiers in Veterinary Science, 8, 630465.
Gilbert-Gregory, S. E., Stull, J. W., Rice, M. R., & Herron, M. E. (2016). Effects of trazodone on
behavioral signs of stress in hospitalized dogs. Journal of the American Veterinary
Medical Association: JAVMA, 249(11), 1281–1291.
Goddard, M. E., & Beilharz, R. G. (1984). A factor analysis of fearfulness in potential guide dogs.
Applied Animal Behaviour Science, 12(3), 253–265.
Goddard, M. E., & Beilharz, R. G. (1985). A multivariate analysis of the genetics of fearfulness in
potential guide dogs. Behavior Genetics, 15(1), 69–89.
Gold, P., Goodwin, F., & Chrousos, G. (1988). Clinical and biochemical manifestations of
depression. The New England Journal of Medicine, 319(7), 413–420.
Gruen, M. E., Roe, S. C., Griffith, E., Hamilton, A., & Sherman, B. L. (2014). Use of trazodone to
facilitate postsurgical confinement in dogs. Journal of the American Veterinary Medical
Association, 245(3), 296–301.
Gruen, M. E., & Sherman, B. L. (2008). Use of trazodone as an adjunctive agent in the treatment
of canine anxiety disorders: 56 cases (1995-2007). Journal of the American Veterinary
Medical Association, 233(12), 1902–1907.
Gunter, L. M., Feuerbacher, E. N., Gilchrist, R. J., & Wynne, C. D. (2019). Evaluating the effects of
a temporary fostering program on shelter dog welfare. PeerJ, 7, e6620.
Harvey, N. D. (2021). A Simple Guide to Inter-rater, Intra-rater and Test-retest Reliability for
Animal Behaviour Studies. School of Veterinary Medicine and Science, University of
Nottingham.
Haverbeke, A., Diederich, C., Depiereux, E., & Giffroy, J. M. (2008). Cortisol and behavioral
responses of working dogs to environmental challenges. Physiology & Behavior, 93, 59–
67.
125
Hawes, S. M., Camacho, B. A., Tedeschi, P., & Morris, K. N. (2019). Temporal trends in intake and
outcome data for animal shelter and rescue facilities in Colorado from 2000 through
2015. Journal of the American Veterinary Medical Association: JAVMA, 254(3), 363–372.
Hawes, S. M., Hupe, T. M., Gandenberger, J., & Morris, K. N. (2021). Temporal trends in intake
category data for animal shelter and rescue organizations in Colorado from 2008 to
2018. Journal of the American Veterinary Medical Association: JAVMA, 260(3), 311–319.
Hennessy, M. B. (2013). Using hypothalamic–pituitary–adrenal measures for assessing and
reducing the stress of dogs in shelters: A review. Applied Animal Behaviour Science,
149(1), 1–12.
Hennessy, M. B., Davis, H. N., Williams, M. T., Mellott, C., & Douglas, C. W. (1997). Plasma
cortisol levels of dogs at a county animal shelter. Physiology & Behavior, 62(3), 485–490.
Hennessy, M. B., Morris, A., & Linden, F. (2006). Evaluation of the effects of a socialization
program in a prison on behavior and pituitary–adrenal hormone levels of shelter dogs.
Applied Animal Behaviour Science, 99(1), 157–171.
Hennessy, M. B., Voith, V. L., Hawke, J. L., Young, T. L., Centrone, J., McDowell, A. L., Linden, F., &
Davenport, G. M. (2002). Effects of a program of human interaction and alteration in diet
composition on activity of the hypothalamic-pituitary-adrenal axis in dogs housed in a
public animal shelter. Journal of the American Veterinary Medical Association: JAVMA,
221(1), 65–71.
Hennessy, M. B., Williams, M. T., Miller, D. D., Douglas, C. W., & Voith, V. L. (1998). Influence of
male and female petters on plasma cortisol and behaviour: Can human interaction
reduce the stress of dogs in a public animal shelter? Applied Animal Behaviour Science,
61(1), 63–77.
Hennessy, M. B., Willen, R. M., & Schiml, P. A. (2020). Psychological Stress, Its Reduction, and
Long-Term Consequences: What Studies with Laboratory Animals Might Teach Us about
Life in the Dog Shelter. Animals, 10(11), 2061.
Henry, J. P., & Stephens, P. M. (1977). Stress, health, and the social environment: A sociobiologic
approach to medicine (K. E. Schaefer, Ed.). Springer-Verlag New York Inc.
Herron, M. E., Lord, L. K., & Husseini, S. E. (2014). Effects of preadoption counseling on the
prevention of separation anxiety in newly adopted shelter dogs. Journal of Veterinary
Behavior, 9(1), 13–21.
Hessing, M. J., Hagelsø, A. M., Schouten, W. G., Wiepkema, P. R., & Van Beek, J. A. (1994).
Individual behavioral and physiological strategies in pigs. Physiology & Behavior, 55(1),
39–46.
126
Hetts, S., Derrell Clark, J., Calpin, J. P., Arnold, C. E., & Mateo, J. M. (1992). Influence of housing
conditions on beagle behaviour. Applied Animal Behaviour Science, 34(1), 137–155.
Hiby, E. F., Rooney, N. J., & Bradshaw, J. W. (2006). Behavioural and physiological responses of
dogs entering re-homing kennels. Physiology & Behavior, 89(3), 385–391.
Horowitz, A. (2009). Attention to attention in domestic dog (Canis familiaris) dyadic play. Animal
Cognition, 12(1), 107–118.
Horváth, Z., Dóka, A., & Miklósi, A. (2008). Affiliative and disciplinary behavior of human
handlers during play with their dog affects cortisol concentrations in opposite directions.
Hormones and Behavior, 54(1), 107–114.
Horváth, Z., Igyártó, B.-Z., Magyar, A., & Miklósi, A. (2007). Three different coping styles in police
dogs exposed to a short-term challenge. Hormones and Behavior, 52(5), 621–630.
Hubrecht, R. C., Serpell, J. A., & Poole, T. B. (1992). Correlates of pen size and housing conditions
on the behaviour of kennelled dogs. Applied Animal Behaviour Science,34(4), 365-383.
Hubrecht, R. C., Wickens, S., & Kirkwood, J. (2017). The welfare of dogs in human care. In J. A.
Serpell (Ed.), The domestic dog: Its evolution, behavior and interactions with people
(second, pp. 271–299). Cambridge University Press.
Jones, A. C., & Gosling, S. D. (2005). Temperament and personality in dogs (Canis familiaris): A
review and evaluation of past research. Applied Animal Behaviour Science, 95(1), 1–53.
Jones, B., & Boissy, A. (2011). Fear and other negative emotions. In M. C. Appleby, J. A. Mench, I.
A. Olsson, & B. O. Hughes (Eds.), Animal welfare (pp. 78–98). CAB International.
Kartashova, I. A., Ganina, K. K., Karelina, E. A., & Tarasov, S. A. (2021). How to evaluate and
manage stress in dogs – A guide for veterinary specialist. Applied Animal Behaviour
Science, 243, 105458.
Kiddie, J. L., & Collins, L. M. (2014). Development and validation of a quality of life assessment
tool for use in kennelled dogs (Canis familiaris). Applied Animal Behaviour Science, 158,
57–68.
Kim, S. A., Borchardt, M. R., Lee, K., Stelow, E. A., & Bain, M. J. (2022). Effects of trazodone on
behavioral and physiological signs of stress in dogs during veterinary visits: A randomized
double-blind placebo-controlled crossover clinical trial. Journal of the American
Veterinary Medical Association: JAVMA, 260(8), 876–883.
127
King, T., Hemsworth, P. H., & Coleman, G. J. (2003). Fear of novel and startling stimuli in
domestic dogs. Applied Animal Behaviour Science, 82(1), 45–64.
Knudson, A. G., Jr. (1971). Mutation and cancer: Statistical study of retinoblastoma. Proceedings
of the National Academy of Sciences of the United States of America, 68(4), 820–823.
Koolhaas, J. M., Korte, S. M., De Boer, S. F., Van Der Vegt, B. J., Van Reenen, C. G., Hopster, H., De
Jong, I. C., Ruis, M. A., & Blokhuis, H. J. (1999). Coping styles in animals: Current status in
behavior and stress-physiology. Neuroscience and Biobehavioral Reviews, 23(7), 925–
935.
Korte, S. M., Buwalda, B., Bouws, G. A., Koolhaas, J. M., Maes, F. W., & Bohus, B. (1992).
Conditioned neuroendocrine and cardiovascular stress responsiveness accompanying
behavioral passivity and activity in aged and in young rats. Physiology & Behavior, 51(4),
815–822.
Korte, S. M., Ruesink, W., & Blokhuis, H. J. (1999). Heart rate variability during manual restraint
in chicks from high- and low-feather pecking lines of laying hens. Physiology & Behavior,
65(4-5), 649–652.
Kuhne, F., Hößler, J. C., & Struwe, R. (2014). Emotions in dogs being petted by a familiar or
unfamiliar person: Validating behavioural indicators of emotional states using heart rate
variability. Applied Animal Behaviour Science, 161, 113–120.
Lamon, T. K., Slater, M. R., Moberly, H. K., & Budke, C. M. (2021). Welfare and quality of life
assessments for shelter dogs: A scoping review. Applied Animal Behaviour Science, 244,
105490.
Lazar, J., Dudley-Sponaugle, A., & Greenidge, K.-D. (2004). Improving web accessibility: A study
of webmaster perceptions. Computers in Human Behavior, 20(2), 269–288.
Lazzaroni, M., Range, F., Backes, J., Portele, K., Scheck, K., & Marshall-Pescini, S. (2020). The
effect of domestication and experience on the social interaction of dogs and wolves with
a human companion. Frontiers in Psychology, 11, 785.
Ley, J., Coleman, G. J., Holmes, R., & Hemsworth, P. H. (2007). Assessing fear of novel and
startling stimuli in domestic dogs. Applied Animal Behaviour Science, 104, 71–84.
Lindsay, S. R. (2005). Handbook of applied dog behavior and training: Procedures and protocols
(S. R. Lindsay (Ed.)) [PDF]. Iowa State University Press.
Lucidi, P., Bernabò, N., Panunzi, M., Villa, P. D., & Mattioli, M. (2005). Ethotest: A new model to
identify (shelter) dogs’ skills as service animals or adoptable pets. Applied Animal
Behaviour Science, 95(1), 103–122.
128
Luescher, A. U., & Reisner, I. R. (2008). Canine aggression toward familiar people: A new look at
an old problem. The Veterinary Clinics of North America. Small Animal Practice, 38(5),
1107–1130.
Luescher, A. U., & Medlock, R. T. (2009). The effects of training and environmental alterations on
adoption success of shelter dogs. Applied Animal Behaviour Science, 117(1), 63–68.
Maestripieri, D., Schino, G., Aureli, F., & Troisi, A. (1992). A modest proposal: Displacement
activities as an indicator of emotions in primates. Animal Behaviour, 44(5), 967–979.
Maier, S. F. (1984). Learned helplessness and animal models of depression. Progress in Neuro-
Psychopharmacology & Biological Psychiatry, 8(3), 435–446.
Manteca, X., Villalba, J. J., Atwood, S. B., Dziba, L., & Provenza, F. D. (2008). Is dietary choice
important to animal welfare? Journal of Veterinary Behavior, 3(5), 229–239.
Marsh, D. M., & Hanlon, T. J. (2007). Seeing what we want to see: Confirmation bias in animal
behavior research. Ethology, 113(11), 1089–1098.
Marston, L. C., & Bennett, P. C. (2003). Reforging the bond—Towards successful canine
adoption. Applied Animal Behaviour Science, 83(3), 227–245.
Martin, P., & Caro, T. M. (1985). On the functions of play and its role in behavioral development.
In J. S. Rosenblatt, C. Beer, M. C. Busnel, & P. J. B. Slater (Eds.), Advances in the study of
behavior (Vol. 15, pp. 59–103). Academic Press.
McCall, G. J. (1984). Systematic Field Observation. Annual Review of Sociology, 10, 263–282.
McNicholas, J., Gilbey, A., Rennie, A., Ahmedzai, S., Dono, J. A., & Ormerod, E. (2005). Pet
ownership and human health: A brief review of evidence and issues. British Medical
Journal: BMJ , 331(7527), 1252–1254.
Menchetti, L., Righi, C., Guelfi, G., Enas, C., Moscati, L., Mancini, S., & Diverio, S. (2019). Multi-
Operator Qualitative Behavioural Assessment for dogs entering the shelter. Applied
Animal Behaviour Science, 213, 107–116.
Menor-Campos, D. J., Molleda-Carbonell, J. M., & López-RodrÃguez, R. (2011). Effects of exercise
and human contact on animal welfare in a dog shelter. The Veterinary Record, 169(15),
388.
Mertens, P. A., & Unshelm, J. (1996). Effects of group and individual housing on the behaviour of
kennelled dogs in animal shelters. Anthrozoos, 9(1), 40–51.
129
Meyer, I., Forkman, B., & Paul, E. S. (2014). Factors affecting the human interpretation of dog
behavior. Anthrozoös, 27(1), 127–140.
Mornement, K. M., Coleman, G. J., Toukhsati, S., & Bennett, P. C. (2014). Development of the
behavioural assessment for re-homing K9’s (B.A.R.K.) protocol. Applied Animal Behaviour
Science, 151, 75–83.
Mornement, K. M., Coleman, G. J., Toukhsati, S. R., & Bennett, P. C. (2015). Evaluation of the
predictive validity of the Behavioural Assessment for Re-homing K9’s (B.A.R.K.) protocol
and owner satisfaction with adopted dogs. Applied Animal Behaviour Science, 167, 35–
42.
Morris, E. M., Kitts-Morgan, S. E., Spangler, D. M., McLeod, K. R., Costa, J. H., & Harmon, D. L.
(2020). The impact of feeding cannabidiol (CBD) containing treats on canine response to
a noise-induced fear response test. Frontiers in Veterinary Science, 7, 569565.
Mugenda, L., Shreyer, T., & Croney, C. (2019). Refining canine welfare assessment in kennels:
Evaluating the reliability of Field Instantaneous Dog Observation (FIDO) scoring. Applied
Animal Behaviour Science, 221, 104874.
Murphy, J. A. (1998). Describing categories of temperament in potential guide dogs for the
blind. Applied Animal Behaviour Science, 58(1), 163–178.
Netto, W. J., & Planta, D. J. (1997). Behavioural testing for aggression in the domestic dog.
Applied Animal Behaviour Science, 52(3), 243–263.
Normando, S., Stefanini, C., Meers, L., Adamelli, S., Coultis, D., & Bono, G. (2006). Some factors
influencing adoption of sheltered dogs. Anthrozoös, 19(3), 211–224.
Overall, K. L. (2017). Appeasement, calming signals, and information capture: How do our
subjects tell us what matters to them? Journal of Veterinary Behavior, 19, v – viii.
Part, C. E., Kiddie, J. L., Hayes, W. A., Mills, D. S., Neville, R. F., Morton, D. B., & Collins, L. M.
(2014). Physiological, physical and behavioural changes in dogs (Canis familiaris) when
kennelled: Testing the validity of stress parameters. Physiology & Behavior, 133, 260–
271.
Pastore, C., Pirrone, F., Balzarotti, F., Faustini, M., Pierantoni, L., & Albertini, M. (2011).
Evaluation of physiological and behavioral stress-dependent parameters in agility dogs.
Journal of Veterinary Behavior, 6(3), 188–194.
Patronek, G. J., & Bradley, J. (2016). No better than flipping a coin: Reconsidering canine
behavior evaluations in animal shelters. Journal of Veterinary Behavior, 15, 66–77.
130
Patronek, G. J., Bradley, J., & Arps, E. (2019). What is the evidence for reliability and validity of
behavior evaluations for shelter dogs? A prequel to “No better than flipping a coin.â€
Journal of Veterinary Behavior, 31, 43–58.
Patronek, G. J., & Glickman, L. T. (1994). Development of a model for estimating the size and
dynamics of the pet dog population. Anthrozoos, 7(1), 25–42.
Patronek, G. J., Glickman, L. T., Beck, A. M., McCabe, G. P., & Ecker, C. (1996). Risk factors for
relinquishment of dogs to an animal shelter. Journal of the American Veterinary Medical
Association: JAVMA, 209(3), 572–581.
Pedretti, G., Canori, C., Biffi, E., Marshall-Pescini, S., & Valsecchi, P. (2023). Appeasement
function of displacement behaviours? Dogs’ behavioural displays exhibited towards
threatening and neutral humans. Animal Cognition, 26(3), 943–952.
Pet Statistics. (n.d.). ASPCA. Retrieved August 2, 2023, from https://www.aspca.org/helping-
people-pets/shelter-intake-and-surrender/pet-statistics
Polgár, Z., Blackwell, E. J., & Rooney, N. J. (2019). Assessing the welfare of kennelled dogs-A
review of animal-based measures. Applied Animal Behaviour Science, 213, 1–13.
Popescu, S., Borda, C., El Mahdy, C., Lazar, E. A., Blaga Petrean, A., Campean, I. D., & Spinu, M.
(2018). The effect of a social enrichment programme on the behaviour of dogs from a
private shelter. Bulletin UASVM Veterinary Medicine, 75(2), 222–230.
Powell, L., Houlihan, C., Stone, M., Gitlin, I., Ji, X., Reinhard, C. L., & Watson, B. (2021). Animal
shelters’ response to the COVID-19 pandemic: A pilot survey of 14 shelters in the
northeastern United States. Animals, 11(9), 2669.
Prato-Previde, E., Custance, D. M., Spiezio, C., & Sabatini, F. (2003). Is the dog-human
relationship an attachment bond? An observational study using Ainsworth’s strange
situation. Behaviour, 140(2), 225–254.
Previde, E. P., & Valsechhi, P. (2007). Effects of abandonment on attachment behavior of adult
pet dogs. Journal of Veterinary Behavior: Clinical Applications and Research, 2(3), 87–88.
Protopopova, A. (2016). Effects of sheltering on physiology, immune function, behavior, and the
welfare of dogs. Physiology & Behavior, 159, 95–103.
Protopopova, A., & Gunter, L. M. (2017). Adoption and relinquishment interventions at the
animal shelter: A review. Animal Welfare , 26(1), 35–48.
131
Protopopova, A., Hauser, H., Goldman, K. J., & Wynne, C. D. (2018). The effects of exercise and
calm interactions on in-kennel behavior of shelter dogs. Behavioural Processes, 146, 54–
60.
Protopopova, A., Mehrkam, L. R., Boggess, M. M., & Wynne, C. D. (2014). In-kennel behavior
predicts length of stay in shelter dogs. PloS One, 9(12), e114319.
Protopopova, A., & Wynne, C. D. (2014). Adopter-dog interactions at the shelter: Behavioral and
contextual predictors of adoption. Applied Animal Behaviour Science, 157, 109–116.
Protopopova, A., & Wynne, C. D. (2015). Improving in-kennel presentation of shelter dogs
through response-dependent and response-independent treat delivery. Journal of
Applied Behavior Analysis, 48(3), 590–601.
Pullen, A. J., Merrill, R. J., & Bradshaw, J. W. (2010). Preferences for toy types and presentations
in kennel housed dogs. Applied Animal Behaviour Science, 125(3), 151–156.
Q2 2023 Analysis. (2023). Shelter Animals Count. https://www.shelteranimalscount.org/wp-
content/uploads/2023/07/Q2-Report.pdf
Reid, P., & Collins, K. (2012). Assessing conspecific aggression in fighting dogs. Proceedings of
the AVSAB/ACVB Meeting, 37–39.
ŘezáÄ, P., Viziová, P., DobeÅ¡ová, M., HavlÃÄek, Z., & PospÃÅ¡ilová, D. (2011). Factors affecting dog–
dog interactions on walks with their owners. Applied Animal Behaviour Science, 134(3),
170–176.
Rodriguez, J. R., Davis, J., Hill, S., Wolf, P. J., Hawes, S. M., & Morris, K. N. (2022). Trends in intake
and outcome data from U.S. animal shelters from 2016 to 2020. Frontiers in Veterinary
Science, 9, 863990.
Rooney, N. J., Bradshaw, J. W., & Robinson, I. H. (2000). A comparison of dog–dog and dog–
human play behaviour. Applied Animal Behaviour Science, 66(3), 235–248.
Rooney, N. J., Gaines, S. A., & Bradshaw, J. W. (2007). Behavioural and glucocorticoid responses
of dogs (Canis familiaris) to kennelling: Investigating mitigation of stress by prior
habituation. Physiology & Behavior, 92(5), 847–854.
Rowan, A., & Kartal, T. (2018). Dog Population & Dog Sheltering Trends in the United States of
America. Animals, 8(5), 0068.
Ruis, M. A., te Brake, J. H., van de Burgwal, J. A., de Jong, I. C., Blokhuis, H. J., & Koolhaas, J. M.
(2000). Personalities in female domesticated pigs: Behavioural and physiological
indications. Applied Animal Behaviour Science, 66(1), 31–47.
132
Russo, A., Dowling-Guyer, S., & McCobb, E. (2021). Community programming for companion dog
retention: A survey of animal welfare organizations. Journal of Applied Animal Welfare
Science: JAAWS, 1–15.
Sales, G., Hubrecht, R., Peyvandi, A., Milligan, S., & Shield, B. (1997). Noise in dog kennelling: Is
barking a welfare problem for dogs? Applied Animal Behaviour Science, 52(3), 321–329.
Salman, M. D., New, J. G., Jr, Scarlett, J. M., Kass, P. H., Ruch-Gallie, R., & Hetts, S. (1998). Human
and animal factors related to relinquishment of dogs and cats in 12 selected animal
shelters in the United States. Journal of Applied Animal Welfare Science: JAAWS, 1(3),
207–226.
Scheifele, P., Martin, D., Clark, J. G., Kemper, D., & Wells, J. (2012). Effect of kennel noise on
hearing in dogs. American Journal of Veterinary Research, 73(4), 482–489.
Schouten, W. G., & Wiegant, V. M. (1997). Individual responses to acute and chronic stress in
pigs. Acta Physiologica Scandinavica. Supplementum, 640, 88–91.
Serpell, J. A. (1996). Evidence for an association between pet behavior and owner attachment
levels. Applied Animal Behaviour Science, 47(1), 49–60.
Serpell, J. A. (2019). How happy is your pet? The Problem of subjectivity in the assessment of
companion animal welfare. Animal Welfare , 28(1), 57–66.
Shabelansky, A., Dowling-Guyer, S., Quist, H., D’Arpino, S. S., & McCobb, E. (2015). Consistency
of shelter dogs’ behavior toward a fake versus real stimulus dog during a behavior
evaluation. Applied Animal Behaviour Science, 163, 158–166.
Sherman, B. L., Gruen, M. E., Case, B. C., Foster, M. L., Fish, R. E., Lazarowski, L., DePuy, V., &
Dorman, D. C. (2015). A test for the evaluation of emotional reactivity in Labrador
retrievers used for explosives detection. Journal of Veterinary Behavior, 10(2), 94–102.
Shih, H. Y., Paterson, M. B., Georgiou, F., & Phillips, C. J. (2021). Do canine behavioural
assessments and characteristics predict the human-dog interaction when walking on a
leash in a shelter setting? Animals, 11(1), 0026.
Shiverdecker, M. D., Schiml, P. A., & Hennessy, M. B. (2013). Human interaction moderates
plasma cortisol and behavioral responses of dogs to shelter housing. Physiology &
Behavior, 109, 75–79.
Sloan Wilson, D., Clark, A. B., Coleman, K., & Dearstyne, T. (1994). Shyness and boldness in
humans and other animals. Trends in Ecology & Evolution, 9(11), 442–446.
133
Smaldino, P. E., Palagi, E., Burghardt, G. M., & Pellis, S. M. (2019). The evolution of two types of
play. Behavioral Ecology, 30(5), 1388–1397.
Smith, S. M., & Vale, W. W. (2006). The role of the hypothalamic-pituitary-adrenal axis in
neuroendocrine responses to stress. Dialogues in Clinical Neuroscience, 8(4), 383–395.
Sommerville, R., O’Connor, E. A., & Asher, L. (2017). Why do dogs play? Function and welfare
implications of play in the domestic dog. Applied Animal Behaviour Science, 197, 1–8.
Stahl, S. M. (2021). Stahl’s essential psychopharmacology: Neuroscientific basis and practical
applications. Cambridge University Press.
Stelow, E. (2018). Diagnosing Behavior Problems: A Guide for Practitioners. The Veterinary
Clinics of North America. Small Animal Practice, 48(3), 339–350.
Stone, H. R., McGreevy, P. D., Starling, M. J., & Forkman, B. (2016). Associations between
Domestic-Dog Morphology and Behaviour Scores in the Dog Mentality Assessment. PloS
One, 11(2), e0149403.
Sundman, A. S., Johnsson, M., Wright, D., & Jensen, P. (2016). Similar recent selection criteria
associated with different behavioural effects in two dog breeds. Genes, Brain, and
Behavior, 15(8), 750–756.
Svartberg, K., & Forkman, B. (2002). Personality traits in the domestic dog (Canis familiaris).
Applied Animal Behaviour Science, 79(2), 133–155.
Svartberg, K., Tapper, I., Temrin, H., Radesäter, T., & Thorman, S. (2005). Consistency of
personality traits in dogs. Animal Behaviour, 69(2), 283–291.
Szydlowski, M., & Gragg, C. (2020). An overview of the current and potential effects of COVID-19
on US Animal Shelters. In AIJR Preprints (157, version 1). University of Exeter.
https://preprints.aijr.org/index.php/ap/preprint/view/157
Tami, G., & Gallagher, A. (2009). Description of the behaviour of domestic dog (Canis familiaris)
by experienced and inexperienced people. Applied Animal Behaviour Science, 120, 159–
169.
Taylor, K. D., & Mills, D. S. (2006). The development and assessment of temperament tests for
adult companion dogs. Journal of Veterinary Behavior, 1(3), 94–108.
Taylor, K. D., & Mills, D. S. (2007). The effect of the kennel environment on canine welfare: A
critical review of experimental studies. Animal Welfare, 16, 435–447.
134
Titulaer, M., Blackwell, E. J., Mendl, M., & Casey, R. A. (2013). Cross sectional study comparing
behavioural, cognitive and physiological indicators of welfare between short and long
term kennelled domestic dogs. Applied Animal Behaviour Science, 147(1), 149–158.
Tod, E., Brander, D., & Waran, N. (2005). Efficacy of dog appeasing pheromone in reducing stress
and fear related behaviour in shelter dogs. Applied Animal Behaviour Science, 93(3),
295–308.
Tuber, D. S., Sanders, S., Hennessy, M. B., & Miller, J. A. (1996). Behavioral and glucocorticoid
responses of adult domestic dogs (Canis familiaris) to companionship and social
separation. Journal of Comparative Psychology, 110(1), 103–108.
Tuyttens, F. A., de Graaf, S., Heerkens, J. L., Jacobs, L., Nalon, E., Ott, S., Stadig, L., Van Laer, E., &
Ampe, B. (2014). Observer bias in animal behaviour research: Can we believe what we
score, if we score what we believe? Animal Behaviour, 90, 273–280.
van Oortmerssen, G. A., & Bakker, T. C. (1981). Artificial selection for short and long attack
latencies in wild Mus musculus domesticus. Behavior Genetics, 11(2), 115–126.
van Rooij, S. J., Stevens, J. S., Ely, T. D., Hinrichs, R., Michopoulos, V., Winters, S. J., Ogbonmwan,
Y. E., Shin, J., Nugent, N. R., Hudak, L. A., Rothbaum, B. O., Ressler, K. J., & Jovanovic, T.
(2018). The role of the hippocampus in predicting future posttraumatic stress disorder
symptoms in recently traumatized civilians. Biological Psychiatry, 84(2), 106–115.
Vanbelle, S. (2016). A New Interpretation of the Weighted Kappa Coefficients. Psychometrika,
81(2), 399–410.
vonHoldt, B. M., & Driscoll, C. A. (2017). Origins of the dog: Genetic insights into dog
domestication. In J. A. Serpell (Ed.), The domestic dog: Its evolution, behavior and
interactions with people (second, pp. 22–41). Cambridge University Press.
Walker, J. K., Dale, A. R., D’Eath, R. B., & Wemelsfelder, F. (2016). Qualitative Behaviour
Assessment of dogs in the shelter and home environment and relationship with
quantitative behaviour assessment and physiological responses. Applied Animal
Behaviour Science, 184, 97–108.
Walker, J. K., Waran, N. K., & Phillips, C. J. (2014). The effect of conspecific removal on the
behaviour and physiology of pair-housed shelter dogs. Applied Animal Behaviour
Science, 158, 46–56.
Wechsler, B. (1995). Coping and coping strategies: A behavioural view. Applied Animal Behaviour
Science, 43(2), 123–134.
135
Weiss, E. (2007). SAFER Guide and Forms [pdf file]. American Society for the Prevention of
Cruelty to Animals: ASPCA. Retrieved from https://aspcapro.org/sites/default/files/safer-
guide-and-forms.pdf
Wells, D. L. (2004). A review of environmental enrichment for kennelled dogs, Canis familiaris.
Applied Animal Behaviour Science, 85(3), 307–317.
Wells, D. L. (2009). Sensory stimulation as environmental enrichment for captive animals: A
review. Applied Animal Behaviour Science, 118(1), 1–11.
Wells, D. L., Graham, L., & Hepper, P. G. (2002). The influence of auditory stimulation on the
behaviour of dogs housed in a rescue shelter. Animal Welfare, 11, 385–393.
Wells, D. L., & Hepper, P. G. (2000). The influence of environmental change on the behaviour of
sheltered dogs. Applied Animal Behaviour Science, 68(2), 151–162.
Wenstrup, J., & Dowidchuk, A. (1999). Pet overpopulation: Data and measurement issues in
shelters. Journal of Applied Animal Welfare Science: JAAWS, 2(4), 303–319.
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. Applied Animal Behaviour Science, 141(3), 158–172.
Woodruff, K., & Smith, D. R. (2020). An estimate of the number of dogs in US shelters in 2015
and the factors affecting their fate. Journal of Applied Animal Welfare Science: JAAWS,
23(3), 302–314.
Zeigler, H. P. (1964). Displacement activity and motivational theory: A case study in the history
of ethology. Psychological Bulletin, 61(5), 362–376.
136
APPENDIX A: ASSESSMENT FORM DRAFTS
A.1: Draft 1
137
A.2: Draft 2
138
A.3: Draft 3
139
A.4: Draft 4
140
A.5: Draft 5
141
A.6: Draft 6
142
A.7: Draft 7
143
A.8: Draft 8
144
A.9: Draft 9
145
A.10: Draft 10
146
A.11: Draft 11
147
B.1: Anxiety Score
APPENDIX B: PILOT DOCUMENTS
Ethogram of anxiety-associated behaviors (1). The Anxiety Score, observed during the 2-minute
acclimation in Draft 9 of the assessment form, is calculated based on the frequency and
intensity of observed anxiety-associated behaviors.
Anxiety Score
1. None; No anxiety behaviors
2. Occasional and mild
3. Some of the time and mild/ occasional and moderate
4. Most of the time and mild/some of the time and moderate/ occasional and severe
5. Some of the time and severe/ most of the time and moderate
6. Most of the time and severe
Anxiety Score Ethogram (1)
Active anxiety-associated behaviors
Active responses
Startling, bolting, vigilance, scanning, pacing, aimless
activity, stereotypic circling, retreat/escape attempts,
digging, climbing
Inactive anxiety-associated behaviors
Decreased activity
Lowered body postures
Autonomic/conflict behaviors
Freezing, positioning in corners/against wall/at door
Crouching, tail tucking, ears back
Panting, shaking, salivating, yawning, lip licking,
elimination
(1) Sherman, B. L., Gruen, M. E., Case, B. C., Foster, M. L., Fish, R. E., Lazarowski, L., DePuy, V., &
Dorman, D. C. (2015). A test for the evaluation of emotional reactivity in Labrador
retrievers used for explosives detection. Journal of Veterinary Behavior, 10(2), 94–102.
148
B.2: Kennel Cards
Purina fecal scoring chart. (2021). Purina Institute.
https://www.purinainstitute.com/sites/default/files/2021-04/fecal-chart.pdf
149
APPENDIX C: TRAINING DOCUMENTS
C.1: Protocol
Inclusion:
• Healthy dogs 12 weeks or older maintained in solitary housing at Capital Area Humane
Society (Lansing, MI) overnight for at least one night prior to assessment.
• This includes but is not limited to strays, owner-surrenders, transfers, returns, and
heartworm positive dogs.
Exclusion:
• Dogs group housed
• Dogs less than 12 weeks of age
• Dogs that appear ill (e.g., vomiting, diarrhea, difficulty breathing, extreme lethargy,
difficulty locomoting)
• Dogs that are removed from home by authorities (i.e., cruelty cases)
Data collection:
• Video assessments
• Assessment forms filled out in-field by assessor and observer
• For a sub-population of enrolled dogs, accelerometer data
• Pet Point records
Assessment Schedule:
• Assessments will take place Mon, Tues, Thurs, Fri from 1pm-4pm.
o Assessor will contact intake staff for a list of eligible dogs by 10am the day of
assessments.
o A maximum of five dogs will be assessed daily.
â–ª
If there are more than five dogs that are eligible, 5 will be randomly
chosen. Those that are not assessed are eligible on the following
assessment date.
o Enrolled dogs will be assessed once.
• Assessments will take place in the education room.
o
If the room is unavailable, assessments will take place outside in the fenced yard
by shelter intake entry.
• For 80% power, we would like to assess 46 dogs inside and 46 dogs outside. This would
equate to a minimum of 10 assessment days inside and 10 outside (5 weeks total).
Assessment Protocol:
• The assessor will conduct the assessment and handle the dog. The assessor will always
be the same female person.
o The assessor will record the assessment using a GoPro camera attached to a
Chesty mount.
o The assessor will rate the dog using the assessment form.
150
• The observer will witness the assessment and attempt to remain “invisible†to the dog
during sub-tests (minimize noise and movement, ignore the dog). The observer will
always be female.
o The observer will record the assessment using a GoPro camera attached to a clip
mount.
o The observer will rate the dog using the assessment form.
â–ª
If the observer cannot see the dog’s response, she will leave the category
blank.
o Observer will be as consistent as possible with her location during tests.
â–ª During In-Kennel: Exit, observer will follow the assessor, stopping
approximately 4’-0†away from the kennel door on the far side of the
walkway. The aim is to maintain a good sightline to observe the dog while
also standing as far away as possible to minimize influence on the dog’s
behavior.
â–ª During In-Kennel: Return, observer will proceed ahead of assessor to
position herself approximately 4’-0†away from the kennel door at the
opposite side of the assessor's approach. The aim is to position oneself
for a good view of the return while distancing as much as possible to
minimize influence on the dog’s behavior.
• While the goal is to remain at a distance during the time intervals
that are scored, the observer will not go so far ahead as to lose
sight of the assessor. If the assessor is having a difficult time
returning the dog to the kennel, she may ask for the observer's
assistance. Observer will not intervene unless asked.
• Assessments will consist of several sub-tests, each with a unique set of categories.
• Sub-tests will be conducted by the assessor in sequence except for ‘Activity’ which is
scored during sub-tests ‘In-Kennel: Begin’ and ‘In-Kennel: Return’.
• Room Setup:
o Cloth mat for Settle
o Water bowl
o Bluetooth speaker positioned in the middle of the room under a plastic crate
o Comfy chair for observer
o Low camp chair for assessor
o Metal folding chair with attached rope near tethered leash that is secured in
door jamb
o Rubber floor dots to mark location of fake dog, and location of assessor during
distraction test
o Toys located where dog cannot reach (tennis ball, rope, stuffless soft toy)
o Fake dog behind cardboard visual barrier
• Dog will have 1 minute of acclimation in room off-leash. Assessor and Observer will sit in
a relaxed fashion and encourage interaction with the dog if the dog approaches either
human. Assessor will sit in the center of the room during acclimation. During
acclimation, A will connect her phone to the Bluetooth speaker.
151
Sub-Tests:
1. In-Kennel: Begin
a. Assessor (A) and observer (O) will start their cameras and A will say the time
using a wristwatch. Before beginning approach, A and O will show the form
header to the camera lens to identify the dog, date, and time of the assessment.
b. A will approach the dog’s kennel door and pin the “Research in Progress†sign to
the kennel door.
c. Immediately after A calls the dog in a friendly manner, location after approach,
presentation with closed door, and overall body language is observed.
d. A will offer the dog a treat through the kennel door for scoring of ability to take
treat. If the dog does not take the treat, A will toss the treat into the kennel.
e. A will then open the kennel door (observing presentation with open door) and
use a slip lead to guide the dog out of the kennel. Latency to exit is scored from
when the door is open until the dog has all four paws past the door threshold.
i. If the dog does not exit from the front of the kennel after 90 sec., A will
attempt to guide the dog out of the kennel through the back door. If after
90 sec. trying at the other door the dog is still unable to exit, the
assessment will end.
2. Play
a. A will grab toys (a rope, a ball, and a stuffing-free soft toy) and toss one at time
slightly out in front of her then start the timer. A will use the toys to encourage
the dog to play. Approach and overall body language are scored within the first
30 sec.
b. A will continue to encourage play for the following 2.5 min. If A ends play early,
she will verbally cue O.
c. After a total of 3 min of play time has passed, A will pick up the toys to observe
ease of putting the toys away. If the dog is engaged with a toy, A will offer a
trade for a treat.
3. Settle
a. A will show the dog a large Milkbone and then place it on a floor mat. Time starts
once the treat is on the ground. A will then sit down in the low camp chair. A will
ignore the dog for the duration of the settle test even if the dog solicits
attention.
b. Response to treat will be observed for the 2 min. duration, as well as
miscellaneous behaviors.
c. At 1:45, activity will be observed and scored until 2:00 min have passed.
4. Acoustic Startle
a. Without getting up from the chair, A will use a phone app to play a loud buzzer
noise.
b. The immediate response of the dog will be observed from the instant the noise
begins to 3 seconds after the noise has ended.
5. Visual Startle
a. A will clip the dog’s collar to a leash and walk the dog over to the tethered leash,
connecting the dog’s leash to the tethered leash at the loose end. A will then pull
152
the rope attached to the metal folding chair so that the chair scoots toward her
location by the dog approximately 6â€. Reaction will be scored while the chair is
moving up to 3 seconds after it has stopped.
6. Distraction
a. A will leave the dog tethered and remove the fake dog from behind a visual
barrier, walking it to the center of the puzzle floor mat, keeping the fake dog
between herself and the dog. Overall body language is scored while the fake dog
stimulus is moving.
b. Once the fake dog is positioned on the dot, A will walk to the location where the
dog’s leash is attached to the tethered leash and call to the dog, offering a treat.
A will attempt to obtain the dog’s attention for 5-10 sec after which she will
unclip the leash at the handle, leaving it attached to the dog’s collar.
c. A will remain where she is, and observe the dog’s behavior for 5-10 seconds,
after which she will walk the fake dog for an additional 6’-0â€. Behavior while the
dog is off leash is not scored but written down anecdotally.
7. In-Kennel: Return
a. A will return the dog to its kennel using the clipped leash (not the slip lead).
b. Once A is approximately 5’-0†from the dog’s kennel entry, approach to kennel
entry is observed.
c. Once the dog is in the kennel, A will give the dog a medium size Milkbone while
removing the leash followed by closing the kennel door.
i. If the dog has not entered the kennel 10 sec. after the door has opened,
the Milkbone can be used as a lure.
8. Activity
a. The activity category represents miscellaneous behaviors that can correlate with
maladaptive coping. These behaviors are circled if observed during either the In-
Kennel: Begin or In-Kennel: Return tests.
153
C.2: Picto-ethogram
Images adapted from “Doggie Language†by Lili Chen (doggiedrawings.net/freeposters);
Downloaded under Creative Commons License (CC BY-NC-ND 3.0 DEED)
Image 1
Image 2
Image 3
Lowered head and body, backward lean, tail
tuck, flattened ears
Lying down/
lowered head/unresponsive
Forward lean with lowered head and body,
outstretched neck, possible tail tuck
Image 4
Avoidant, standing (with or without paw lift)
Image 5
Avoidant, sitting
Image 6
Exaggerated yawn or panting with facial
muscle tension
Image 7
Relaxed (no muscle tension in body or face)
Image 8
Loose/wriggly body, no muscle tension in
body
154
Image 9
Bow
Image 10
Image 11
Panting with
no facial muscle
tension
Alert (some muscle tension), leash tension
may be present
Image 12
Head tilt
Image 13
Alert (some muscle tension), lying or sitting
Image 14
Image 15
Front paws leave ground at same
time/jumping or hard pulling with skittering
of paws for multiple steps
Extreme tension/showing teeth/backward
lean/
cowering
155
Image 16
Extreme tension/showing teeth/forward
movement
Image 17
Shake-off
Image 18
Image 19
Lip lick
Sniffing
Image 20
Scratching
156
C.3: Ethogram
Images adapted from “Doggie Language†by Lili Chen (doggiedrawings.net/freeposters);
Downloaded under Creative Commons License (CC BY-NC-ND 3.0 DEED)
HAI PROMPT
HAI
1
Location
after
approach
HAI
1
Presentatio
n with
closed door
RESPONSE DESCRIPTION
OBSERVATION PERIOD
RESPONSE
Assessor approaches the front of the kennel and calls to the dog; observe
the predominant behavior after the dog has noticed the assessor up until
assessor offers treat
Front
Dog remains near the front of the kennel
Dog remains in the far back of the kennel
Dog remains in the middle third of the kennel
Dog moves from front to back or vice versa more than
once
Back
Middle
Not stationary
Observe the dog while assessor is directing attention toward the dog after
approach up until assessor offers treat
Friendly/
attentive/
neutral
Dissociated
(fearful/
stressed)
Aware (fearful/
stressed)
Hyper-active
Dog remains at the front of the run; orients gaze toward
assessor; may be standing, sitting, or lying; moderate to
no muscle tension in the body and face; tail base and
ears in neutral to moderately perked position; dog may
put paws on front door or make short duration
vocalizations; dog may shift weight around with loose
wriggly body motions; dog may move around and
vocalize but does not repeatedly exhibit these behaviors
for the duration of the observation
Dog makes no attempt to move or moves very slowly,
requiring a lot of encouragement from the assessor; dog
may be lying or sitting with head up or down, eyes open
Dog may be cowering with tucked tail or leaning forward
with head kept low; dog may or may not retreat to the
back of the kennel; dog may be panting with tense facial
muscles and wide open eyes; dog may keep head low
and turn head to the side to look at
environment/human; dog may flatten ears down and
look down, moving the head lower; dog may lift front
paw or show belly/chest to assessor
Dog may be jumping, pacing, or spinning; dog may or
may not be vocalizing; dog might paw repeatedly in
quick movements at the floor or door of the kennel; dog
may remain with front paws on door while continuously
vocalizing; dog may make tight back and forth
locomotion
157
Aggression
Dog might lunge forward with teeth bared; dog might
snap jaws or growl; high muscle tension in body and
face; dog might freeze with stiff body and whale eye
(wide open eyes with whites showing)
Not enough of the dog is visible to be able to interpret
the body language
Hidden from
view
Circle one or more images that represent the overall body language
observed with kennel door closed
See document: Body Language Picto-
ethogram
HAI
1
Overall
body
language
HAI
1
Ability to
take treat
Observe dog’s reaction to treat ending when the assessor moves to open
the door
Consumes treat
from hand
Takes treat through kennel door from hand and
consumes
HAI
1
Location
with open
door
HAI
1
Presentatio
n with
open door
Takes treat from
hand but does
not consume
Consumes treat
from floor
Takes treat through kennel door from hand and drops it
to the floor but does not consume
Eats treat from the floor after assessor has dropped or
tossed it into the kennel or after the dog has dropped it
to the floor
Does not take the treat from the hand nor consumes it
from the ground
No interest in
treat
Observe the dog's location while opening the kennel door until the lead is
attached to the dog
Front
Back
Middle
Not stationary
Observe the dog's behavior while opening the kennel door until the lead is
attached to the dog
See ‘Location with closed door, HAI 1’
See ‘Presentation with closed door, HAI 1’
Friendly/
attentive/
neutral
Dissociated
(fearful/
stressed)
Aware (fearful/
stressed)
158
HAI
1
Latency to
exit
Hyper-active
Aggression
Hidden from
view
Observe the dog from when attaching the lead until all four paws are past
the kennel door threshold. Front/primary kennel exit is attempted for 90
sec. followed by a 90 sec. attempt at back/alternate exit if needed.
Dog might move excitably to the extent that the
Bolts
assessor has difficulty securing the dog; dog may burst
from the kennel with sufficient speed and/or strength to
cause assessor to counterbalance against the force of
the dog; dog may push nose through door opening as
assessor is trying to secure the dog; dog may jump with
front paws leaving the ground while assessor is trying to
secure the dog; dog may leap past door threshold
before assessor has moved out of the way; dog may
immediately react to neighbor dog in the instant that
they are passing through threshold; dog may push into
assessor or jump into/over assessor; dog may lower
head to push forward before assessor is out of the way
Assessor is easily able to secure dog and dog exits with
neutral body in a timely manner
Dog may pull back from the lead; dog may sit in kennel
after being secured with lead; dog may require treats to
accept lead or exit kennel
Dog won't leave the kennel from neither the front nor
the back
Dog exhibits severe aggressive behaviors that cause the
assessor to make a judgement call to leave the dog in
the kennel and end the assessment
Unsafe to allow
exit
Refuses to exit
Delayed
Efficient
HAI
2
Initial
reaction to
play
The dog's reaction during the first 30 sec. of play-time is observed.
Engages in play
Approach - does
not engage
Dog approaches human or toys and engages in play
behavior. Play behavior could include bowing, grabbing
toys in mouth with loose/wiggly body, chasing, engaging
in fetch, tug, or keep-away; dog may or may not shake
the toys; dog may individually play or socially play with
the assessor
Dog approaches human or toys but does not engage in
play behavior. Dog may sniff the toys or grab the toys
with the mouth but does not engage with play behavior;
dog may solicit attention from the human with the
159
motivation for interaction that does not involve play
(e.g., petting)
No approach -
does not engage
Dog does not approach human or toys. Approach is
defined as the dog coming close enough to smell or
touch object of attention
Circle one or more images that represent the overall body language
observed during the first 30 sec. of play time
HAI
2
Overall
body
language
See document: Body Language Picto-
ethogram
HAI
2
Ease of
putting
away toys
HAI
2
Was play
ended
early?
How much effort is required to redirect the dog away from the toys when
play-time is over?
Easy - not
engaged with toy
Easy with trade
for treat
Dog has no interest in the toys and assessor is able to
pick them up and put them away with no effort required
Dog is interested in the toys but will calmly allow
assessor to pick them up and put them away in
exchange for a treat
Dog does not stop playing with the toys and may
repeatedly grab at the toys or assessor as they are
getting picked up. Assessor requires multiple distractions
(scattered treats) to successfully put away the toys
Difficult -
multiple trades
required
Were the toys put away before 3 min. had passed?
Yes - over-
stimulated
Assessor makes a judgment call to end play before the 3
min. have passed due to the dog's high level of arousal;
dog may be hard mouthing the assessor, humping, or
tugging at clothing
Assessor makes the judgment call to end play before the
3 min. have passed due to severe and sustained anxiety-
associated behaviors such as escape behaviors
Assessor does not put the toys away until 3 min. have
passed
Yes - fearful
No
160
HAI
3
Response
to treat
HAI
3
Misc.
behavior
Observe response to treat beginning immediately after treat is placed on
ground for the duration of 2 minutes. Select predominant response to
treat or the most progressive behavior in terms of consumption
(consuming while sitting/lying being the most progressive followed by
consuming while standing with the least progressive being no
consumption)
Consumes while
standing or
moving
Consumes while
lying/sitting or
plays with treat
Ignores
treat/brief
interest or holds
treat in mouth
without
consuming
Dog will break the treat apart with teeth and swallow
but remains standing; dog may or may not eat all of the
treat; dog may break pieces of treat while locomoting
and consume
Dog will lie down or sit to bite the treat and swallow;
dog may grip treat between the front paws; dog may or
may not eat all of the treat; Dog engages with the treat
as if it were a toy but does not chew or swallow; dog
may consume while in a bowed position
Dog never approaches treat during the assessment; dog
might follow assessor to the treat location but does not
orient toward the treat after it is placed on the ground;
Dog will approach and sniff the treat at least once during
the assessment; dog might approach and sniff treat
multiple times with or without briefly picking it up and
putting it back down without carrying it to another
location; Dog will pick up the treat in mouth but does
not consume any portion of it; dog may pace with treat
in mouth; dog may pick up treat in mouth and put it
back down; dog does not play with the treat
Over the course of 2 min., does the dog display specific behaviors that may
be indicative of a negative emotional state?
Pawing at exit
Pacing
Whining
Barking
Other (write)
None
Dog will stand at exit (door, gate) and use one or both
paws to scratch at the door or ground
Walks back and forth or along perimeter of yard at a
brisk pace with or without treat in mouth
High pitch, long vocalization with closed mouth
Loud, short vocalizations with open mouth
Note any observations of interest
No behavior outside of what is already captured in the
form is of interest
HAI
3
Activity
(last 15
sec.)
Observe the activity and manner of the dog for the last 15 sec. of the 2
min. settle test
Neutral -
stationary
Dog remains in one spot engaging in dog-typical
activities
161
HAI
4
Reaction to
acoustic
startle
HAI
5
Reaction to
visual
startle
Neutral - active
Dog is locomoting calmly and engaging in dog-typical
activities
Dog remains in one spot and exhibits anxiety-associated
behaviors
Stressed -
stationary
Stressed - active Dog is locomoting and exhibiting anxiety-associated
behaviors
Observe the reaction of the dog while a buzzer noise is played up until 3
sec. after the sound has ceased.
Retreat
Dog takes multiple steps backward or quickly turns and
rapidly moves away from stimulus; dog may show
escape behaviors by hiding or rapidly moving toward an
exit
Dog reacts automatically to stimulus with full body
movement
Immediate
approach
Freeze - upright Dog immediately stops what he/she is doing and does
not move, holding the body and head still in an upright
posture for the duration of the noise and the
subsequent 3 sec.; see also 'anxiety-associated
behaviors, freeze'
Dog immediately stops what he/she is doing and does
not move, holding the body and head still in a lowered
posture for the duration of the noise and the
subsequent 3 sec.; see also 'anxiety-associated
behaviors, freeze'
Freeze - cower
Stops
briefly/orients
Flinches/startles Dog alters activity in acknowledgement of the stimulus
by flinching (dog reacts automatically to stimulus with
tense facial muscles, lowering ears, and/or stepping
backward) or by startling (automatic full body reaction
such as jumping, or lowering body) with quick recovery
to movement; dog may show additional anxiety-
associated behaviors after immediate reaction
Dog alters activity in acknowledgement of the stimulus
by momentarily pausing with rapid recovery to
movement, by turning head toward stimulus, by perking
ears, or by turning head or body toward stimulus; dog
will have neutral to attentive body language
Dog does not acknowledge the stimulus
No reaction
Observe the immediate reaction of the dog to a sudden moving chair up
until 3 sec. after the chair has stopped moving
Retreat
Immediate
approach
See ‘Reaction to acoustic startle, HAI 4’
162
Freeze - upright
Freeze - cower
Flinches/startles
Stops
briefly/orients
No reaction
Select the most predominant body language of the dog while the dog is
tethered and the fake dog is moving. End observation when fake dog stops
moving.
Image A
Lowered head and body, backward lean, tail tuck,
flattened ears
HAI
6
Overall
body
language
(while
stimulus is
moving)
Image B
Forward lean with lowered head and body, outstretched
neck, possible tail tuck
Image C
Relaxed (no muscle tension in body or face)
Image D
Alert (some muscle tension), leash tension may be
present
Image E
Front paws leave ground at same time/jumping or hard
pulling with skittering of paws for multiple steps
Image F
Extreme tension/showing teeth/backward lean/
cowering
163
Image G
Extreme tension/showing teeth/forward movement
HAI
6
Can you
obtain the
dog’s
attention
while
stimulus is
stationary
(5-10 sec.)
HAI
6
Note
behavior
after let off-
leash
HAI
7
Approach
to kennel
entry
Once assessor is positioned near the dog and calls to it, offering a treat, is
the assessor able to obtain the dog's attention within 10 sec.?
No - oriented
toward stimulus
No - retreat from
stimulus
Dog remains oriented toward fake dog and does not
approach the assessor or approaches briefly after which
the dog immediately goes back toward the fake dog; dog
may look at assessor briefly but not approach assessor;
dog may run to the assessor and quickly consume treat
but immediately move toward the fake dog
Dog keeps distance between him/herself and the fake
dog and does not approach the assessor or will stand
near the assessor but ignore the treat and remain
looking at the fake dog while exhibiting anxiety-
associated behaviors
No - uninterested Dog does not approach the fake dog nor the assessor
Yes
and performs dog-typical activities with a neutral body
language
Dog approaches assessor and orients toward assessor;
dog may or may not consume treat; dog shows neutral
or attentive body language toward assessor; dog may
look back at the fake dog or move toward the fake dog
after giving attention to the assessor in a calm manner
After dog is untethered observe approach behavior; observe the dog's
behavior while walking the fake dog when dog is untethered
Write-in
response
Observer behavior of dog when within approximately 5'-0" of the kennel
entry up until all four paws have passed the entry threshold
Write-in observed approach behaviors and interactions
with the fake dog
Actively pulling
towards entry
Actively pulling
away from entry
Dog enters kennel with hard pulling on the lead and
does not pause to allow assessor to remove lead but
continues to pull causing difficulty in detaching the lead
Dog pulls past entry on either side of the door; Dog
might back-up resulting in a pull backward on the leash;
dog may require high reward treat to be coaxed into
kennel
164
HAI
7
In-kennel
behavior
post-return
Actively
pulling/darting in
multiple
directions
Dog moves rapidly to end of leash and abruptly changes
direction, repeating this movement pattern in multiple
directions (toward or away from kennel entry); dog may
enter kennel efficiently after darting in the vicinity of the
entry
Dog may divert from crossing the kennel threshold or
pause outside the kennel until lured in with treat
Dog enters the kennel in a timely manner
Requires
encouragement
Cooperative
Observe the behavior of the dog from when it is fully in the kennel to when
the door is closed and latched
Hyper-active
Dog exhibits high energy movement as assessor
attempts to close the door; dog may vocalize while
assessor closes the door; dog may paw repeatedly at the
door or floor; dog may dart back and forth in kennel;
dog may excessively jump or spin in kennel
Dog has lowered body posture; dog may have tucked tail
and/or flattened ears; dog may quickly move to the back
of the kennel and remain; dog may appear despondent
Dog exhibits neutral or attentive body language and
engages in dog-typical activities
Fearful
Calm
HAI
8
In-kennel
misc.
behavior
While observing the dog when it is in the kennel, select behaviors if
observed
Spinning
Dog turns multiple quick, tight circles
Dog repeatedly jumps, frequently having all four paws
off the ground
High pitch, long vocalization with closed mouth
Excessive
jumping
Whining
Excessive barking Multiple loud, short vocalizations with open mouth
Walks back and forth along kennel run at a brisk pace
Pacing
Dog paws at door or ground with both front paws as if
Frantic pawing at
digging
door
Dog has defecated in kennel and the fecal matter has
Smeared feces in
been disturbed from its original shape
kennel
Note any observations of interest
Other (write)
While moving the dog through the shelter, select behaviors if observed
Dog directed
reactivity (e.g.,
barrier
aggression)
Other (write)
Dog bares teeth toward another dog and lunges toward
the dog; dog may growl or bark; dog may bite at the
fencing of another dog's kennel
Note any observations of interest
HAI
8
Out-of-
kennel
misc.
behavior
165
C.4: Glossary
Aggression
Anxiety-
associated
behaviors
Growling, snarling, snapping, lunging with lips drawn away from teeth,
hard stare with freeze
Fight: body stiffens, facial tension, eyes stare, may struggle, may growl and
bare teeth, tail suddenly stops moving, mouth closes, lips tense, begins to
growl, lunge forward, bite or snap (1)
Flight: body stiffens, facial tension, eyes wide, avoids eye contact, turns
head away, turns whole body away, creeping, crouching, tail tucked, moves
away, tries to escape, retreats as far as possible, roll over to evade contact
(1)
Fret/fidget: restless, eyes wide, lip licking, looking around, scanning,
pacing, pawing, sniffing, blinking, shaking off (1)
Freeze: body stiffens, facial tension, freezes in place, may tremble, eyes
wide, mouth closed, bark, whine, or scream, shut down, refuses treats,
may appear to be "sleeping", helpless (1)
Includes active and inactive anxiety-associated behaviors (2)
startling, bolting, vigilance, scanning, pacing, aimless activity, stereotypic
circling, retreat/escape attempts, digging, climbing (2)
Anxiety-
associated
behaviors, active
Anxiety-
associated
behaviors,
inactive
Approach
Approach, during
play
Attention, obtain Dog approaches human and remains near the human while leash is
decreased activity: freezing, positioning in corners/against wall/at door (2)
lowered body postures: crouching, tail tucking, ears back (2)
autonomic/conflict behaviors: panting, shaking, salivating, yawning, lip
licking, elimination (2)
Dog moves toward the object of interest
Dog moves toward toys or humans and gets close enough to sniff or touch
Attentive
Aware
(fearful/stressed)
Barrier
aggression
Calm
Dissociated
(fearful/stressed)
Dog-typical
activities
Dog-directed
reactivity
removed (dog may or may not eat the treat that is offered)
Orients to human or environment with alert gaze and neutral body
language
Dog is primarily exhibiting anxiety-associated behaviors
Aggression directed at dogs while separated by a physical barrier
Dog is friendly, attentive, or neutral
Dog is lethargic and appears apathetic
Grooming, eating, drinking, exploring using smell, sight, sound, touch, and
taste, playing, resting, sleeping, socializing
Aggression directed at dogs
166
Engage, during
play
Fearful
Flinches
Freeze
Friendly
Frustration
Hyper-active
Mounting
Neutral
Neutral body
language
Over-stimulated
Piloerection
Play behaviors
Posturing
Retreat
Startle
Stressed
Uninterested
Dog exhibits a sequence of play behaviors
See Dissociated and Aware (fearful/stressed)
Dog reacts automatically to stimulus with tense muscle movement in face,
head, or shoulders
Dog immediately stops what he/she is doing and does not move, holding
the body and head still. See also 'anxiety-associated behaviors, freeze' with
slow recovery to movement
Orients to human with soft gaze no muscle tension in body and face; body
may be neutral or loose and wiggly
Dog is obsessively directing energy into vocalizing, mouthing, jumping,
digging and/or humping
Dog is expending high levels of energy and/or exhibiting severe frustration
Dog leans chest into object or human and grips with the front legs
Dog is engaging in dog-typical activities with neutral body language
Overall posture is relaxed or loose, no tension in the mouth and lips, hair
coat lies flat, tail base is neither overtly raised or tucked, eyelids are soft,
stance is balanced on all four paws, ears are held in natural position
Dog is aroused beyond his/her ability to appropriately engage with the
human or environment; exhibiting frustration
The hair along the back of the dog is raised in comparison to the natural
hairline of the dog (i.e., raised hackles)
Bowing, prancing with toy in mouth, chasing toys with loose movement,
hip bumps (side approach hip sway against human or dog) with loose body
and relaxed jaw
Dog holds head high and over object with hard staring at object, oriented
so that the chest directly faces the direction of the gaze
Dog takes multiple steps backward or quickly turns and rapidly moves away
from stimulus
Dog reacts automatically to stimulus with full body movement (e.g., jump,
crouch, step backward)
See Dissociated and Aware (fearful/stressed)
Dog ignores human or stimulus and engages in dog-typical activities with
neutral body language
(1) Fear Free shelter course, module 2c. (n.d.). Retrieved 2022, from
https://fearfreeshelters.com/program/
(2) Sherman, B. L., Gruen, M. E., Case, B. C., Foster, M. L., Fish, R. E., Lazarowski, L., DePuy,
V., & Dorman, D. C. (2015). A test for the evaluation of emotional reactivity in Labrador
retrievers used for explosives detection. Journal of Veterinary Behavior, 10(2), 94–102.
167
APPENDIX D: ASSESSMENT SCORE RUBRIC
HAI
PROMPT
RESPONSE
PROMPT SCORE
HAI 1
In-Kennel:
Begin
Presentation
with closed
door (sc01)
Ability to take
treat (sc02)
Presentation
with open
door (sc03)
Latency to exit
(sc04)
HAI 2
Play
Initial reaction
to play (sc05)
Ease of putting
away toys
(sc06)
Was play
ended early?
(sc07)
Friendly/attentive/neutral
Dissociated (fearful/stressed)
Aware (fearful/stressed)
Hyper-active
Aggression
Hidden from view
Score dependent on response to
‘Presentation with closed door,
HAI 1 (sc01)’
Consumes treat from hand
Takes treat from hand but does
not consume
Consumes treat from floor
No interest in treat
Friendly/attentive/neutral
Dissociated (fearful/stressed)
Aware (fearful/stressed)
Hyper-active
Aggression
Hidden from view
Bolts
Efficient
Delayed
Refuses to exit
Unsafe to allow exit
Engages in play
0
-3
-1.5
1.5
3
null
If
Q1=0 pts
0
0
0
0
0
If
Q1<0 pts
If
Q1>0 pts
0
-1
-2
-3
0
1
2
3
-3
-1.5
1.5
3
null
3
0
-3
null*
null*
-12 TO +12 HAI 1 SCORE RANGE
0
*Assessment ends
Approach - does not engage
No approach - does not engage
Easy - not engaged with toy
3
-3
-3
0
Easy with trade for treat
Difficult - multiple trades required 3
3
Yes - over-stimulated
Yes - fearful
No
-3
0
-9 TO +9
HAI 2 SCORE RANGE
168
HAI 3
Settle
Response to
treat (sc08)
Consumes while standing or
moving
Consumes while lying/sitting or
plays with treat
Ignores treat/brief interest or
holds treat in mouth without
consuming
Neutral - stationary
Neutral - active
Stressed - stationary
Stressed - active
Activity (last 15
sec.) (sc09)
HAI 4
Acoustic
Startle
Reaction to
acoustic startle
(sc10)
HAI 5
Visual
Startle
Reaction to
visual startle
(sc11)
Retreat
Immediate approach
Freeze - upright
Freeze - cower
Flinches/startles
Stops briefly/orients
No reaction
Retreat
Immediate approach
Freeze - upright
Freeze - cower
Flinches/startles
Stops briefly/orients
No reaction
HAI 6
Distraction
Overall body
language
(while stimulus
is moving)
(sc12)
Image A
Image B
Image C
Image D
Image E
169
3
0
-3
0
0
-3
3
-6 TO +6
-3
1.5
3
-1.5
0
0
0
-3 TO +3
-3
1.5
3
-1.5
0
0
0
-3 TO +3
-2
-1
0
0
1
HAI 3 SCORE RANGE
HAI 4 SCORE RANGE
HAI 5 SCORE RANGE
Images adapted from
“Doggie Language†by
Lili Chen
(doggiedrawings.net/
freeposters);
Downloaded under
Creative Commons
License (CC BY-NC-ND
3.0 DEED)
Can you obtain
the dog's
attention while
on-leash?
(stimulus is
stationary, 5-
10 sec.) (sc13)
Approach to
kennel entry
(behavior
when within
5'-0" of kennel
door) (sc14)
HAI 7
In-Kennel:
Return
Image F
Image G
No - oriented toward stimulus
No - retreat from stimulus
No - uninterested
Yes
-3
3
3
-3
0
0
Actively pulling towards entry
-6 TO +6
-3
HAI 6 SCORE RANGE
Actively pulling away from entry 1.5
Actively pulling/darting in
multiple directions
Requires encouragement
Cooperative
Hyper-active
0
0
3
3
Fearful
Calm
In-kennel
behavior post-
return
(behavior
while
unclipping the
leash to closing
of door) (sc15)
-3
0
-6 TO 6
HAI 7 SCORE RANGE
-45 TO 45 TOTAL SCORE RANGE
170
APPENDIX E: REFINED ASSESSMENT FORM
Overall body language images adapted from “Doggie Language†by Lili Chen
(doggiedrawings.net/freeposters); Downloaded under Creative Commons License (CC BY-NC-ND
3.0 DEED)
Dog:
Date:
Time:
Assessor:
HAI
PROMPT
RESPONSE
HAI 1:
Approach
Presentation with closed
door
Friendly/attentive/neutral
Dissociated (fearful/stressed)
Aware (fearful/stressed)
Hyper-active
Aggression
Hidden from view
Overall body language
(during initial approach
to kennel) – circle all
that apply
Presentation with open
door
Latency to exit
Acclimation off-leash (1-2 min.)
HAI 2: Play
(3 min.)
Initial reaction to play
(first 30 sec.)
Friendly/attentive/neutral
Dissociated (fearful/stressed)
Aware (fearful/stressed)
Hyper-active
Aggression
Hidden from view
Bolts
Efficient
Delayed
Refuses to exit
Unsafe to allow exit
Engages in play
Approach - does not engage
171
PROMPT
SCORE
0
-3
-1.5
1.5
3
NULL
N/A
0
-3
-1.5
1.5
3
NULL
3
0
-3
END
END
0
TBD
Overall body language
(first 30 sec.) – circle all
that apply
Ease of putting away
toys
Response to treat
Circle Y (yes) if behavior
is present or N (no) if
absent
Activity (last 15 sec.)
HAI 3:
Settle with
treat
(2 min.)
HAI 4:
Model dog
Can you obtain the dog's
attention while on-
leash? (model dog is
stationary, 5-10 sec.)
HAI 5:
Return
Approach to kennel
entry (within 5 ft. of
kennel)
TBD
N/A
-3
0
3
3
0
-3
Y / N
Y / N
Y / N
Y / N
0
0
-3
3
3
-3
0
0
-3
1.5
3
0
0
No approach - does not engage
Easy - not engaged with toy
Easy with trade for treat
Difficult - multiple trades required
Consumes while standing or moving
Consumes while lying/sitting or plays
with treat
Ignores treat/brief interest or holds treat
in mouth without consuming
Attention seeking through close
proximity/touch
Vigilance to environment
Vocalizing
Pacing and/or pawing at exit
Neutral - stationary
Neutral - active
Stressed - stationary
Stressed - active
No - oriented toward stimulus
No - retreat from stimulus
No - uninterested
Yes
Actively pulling towards entry
Actively pulling away from entry
Actively pulling/darting in multiple
directions
Requires encouragement
Cooperative
172