**Comparison of Canine Behaviour Scored Using a Shelter Behaviour Assessment and an Owner Completed Questionnaire, C-BARQ**

#### **Liam Clay 1,\*, Mandy B. A. Paterson 1,2 , Pauleen Bennett <sup>3</sup> , Gaille Perry <sup>4</sup> and Clive J. C. Phillips <sup>1</sup>**


Received: 23 August 2020; Accepted: 21 September 2020; Published: 3 October 2020

**Simple Summary:** In shelters, it is usual to conduct a standardised behaviour assessment to identify adoption suitability. The information gathered from the assessment is used to identify the behaviour of the dogs, suitability for adoption and to help to match the dog to an ideal home environment. We investigated if the dogs' behaviour in the home as reported by owners was reflected in the Royal Society for the Prevention of Cruelty to Animals (RSPCA) Queensland behaviour assessment, conducted on the same dogs during a visit to the shelter. A total of 107 owners and their dogs aged 1–10 years were assessed in-home, by the owners, and in the shelter, by a researcher. The owners completed a questionnaire (Canine Behavioural Assessment and Research Questionnaire (C-BARQ)) prior to the standardised behavioural assessment conducted at the RSPCA Queensland. Regression analysis identified positive correlations between the two for fear, arousal, friendliness and anxiousness, identified in in-home behaviour and the behaviour assessment. This research therefore allowed a greater understanding of current canine behaviour assessment protocols used at the RSPCA Queensland in regard to the predictability of behaviour, behavioural problems and the efficiency and effectiveness of testing procedures.

**Abstract:** In shelters, it is usual to conduct a standardised behaviour assessment to identify adoption suitability. The information gathered from the assessment is used to identify the behaviour of the dogs, its suitability for adoption and to match the dog with an ideal home environment. However, numerous studies have demonstrated a lack of predictability in terms of the post-adoption behaviour in these assessments. We investigated if the owners' perception of dogs' behaviour in the home was reflected in the RSPCA Queensland behaviour assessment, conducted on the same dogs during a visit to the shelter. A total of 107 owners and their dogs aged 1–10 years were assessed in-home and in the shelter. The owners of the dogs completed a questionnaire (the Canine Behavioural Assessment and Research Questionnaire (C-BARQ) survey) 1–2 weeks before bringing their dog to the shelter for the standardised behavioural assessment conducted at the RSPCA Queensland. An ordinal logistic regression analysis identified positive correlations for fear, arousal, friendliness and anxiousness, identified in in-home behaviour and the behaviour assessment. Furthermore, the behaviours of friendliness, fearfulness, arousal, anxiousness, and aggression were positively predictive between home behaviour and tests in the behaviour assessment. This research therefore led to a greater understanding of current canine behaviour assessment protocols used at the RSPCA Queensland in regard to the predictability of behaviour, behavioural problems and the efficiency, effectiveness and predictability of current behaviour testing procedures.

**Keywords:** dog behaviour; behaviour problems; behaviour assessment; canines; shelters; predict; home behaviour

#### **1. Introduction**

The Royal Society for the Prevention of Cruelty to Animals (RSPCA) Australia accepted 33,863 dogs to its shelters during the period 2018–2019 [1]. Sources of admitted dogs in Queensland include councils, owner surrenders, humane officer admission (employees of the RSPCA with investigative powers under the Queensland Animal Care and Protection Act 2001) and euthanasia requests [2], with age at admission being variable, but with over 74% adult dogs. Dogs are surrendered for numerous reasons: human-related (unwanted, changed circumstances, financial, owner's health, and ex-commercial/racing), or dog-related (medical and behavioural problems) [3]. After surrender, dogs are housed in the shelter until their suitability for adoption is determined, and if suitable, adopted.

The procedures used to identify dogs suitable for adoption include a medical check, behavioural assessment, in-kennel monitoring, and monitoring by shelter staff when interacting with the dog. Behavioural assessments are the preferred method in many shelters to give an overview of the dog's behaviour for potential adopters [4,5]. They assess the dog's reactions to diverse novel stimuli typical of everyday life situations and their ability to cope in challenging situations [6], usually 3–5 days after entering the shelter [5].

The testing procedures have a risk of both false positives and negatives [7,8], that is, running the risk of falsely identifying a behavioural problem that does not exist or deeming a dog suitable for adoption when it is not. These problems may arise due to the stress experienced by the dog from living in the shelter [9], and because certain behaviours are multifactorial and a test carried out at a single point in time may not be able to accurately capture this behaviour. Few studies have evaluated the effect of the timing of behaviour assessments, for example immediately on shelter admission [10].

Measurements used in the assessments need to be appropriate and meaningful, providing both quantitative and qualitative data [11]. Qualitative measurements include history-taking measures, which provide a reflection of previous home environment and behaviour. Current procedures used by RSPCA Queensland are primarily quantitative measures, which are in line with the behaviour assessments reported in the literature that use a direct measure of behaviour by observing the dog's response to several testing procedures [4,12–17]. Other measures focus on the assessment of behaviours in everyday situations, using a questionnaire for the dog's owner to complete [18–21]. A widely used questionnaire is the Canine Behavioural Assessment and Research Questionnaire (C-BARQ), which includes items focusing on behaviour associated with aggression, fear and anxiety, trainability, excitability, separation, attachment, attention-seeking, and chasing [18]. It has been extensively evaluated and used to validate quantitative behaviour assessments focusing on areas of behaviour issues and service dogs [22–26].

In order to further investigate the accuracy with which behaviour assessments used in shelters identify behaviours exhibited elsewhere, this study adopted a novel approach to help to determine whether previous home behaviours are accurately reflected in these shelter assessments. The study asked owners to complete a validated questionnaire (C-BARQ) about their dog's behaviour and then to bring the dog into a shelter where the dog underwent the standardised behaviour assessment. The aim of this study was to determine if the dogs' behaviour in the home was reflected in the RSPCA Queensland behaviour assessment, conducted on the same dogs during a visit to the shelter.

#### **2. Materials and Methods**

#### *2.1. Ethics*

This study was conducted with the approval of the University of Queensland's Human and Animal Ethics Committees (approval numbers 2018001353 and SVS/290/18, respectively). The study complies with provisions contained in Australia's National Statement on Ethical Conduct in Human Research and with Queensland regulations governing experimentation on humans.

#### *2.2. Subjects*

Companion dog owners from the general public (n: 107) were invited via social media to participate in this study. The RSPCA and the University of Queensland media outlets were used to attract participants. Participants had to have owned their current dog for at least 6 months, be over the age of 18 years and willing to complete a questionnaire and bring their dog into the shelter to undergo a non-invasive behaviour assessment. Participants received an information sheet, and if willing to have their dog participate in the study, they signed a consent form outlining that the testing would be used for research purposes. Each participating dog was allocated a number which was used to tie the C-BARQ and assessments to the same dog. Apart from the consent form, all information was non-identifiable and most of the questions focused on information about the dog, not the owner. Owners of dogs had to complete and submit the C-BARQ questionnaire before an appointment was made for the shelter assessment. C-BARQ focuses on the dog's interactions in numerous situations. The shelter assessment used was the standardised assessment used on all in-coming dogs.

#### Dogs

Dogs were required to be older than 6 months and younger than 13 years of age. Any breed was allowed in the study. Dogs were also required to have no medical conditions nor be on any medication that had the potential to influence behaviour. Dogs previously adopted from shelters were allowed in the study and were initially categorised separately to identify any variability. However, there were no differences between groups, therefore, separate categories were dropped. All dogs were required to be with the owners for at least 6 months.

#### *2.3. Behaviour Assessment*

The dogs were brought into the shelter by their owner for the formal behaviour assessment. It was conducted in a room (4.5 m × 4.7 m) in a separate building, approximately 50 m from the shelter offices and kennels to minimise disturbance. The dogs were initially left in the room by themselves for 15 min to allow them to acclimatise to the room while the researcher watched their behaviour from the next room via a video link (4× Go pro Hero 4 Silver positioned an equal distance apart). The owner waited in an adjoining area for the period of acclimatisation and assessment.

The behavioural assessment used in this study was the standard assessment used by the RSPCA Queensland for shelter dogs. The assessments were conducted, recorded and scored by the lead researcher (LC), who was formally trained in the assessment regimen. Reviewed behaviours included room exploration, leash manners, sociability, tolerance, play behaviour with toys, the response to unusual/unpredictable stimuli, possessive behaviours, toddler and stranger interaction, time alone and social interactions with other dogs [27] (Appendix A). In each test, the dog's behaviours were scored for friendliness, socialisation, fearfulness, arousal and aggressiveness. The assessment comprised nine different tests performed over a 15 min period. The equipment used was in line with the RSPCA Queensland's protocol and included a 1.8 m leash, a tennis ball, a plush squeaky toy, rope, plastic hand on an extend pole, bowl, raw hide or bone, and the combination of wet and dry dog food. The details of the RSPCA Queensland assessment tests can be found in Clay et al. [27]. All the tests were recorded by video (Go Pro Hero 4, Model: HERO4 Black, Manufacture: Hong Kong, China) and reviewed later.

#### *2.4. Owner Questionnaire, C-BARQ*

Owners rated the behaviour of their dog at home based on behavioural interactions in relation to attachment or attention seeking, sociability, touch sensitivity, excitability, chasing, fear, aggression, and separation-related behaviours. The owners' information on their dog's behaviour was categorised into predetermined behavioural categories on a score of 0–4 (Appendix B). The C-BARQ questionnaire used had the 102 question format [24] and was scored on a scale between 0 and 4 (aggression: 0, none—4, serious, separately scored for stranger-, owner-, dog and familiar dog-directed aggression; fear: 0, no fear or anxiety—4, extreme fear, both stranger, non-social and dog fear; separation-related problems: from 0, never, to 4, always; attachment/attention-seeking: from 0, never, to 4, always; touch sensitivity: from 0, never, to 4, always; excitability: from 0, calm, to 4, extremely excitable; chasing, energy, and trainability: from 0, never, to 4, always).

#### *2.5. Behaviour Scoring*

The formal behaviour assessments were scored for dog behaviour during all tests, as described in Clay et al. [27]. The ethogram comprised 48 behaviours, determined following the preliminary observation of dogs during the formal behaviour assessment, classified as either long duration behaviours (for which the duration was recorded) or events (for which the number of occurrences was recorded). The behaviours focused on eight components: activities of the mouth, body, tail position, tail movement, ears, eyes, position in room, and movement (Table 1). The descriptions of each behaviour were presented in a previous study [27]. Behaviour recording was assisted by coding software BORIS [28], which recorded the frequency and duration of each behaviour using continuous input from the coder. Two behaviour variables with no or only one occurrence were discarded: squint and whale eyes. From the coded behaviours, using similar principles to our previous articles [27,29], the proportion of the time and frequency of the five behavioural categories (anxiety, fear, friendliness, arousal, aggression) were derived. The descriptions of each behaviour are presented in Table 1 and their connection to behavioural categories (anxiety, fear, friendliness, arousal, aggression) in Table 2 are based off the literature described in a previous article (27).


**Table 1.** Behaviours of dogs (n = 107) recorded for each body part, as well as the position in the room and movement types.


**Table 2.**The behaviours contributing to the behavioural states fear, anxiety, aggression, arousal, and friendliness.

#### *2.6. Statistical Analysis*

Statistical analysis was conducted using Minitab 18. Behaviours were analysed as the percentage of the total observation time (long duration behaviours) or the percentage of the frequency of occurrence (events) during the overall behaviour assessment and within the individual tests. The C-BARQ questionnaire has predetermined categories that were calculated after the 102 questions were complete. Descriptive analysis was used for behaviour in assessments.

Spearman's rank order correlations were computed between C-BARQ and the formal behaviour assessment variables. As comparisons with 79 other behaviours were made for each behaviour in each test of the behaviour assessment, results were corrected for false discovery using the Benjamini–Hochberg procedure [30]. The Bonferroni correction was rejected as it assumes the independence of the individual tests. The Benjamini–Hochberg procedure ranks the *p* values for each test and compares the *p* values to critical values [(rank/no. tests) × false discovery rate (selected as 0.20 as recommended by McDonald [30]). All *p* values up to the critical one were considered to indicate a significant difference [30].

Ordinal logistic regression was used to compare the temperament/behavioural information from owner-reported temperament/behaviour with derived behaviours from the shelter assessment, both overall and within the different tests. The Benjamini–Hochberg was used to correct for false discovery as with Spearman rank correlations.

#### **3. Results**

#### *3.1. Descriptive Statistics*

The sample included 107 companion dogs (males: 52, females 57, desexed: 103, intact: 6) who were over the age of 6 months and under 13 years (mean: 5 years 3 months). Sources of the dogs included: shelters (44.9%), breeders (23.8%), other (online, private sales, or did not disclose) (11.9%), neighbour, friend, or relative (10.1%), and under 5% were from pet stores or were stray dogs.

A variety of breeds were included in the study, determined by the C-BARQ questionnaire completed by the owners; mixed breeds (19.3%), Border collie (10.1%), Kelpie (8.3%), Staffordshire bull terrier (8.3%), German shepherd (5.5%), Australian cattle dog (3.7%), and Rottweiler (3.7%). All other breeds represented less than 3% of the population of dogs. Mean weight of the dogs was 21.8 ± 1.06 kg.

With respect to the household environment, 64.2% had other dogs in the household; 35.8% were single dog homes. Of the total population, 69.7% of the households had no children and 30.28% had children living in the home. With regard to the living arrangements for the dogs, 80.7% were classified as inside/outside, 12.8% were only inside, 4.6% were only outside and 1.8% had no classification.

#### *3.2. Owner Questionnaire*

All owners completed the C-BARQ questionnaire (107 participants). Many owners indicated that their dogs displayed no signs of fear (score 0) in situations with other unknown dogs (46%), strangers (68%) and non-social interactions (56%), with the second highest occurrence being the dog displaying minimal signs of fear (score 1) in the above situations (Appendix C). When owners did report that some fear was displayed, it was most likely to be dog directed, then non-social and least likely to be stranger directed.

It was mostly reported that little aggression was observed. In particular, owner-directed aggression was very rare, only 5% of owners reported this, and stranger-directed aggression was also quite rare, with only 28% of owners reporting this, and mostly at low levels. However, dog-related aggression (unfamiliar dogs) was relatively common, reported by 60% of owners, but less towards familiar dogs (34% of owners). Separation-related behaviours were even less common, reported by 23% of owners, but attention-seeking, chasing, excitable and energetic behaviours were relatively common, with most owners reporting some occurrence. Touch sensitivity was less common, with most owners reporting that it was never or seldom seen. Dogs were reported to be trainable most of the time, but never always.

#### *3.3. Formal Behaviour Assessment*

In the overall formal behaviour assessment, dogs spent 41.2% of their time in friendly behaviours, 28.4% displaying fear, 14.3% in a state of high arousal, 13.5% displaying anxiousness, and 2.5% in aggression. Considering the frequency of the behaviours, there was a mean of 37.6% incidents of friendly behaviours, 30.3% incidents of fear-related behaviours, 15.4% incidents of high arousal behaviours, 13.7% incidents of anxiety-related behaviours, and 3.5% incidents of aggressive behaviours.

In individual tests, the major behaviours that had the highest occurrences were friendly and fearful, whereas anxiousness, arousal and aggression had lower instances (Appendix D). However, there were higher instances of arousal in the toy interaction test which reflects the purpose of the test.

#### *3.4. Relationships between Owner-Reported Dogs' Behaviour in the Home and Behaviours Derived from the Formal Behaviour Assessment in the Shelter*

All correlations were corrected using Bonferroni correction and varied in strength. Considering the overall behaviour assessment, there were positive Spearman rank correlations between the fear displayed in the assessment and the fear in non-social situations and stranger situations reported by the owner (Table 3). A friendly classification in the shelter assessment correlated negatively with stranger-directed fear reports by the owner. Aggression in the shelter correlated positively with touch sensitivity reports by the owner, both in the overall assessment and in the touch sensitivity test. In the latter test, friendliness correlated with the non-social fear reports by the owner.


**Table 3.** Significant (*p* < 0.01) Spearman rank correlations between the owner-reported dogs' temperament/behaviour in the home and the behaviours derived from the formal behaviour assessment at the shelter.

In the Play interactions test in the shelter, fear correlated positively with stranger-directed and non-social fear and aggression in the home. Friendliness in this test correlated negatively with stranger-directed fear reports by the owner. In the Response to unusual/unpredictable stimuli test in the shelter, fear correlated positively with stranger-directed fear reports by the owner, which also correlated negatively with friendliness in the behaviour assessment. In the Food possession test in the shelter, friendliness correlated negatively with stranger-directed fear, and in the Toddler doll test, fear correlated positively with non-social fear reports by the owner, and aggression correlated with touch sensitivity reports by the owner.

#### *3.5. Predictability of Behaviour Assessment*

In the home environment, dogs whose owners reported low levels of stranger-directed fear had high levels of friendliness in the Overall shelter test and in the Response to Unusual/Unpredictable Stimulus, Food Possession, Stranger, and Toddler doll tests (Table 4). High levels of stranger-directed fear related positively to aggression in the Overall, Play interaction, Response to Unusual/Unpredictable Stimulus and Food Possession tests, to fearfulness in the Touch Sensitivity test and negatively to high arousal in the Toddler doll test. Owner-reported non-social fear and fear in the Exploration of room, Touch sensitivity and Response to unusual stimulus tests were related. Stranger-directed aggression reported by the owner was also related to fearfulness in the Touch sensitivity test. Owner-directed and reported aggression was negatively related to friendliness, fearfulness and high arousal in the Stranger test, and positively related to aggression in that test and the Toddler doll test. Familiar dog aggression reported by the owner was negatively related to friendliness, fearfulness and high arousal in the Toddler doll test and positively related to aggression in that test.

Touch sensitivity reported by the owner was negatively related with friendliness (Overall assessment, Response to unusual stimulus, Toddler doll, Time alone, Dog-to-dog interaction), high arousal (Overall assessment, Toddler doll, Touch sensitivity, Time alone), fearfulness (Touch sensitivity, Dog-to-dog interactions), and anxiety (Response to unusual stimulus, Toddler doll, Dog-to-dog interaction). There was a positive relationship between those related with aggression (Overall assessment, Touch sensitivity, Play interaction, Response to unusual stimulus, Toddler doll tests).

Attachment/attention seeking reported by the owner related negatively with friendliness (Response to unusual stimulus, Toddler doll), fearfulness (Overall assessment, Response to unusual stimulus, Toddler doll, Time alone), high arousal (Overall assessment, Play interaction, Response to unusual stimulus, Toddler doll), anxiety (Response to unusual stimulus, Toddler doll, Time alone). It related positively with aggression (Overall, Response to unusual stimulus, Toddler doll, Dog-to-dog interaction tests).

Excitability related negatively to fearfulness in Touch sensitivity, high arousal in Touch sensitivity, and it related positively to anxiousness in the Exploration of room, high arousal in the Exploration of room, and Time alone tests.

Energetic behaviour was related positively to high arousal in the Exploration of room, and aggression in Dog-to-dog interaction and negatively to friendliness in the Dog-to-dog interaction. Chasing was related negatively to anxiousness in the Toddler doll test.







#### **4. Discussion**

Behavioural assessments are used in the RSPCA Australian shelters to identify behavioural problems, determine suitability for adoption and to monitor the behaviour of each dog over time while in the shelter. The use of the behavioural assessment as a tool in combination with surrender information (home environment, in-home behaviour, and behaviour towards other dogs), veterinary history, in kennel observations, and staff feedback is thought to provide some representation of the dog's behaviour. The behavioural assessment is not being used as a pass–fail tool, rather, it is used as one component of a toolbox to collect information over time. It is important to know how valid it is. The aim of this study was to determine if dogs' home behaviour, measured using information provided by owners using the C-BARQ, was accurately reflected in the standardised RSPCA Queensland behaviour assessment. The study was conducted with dogs owned by members of the general public and therefore not dogs potentially negatively affected by stress due to time in the shelter.

Major themes identified in this study are consistent with the previous findings and results reported in previous studies, particularly in relation to fear, arousal, friendliness, and anxiousness [27,29]. The major tests that were most predictive of behaviour in a home environment were the exploration of room, touch sensitivity, and Response to unusual stimulus in regard to non-social fear. Stranger-directed fear was predictive in tests of touch sensitivity, and response to unusual stimulus response. Touch sensitivity was reflected in the corresponding test in the assessment. Owner-directed aggression was predicted in the stranger and toddler doll tests. Stranger-directed aggression was only identified in touch sensitivity in relation to fear. Excitability and energy were predicted in the exploration of room, touch sensitivity, and time alone tests. Finally, attachment was predicted in the tests related to the response to unusual stimulus, and toddler doll.

Overall friendliness identified during the play interactions, response to unusual stimulus, food possession, stranger, toddler doll and dog-to-dog interactions tests were reflected in the low scoring of the categories of energetic, fear and aggressive-related issues in C-BARQ. Categories of the C-BARQ that were not predicted in the tests were dog rivalry, dog-directed aggression, separation-related behaviours, trainability, and chasing.

There are few studies on the ability of an assessment to reflect previous home behaviour; rather, most literature looks at predicting future behaviour [8,13,14,25,31–35]. In this study, behaviour reported in the home showed a relationship with certain aspects of the behavioural assessment including fear, friendliness, anxiety, arousal and aggression.

The relationship between fear displayed in the assessment and owners' indication of stranger-directed and non-social fear, aligns with previous findings of the predictability of fear [14,36]. In looking at C-BARQ categories, stranger-directed fear and aggression, and non-social fear in the home were related to fear observed in the exploration of room, touch sensitivity, and response to unusual stimulus. Non-social fear, stranger-directed fear, and aggression in the home were associated with increased odds of fearfulness in dogs in the assessment. This consistency of fear responses is to be expected, since the fear response is a manifestation of a survival response in the brain located in the amygdala, with the behavioural response created being very recognisable and easy to identify in all species [37]. Furthermore, the consistency of fear responses indicates a similarity of stimulus features and the demonstration of fearful behaviour requires appropriate environmental stimuli. One might expect to observe some consistency of fear responses in the home environment and shelter, even if people cannot categorise the motives/diagnosis of fear.

Mornement and co-authors [14] argued that general measures of anxiousness and fear measured in the Behaviour assessment for rehoming K9's (B.A.R.K) protocol significantly predicted "Fearful/inappropriate toileting" behaviours post adoption. These results outline the stable predictiveness of fear consistent over a shelter to a post-adoption environment and therefore suggests the stability of fear over longitudinal periods. Foyer and co-authors [38] further reflected this in a study looking at behaviour in the first year of life and in a later temperament test in dogs. Results from the study outlined that dogs scoring high in categories of stranger-directed fear, non-social fear, and dog-directed fear showed a significantly lower rate of success 3 months later in the temperament test due to fear [38]. Therefore, it is of no surprise to observe consistency in the fear response seen in this study.

In relation to the friendliness displayed in the home environment and behaviour assessment, it is no surprise that it reflects previous findings [14]. Mornement and co-authors [14] found that post adoption, dogs greeting visitors in a friendly manner could be predicted by friendliness scores in B.A.R.K. However, it did not appear to be a reliable predictor of problem behaviours, such as overall aggression or destructive behaviour in shelters.

Furthermore, the predictability of behavioural problems outlined in the results using the owner information and the behaviour assessment could be due to the timing of the assessment. The assessment was conducted upon arrival, located in a room which was at a considerable distance from the main shelter. The stress of the shelter may cause the normal behavioural repertoire to change in the dog for the purpose of finding the best coping mechanism to deal with acute stress due to changes in the environment. Therefore, the timing of the assessment (currently at a minimum of 3 days after surrender) may cause the predictability of behaviour post adoption to be poorer due to the changes that stress can cause in normal behaviour. If we take human psychology as an example, humans that go into a novel environment which they have never been in before suffer an acute stress response. Humans, like all animals, need to adapt to a new environment; they can find positive and negative coping mechanisms to help with this which is then reflected in their behaviour [39]. If positive coping mechanisms are not found, then negative coping mechanisms are used, causing problem behaviours and sometimes addiction. Dogs that have never been in the novel environment before, such as the shelter, respond with an acute stress response due to social isolation from previous family, daily routine changes, disturbed feeding, walking, socialising, lack of handling and attachment figures, and sensory overstimulation. The dog must adjust to the new environment and if unable to cope effectively, behavioural problems start to occur. Once adopted, however, dogs then need to adjust back to home behaviour, which can be easy for most dogs but other dogs with behavioural problems may find this difficult. This is consistent with the findings of Mornement and co-authors [14] who indicated a high number of new adopters reporting signs of growling, snapping, and attempting to bite a person.

Not all instances of behaviour seen in the behavioural assessment-reflected responses to the C-BARQ questionnaire, including certain categories of aggression (dog-directed, stranger-directed), separation-related behaviours and possessive behaviours. Only one category of the C-BARQ, owner-directed aggression, showed consistency with the behaviour assessment stranger and toddler doll tests.

One might expect that stranger-directed aggression in these tests would be reported in the C-BARQ but this was not the case. A study by Dalla Villa et al. [25] outlined the use of the Socially acceptable behaviour (SAB) protocol for identifying categories of aggression. The results indicate that only categories of C-BARQ predictive of the SABS were associated with owner-reported aggression towards familiar people and familiar dogs, however, these were not directly measured by any of the SAB subtests. The identification of the category of aggression is difficult as there are numerous such categories [40] and aggression can be multifactorial. Therefore, this could explain the lack of results in the predictability of aggression towards another stimulus e.g., dog-directed and stranger directed. Without thorough examination of the context of aggression, the environment, and a comprehensive understanding of all factors at play, it is very difficult for assessments to correctly identify, let alone predict, categories of aggression.

Separation-related behaviours are difficult for assessments to identify predictably due to the multifactorial nature of the issue. The issue can be easily misclassified due to other underlying problems like attachment-seeking, general anxiety, fears, or phobias [41]. Furthermore, differential diagnosis should always be taken into account before outlining that the individual has separation anxiety. Storengen and co-authors' [42] study of 215 dogs diagnosed with separation anxiety reported that only 18.5% of animals actually had only separation anxiety with no other behavioural problems, whereas 82.8% of the animals had other underlying behavioural problems in addition to separation anxiety, with the most common comorbidity being related to noise sensitivity (43.7%) [42].

Possessive behaviour has been reported in the literature to have a low predictability [13,14,31]. This may be due to the manifestation of the problem being environmentally based [13,31]. Possessive aggression is associated with a need to protect a resource from surrounding threats, however, once a threat is no longer present, the behaviour ceases, therefore it is not often seen in post-adoption environments. The study by Marder and co-authors [13] found that a little over half of the dogs with possessive behaviour in the shelter displayed these issues post adoption, whereas 22% of dogs identified in a shelter with no signs of possessive behaviours exhibited the behaviour post adoption. Furthermore, a study by Mohan-Gibbons [31] into the removal of the test, identified that there was a low risk of injury to handlers, volunteers, staff or adopters and no significant difference in the rate of returns. However, even though it was a low relative risk of occurrence in the home it is predictive, just not perfectly predictive. Possession aggression, however, can be stimulated by environmental or competition in the environment, therefore, if in a stable environment, such behaviours will decrease or cease. Therefore, in the current study, this could explain the low occurrence of possessive aggression, especially in the home environment.

Numerous possibilities exist that consider discrepancies between the behavioural assessment results and owner reports. A possibility is that the current standardised behaviour assessment may be adequate at identifying overall behaviours, however, unable to correctly identify certain behavioural problems. However, behavioural problems, such as dog-directed aggression or separation-related behaviours, may be inaccurately identified due to the misinterpretation of the behaviour by the owner in the home. For example, dogs that are reactive to other dogs at a distance could be misclassified as dog-aggressive or offensive aggressive, when what is being displayed is built-up frustration and hyperactivity towards other dogs. A study that assessed the behaviour of privately owned dogs using the Dutch socially acceptable behaviour test, found that a large portion of aggressive dogs remain undetected and the test was unsuitable for assessing types of aggression apart from fear [23]. The current results agree with this, outlining the high degree of detectability of fear.

There are limitations to this study. One limitation is that all dogs in this study had been in a home environment for over 6 months, and therefore, had an attachment figure. Attachment figures have previously been seen to have a significant impact on inhibitory control, problem-solving tasks and social interactions in comparison to dogs that were in shelters with no attachment figure [43–45]. Another limitation includes that the study population may not be representative of dogs that end up in shelters.

The results from this novel study suggest the benefit of an upon surrender assessment to increase the understanding of behaviour from the previous home environment. Early recognition of behavioural problems that include fear, anxiousness, arousal, and aggression can help dogs cope in the environment and allows behaviour modification to be implemented before the stressors of the shelters begin to have an effect [9].

#### **5. Conclusions**

This study suggested that the standardised behaviour assessment protocol used at an Australian shelter is a useful tool to reflect home behaviour when conducted upon entry to the shelter as mimicked in this study methodology, with friendliness, fearfulness, anxiousness, high arousal and certain categories of aggression measured by the C-BARQ being reflected in the assessment. The identification of behaviours of dogs upon entry can help to create a more comprehensive understanding of the dog's behaviours in the home environment and further identify any behavioural issues/monitored throughout the stay in the shelter plus allow behaviour modification to start upon entry. Information can give a base line for the dogs before entry, thus allowing the longitudinal monitoring of behaviours and behavioural issues. Investigations into longitudinal monitoring from surrender to adoption, and the relationship of individual behavioural change over time, needs to be conducted.

**Author Contributions:** L.C., M.B.A.P., P.B., G.P., and C.C.J.P. conceived the project. L.C. drafted the paper and all authors had input into modifying it into the present format. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** The authors acknowledge the assistance of the RSPCA. We would like to thank James Serpell for his permission to use the C-BARQ for this study.

**Conflicts of Interest:** Mandy Paterson declares that she works for the RSCPA Queensland. Liam Clay declares that RSPCA Queensland funded his studentship. Apart from this, no other author has any conflict.

#### **Appendix A. RSPCA Standardised Behavioural Assessment**

#### *Appendix A.1. Test 1: Exploration of Room*

#### Appendix A.1.1. Exploring the Room

The assessor entered the room, dropped the lead attached to the dog, and sat in the centre on a chair. Then, the observer started a timer and waited for 1 min without any interaction with the dog by either person.

#### Appendix A.1.2. Sociability to Assessor

At the end of exploring the room, the assessor called the dog to them in a friendly voice, remaining in the chair with no other body movement. If there was no response, a second attempt was made, and if still no response the assessor clapped their hands on their lap and said 'come here' in the direction of the dog, trying at least three times to call the dog to them. When the dog came (at the first, second, or third call), the assessor picked up the leash and then stroked the dog from the base of the neck to the tail three times. If the dog did not respond to the first, second, or third call, the assessor approached the dog, picked up the leash, and gave the dog three strokes from the base of the neck to the tail. Following each stroke, the observer and assessor counted 10 s, with the behaviours exhibited noted.

#### *Appendix A.2. Test 2: Tolerance to Handling*

There were three components to the test, namely touch sensitivity to collar, stroke, and feet. The assessor dropped the leash and held the dog's collar. After 3 s, the handler stroked the dog from head to tail. With the dog standing, the other assessor (in the standing position, or crouching if a small breed of dog) picked up the dog's rear inside foot, then the front inside foot, then reached over its back to pick up its rear outside foot, and finally the front outside foot. Each foot was held for 2 s. After picking up all four paws in this manner, the assessor stood for 10 s with no dog interaction and finally removed the dog's leash.

#### *Appendix A.3. Test 3: Startle Response*

There were two components: startle response and recovery to stimulus. At the end of Test 2, the assessor created a loud sound using a book on a bench or a desk (startle response). The assessors recorded recovery.

#### *Appendix A.4. Test 4: Toy Interactions*

Three toys were used in this testing procedure: tennis ball, squeaky toy, and tugging rope. A tennis ball was shown to the dog and gently thrown across the room, and the assessor verbally engaged the dog in play. If the dog picked up the ball, the assessor waited to see if it returned to the assessor without encouragement. If it did not, the assessor encouraged the dog to bring the ball back by calling his/her name and saying "come". If the dog still did not return, the assessor went to the dog.

In both situations, the assessor waited 10 s to see if the dog dropped the ball. If it did not, they asked the dog to "drop it". If the dog did not respond, then a second command was given, "give", and if necessary, a third attempt, "out", was tried. If the dog did not respond to these commands, the

assessor approached the dog carefully and removed the ball from the dog's mouth. These steps were repeated for a second throw and after completion, the assessor waited 10 s with no interaction before moving on to the next toy, the squeaky toy, and after that, the tugging rope. The same sequence was used for each toy. After completing all three toys, the assessor moved on to the next test.

#### *Appendix A.5. Test 5: Response to Unusual*/*Unpredictable Stimulus*

The assessor gently moved the dog to the opposite end of the room and left it standing against the wall. Then, they gently moved one hand over its head, down toward the back to gently tap the rump area, and then ran across the room, laughing and waving arms, followed by suddenly stopping, folding their arms, and ignoring the dog. The tap, run, and freeze series was repeated a second time. The assessor waited for 10 s after the run and freeze, ignoring the dog, before moving onto the next test. The dog was then placed back on the leash.

#### *Appendix A.6. Test 6: Resource Guarding*

There were four components to the test: wet food, dry kibble/biscuits, pig's ear and bone. The assessor tethered the dog to the wall for safety reasons, and proceeded to show the dog wet canned food, smeared in a bowl. The bowl was then placed near the dog at the end of the leash perimeter, allowing the dog to begin eating for 2 s. The assessor then proceeded with a plastic hand, walking to the side of the dog while it was eating. Using the fake hand, the assessor patted the dog on the head, continuing to stroke down its back and body twice. The fake hand was then placed 5 cm in front of the bowl and moved around in a semi-circle. The hand was then placed on the inside edge of the bowl and moved around the edge of the bowl next to the dog's face, without touching it. Finally, the bowl was pulled away from the dog using the fake hand. The bowl was then returned to the dog, which was observed for 10 s.

The assessor then gave the dog a pig's ear or bone, depending on the dog's food interest, and it was allowed to chew it for 30 s. The steps above with wet food were repeated; then, the assessor attempted to retrieve the food, asking the dog to "drop it", "leave it", or "give" before attempting to retrieve it by offering a new food that is novel.

#### *Appendix A.7. Test 7: Stranger Interaction*

There were three components to the test: the entry, approach and exit of a stranger. The assessor placed the dog on a leash as the observer exited the room and returned dressed in a reflective vest, large brimmed hat and using a walking stick. The observer entered the room, and bent down to extend an open flat hand as if to pat the dog on the head. The observer then talked to the dog normally and stopped for 3 s, allowing the dog to approach. If the dog approached, the observer patted the dog on the top of its head for 3 s. If the dog did not approach, it was observed for 10 s, with an emphasis on any interaction between the assessor and/or the observer.

#### *Appendix A.8. Test 8: Fake Toddler Interaction*

There were two components of the test: the approach of the toddler doll and the exit/removal of the toddler doll. The assessor stood and held the dog's leash while the observer exited the area and returned carrying a toddler doll simulating a small child. Once the toddler was within the leash perimeter from the dog, the observer placed the doll on the floor facing the dog, with the doll's arm extended toward the dog. The assessor allowed the dog to approach if it desired. If the dog did not approach the observer, it was observed for 20 s. After this, the assessor picked up the toddler doll and walked back out of the room. The assessor allowed the dog to follow to the door or move away from stimulus.

#### *Appendix A.9. Test 9: Fake Cat*

The assessor stood and held the dog's leash while the observer exited the area and returned carrying a fake cat as if it were a "real" cat. Once the fake cake was within the leash perimeter from the dog, the observer placed the fake cate on the floor facing the dog. The assessor allowed the dog to approach if he/she wanted to. However, if the dog did not approach the observer, the dog was observed for 20 s with the fake cat present.

#### *Appendix A.10. Test 10: Time Alone*

The assessor and observer removed the leash from the dog and left the room for 2 min, with a video camera in the front of the room monitoring behaviour and vocalisations. Then, the assessor and observer re-entered through the same door.

#### *Appendix A.11. Test 11: Behaviour with Another Dog*

There were three components to the test: walking parellel, circling activity, and nose-to-nose interaction. This test was conducted in a yard (10−20 m), allowing adequate space between the test dog and another dog. Each dog had an assessor, who interacted with their dog by giving treats and ignoring the other assessor and dog. The assessor had a short, 1 m leash, so that the dog walked close to the assessor. At the start, both assessors walked parallel to each other, 5 m apart, with the dogs on the outside. If one or both dogs were reactive and pulled toward each other, the distance between the assessors was increased. If both dogs were relaxed and focused on their assessor, the assessors moved the dogs to an exercise circle. If the dogs did not breach a minimum distance of 5 m between them, they were introduced on opposite sides of a fence. Then followed a circling activity, which required one assessor to stand still with their dog on no more than 1.5 m of leash while the other assessor and their dog completed a circle around the assessor. The assessors then swapped places and repeated the circling activity. If no adverse behaviours were displayed, the assessor in the middle of the circle remained at that location, ensuring that the only tension on the leash was from the dog. The other assessor identified the leash threshold of the dog in the centre and moved close enough to allow the dogs to be nose to nose, also ensuring that the only tension on their leads was caused by the dog pulling, not them pulling against the dog. Once the leads became loose, and the dogs stopped pulling against the assessor, the assessors took a step closer to each other, allowing the dogs to interact if they chose. Leashes remained loose. If there were signs of adverse reactions or aggression, the dogs were separated by increasing the threshold.


**Appendix**

 **B**


**Table A1.** *Cont.*




**Table A2.** Number (and %) of respondents (n:107) classifying their dogs in each of five levels on a scale of increasing intensity of behaviour exhibited at home, using the C-BARQ Categories.


anxiety—4stranger,dogaggression,separatelystranger-,dogdog-directed; separation-related problems, from 0 never—4 always; attachment/attention-seeking, from 0 never—4 always; touch sensitivity, from 0 never—4 always; excitability, from 0 calm to 4 extremely excitable; chasing, energy, and trainability, from 0 never—4 always.

#### **Appendix D**


**Table A3.** Percentage of coded durations and frequencies of the five behavioural categories (friendliness, fear, anxiety, arousal and aggression) during each subtest in the standardised behaviour assessment.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Do Canine Behavioural Assessments and Characteristics Predict the Human-Dog Interaction When Walking on a Leash in a Shelter Setting?**

**Hao-Yu Shih 1,\*, Mandy B. A. Paterson 1,2, Fillipe Georgiou <sup>3</sup> and Clive J. C. Phillips <sup>4</sup>**


Received: 18 November 2020; Accepted: 24 December 2020; Published: 25 December 2020

**Simple Summary:** We explored 370 human-dog interactions in an animal shelter when volunteers walked shelter dogs on a leash, considering the effects of canine demographics and the results of the shelter's canine behavioural assessments. Results showed that dogs that were more relaxed during the shelter assessment (i.e., when socialising with humans or being left alone in a new environment) were less reactive on the leash, with lower tension and pulling frequency. Moreover, socialised and relaxed dogs displayed more positive body language, such as tail in a high position, gazing at the handler, and exploring the environment. When walking with these dogs, volunteers utilised fewer verbal guidance cues and body language during the walk. In addition to the canine behaviour assessment, there were correlations between canine demographics (i.e., age, skull shape, body size, and previous ownership history) and the behavioural interaction and humans' perception. Finally, volunteers perceived the walk as less satisfactory when they needed to pull the leash harder during the walk. This research suggests that the RSPCA behavioural assessment may be useful in predicting the behaviour of shelter dogs when walked by volunteers.

**Abstract:** Inappropriate leash reactivity is one of the most common problems in shelter dogs, which negatively affects the health of dogs and reduces their adoptability. We explored 370 human-dog interactions, involving 74 volunteers and 111 dogs, in an animal shelter when volunteers walked shelter dogs on a leash, considering the effects of canine demographics and the results of the shelter's canine behavioural assessments. The interaction was video recorded and coded using ethograms, and a leash tension meter was used to measure the pull strength of dogs and handlers. Results showed that dogs that were more relaxed during the shelter assessment (i.e., when socialising with humans or being left alone in a new environment) were less reactive on the leash, with lower tension and pulling frequency. Moreover, socialised and relaxed dogs displayed more positive body language, such as tail in a high position, gazing at the handler, and exploring the environment. When walking with these dogs, volunteers utilised fewer verbal cues and body language during the walk. In addition to the canine behaviour assessment, there were correlations between canine demographics and the behavioural interaction and humans' perception. Finally, volunteers perceived the walk as less satisfactory when they needed to pull the leash harder during the walk. This research suggests that the RSPCA behavioural assessment may be useful in predicting the behaviour of shelter dogs when walked by volunteers.

**Keywords:** on-leash walk; canine behavioural assessment; leash tension; behaviour; verbal cue; body gesture; human-dog interaction; shelter

#### **1. Introduction**

An animal shelter is a challenging environment for dogs, causing both short-term and long-term stress, with acutely elevated cortisol levels within five days of dogs being transferred into a shelter [1]. Moreover, prolonged confinement leads to more problematic behaviours (e.g., decreased activity or excessive auto-grooming [2]), which compromises the animal's welfare and negatively influences their adoptability [3]. Therefore, positive human-dog interactions that help alleviate the stress of dogs are important in an animal shelter [4]. There are a variety of human-dog interactions that occur routinely in shelters, including petting [4], training [5] and on-leash dog walking [6]. Despite the extended literature on human-dog interactions, limited research has explored the role of human-dog interactions in the shelter context.

Physical characteristics of dogs influence the human-dog interaction. Compared to adult dogs, puppies are more likely to evoke our nurturing instinct and more quickly and easily form a stronger attachment with humans [7,8]. Smaller dogs are generally perceived as less obedient, more anxious, fearful, excitable and aggressive [9], while larger dogs are thought to be more cooperative and playful [10]. Dogs with wider heads are more likely to display self-grooming but less likely to chase [11]. Nevertheless, small-sized dogs are generally preferred by the general public [3], and there is a trend of people favouring breeds with shorter and wider heads [12–14].

To evaluate the behaviour of incoming dogs and their adoptability [15,16], many animal shelters have implemented canine behavioural assessments (e.g., SAFER® [16] and Assess-A-Pet™ [17]). In the Royal Society for the Prevention of Cruelty to Animals Queensland (RSPCA Qld), a similar test is conducted on adult dogs around day 5 after arrival to the shelter [18,19]. The canine behavioural assessments measure how dogs react to different stimuli in different situations [15,20]. Controversies have arisen about whether the assessments conducted in animal shelters do indeed predict future household behaviours after adoption [21–23]. One argument is that certain behaviours, such as aggression [24] and separation anxiety [25], are context- and stimulus-specific. Since the behavioural assessment is usually done by an examiner who may be perceived as unfamiliar and intimidating to the dog, the result may not be transferable to a situation where an owner interacts with the dog [23,24]. Although such arguments have been raised, behavioural assessments still reliably describe animals' behaviours within the shelter [19,26]. For example, fear, anxiety and arousal-related behaviours found in the assessment were also observed when dogs were in the shelter kennels [19].

Behaviours of dogs are also related to how we perceive and interact with them. For instance, owners of aggressive English cocker spaniels were reported to be shy, undisciplined, less emotionally stable and tense [27]. Owners are also less likely to engage in shared activities and training with dogs if their dogs are disobedient, aggressive or bark excessively [28]. When it comes to shelters, a less timid dog prefers to play with humans rather than engaging in independent play [29], and dogs spending more time in front of the kennel, laying proximal to the adopter and reacting to humans' play behaviours attract more preference from potential adopters [30,31]. Problematic behaviours, such as vocalising, house soiling and aggression, negatively affect the owner–dog attachment and significantly increase the risk of unsuccessful dog ownerships, such as owners surrendering dogs to shelters or dogs being returned after adoption [32–34].

In many dog shelters, on-leash walking is an important part of the human-dog interactions that improves the health of shelter dogs, reduces their in-kennel stress level and facilitates their socialisation [35, 36]. However, leash reactivity is one of the most common problems in shelter dogs [37]. Dogs often lunge forwards during on-leash walks, which may damage the soft tissue around the neck and trachea [38] and contribute to increased intraocular pressure [39]. In addition, leash reactivity is reported to be a common problem after adoption [37] and is related to the failure of dog ownerships [40].

An equine rein tension meter measures the force exerted on the rein, which enables researchers and trainers to monitor how a rider communicates with the horse [41,42]. A similar concept is adopted in this study using a custom-made leash tension meter to capture the leash tension when walking a dog on a leash. Moreover, this canine leash tension meter includes an accelerometer in the device which differentiates between human and dog pulling during the walk [43].

Apart from the leash tension meter, behavioural observation using video recording was also used [30, 44]. For dogs, the position of the tail, facial expression and ongoing behaviours were recorded, which are directly related to the animal's welfare [45,46]. In humans, body gestures and verbal cues were measured as they influence the response of the dogs [47,48]. It was hypothesised that dog tolerance to physical contact would be related to higher leash tension, while dogs that were more relaxed in an unfamiliar environment would have lower leash tension [19]. We also hypothesised that dogs would show more friendly signs and have a lower leash tension if the dogs were more accustomed to human interaction. Finally, we hypothesised that handlers would use fewer verbal and gestural cues when dogs were less easily aroused as scored in the behavioural assessment [19] and were older, presumably because they were generally calmer and more obedient [28]. Additionally, handler satisfaction would be negatively associated with leash tension [28].

This article describes the relationship between the results of canine behavioural assessment, canine demographics and human-dog interactions when the dog is walked on a leash, with the focus mainly on canine behavioural assessment. Therefore, results regarding the behavioural assessment will be presented in the main body of the manuscript, followed by relevant demographics. More details on demographic characteristics are presented in the appendices.

This article is a part of a larger research project that explores the behavioural interaction between shelter dogs and volunteers during walks. In this paper, emphasis will be placed on how canine demographics and behavioural assessment results influenced the behavioural interaction while shelter dogs were being walked on leash by volunteers. The effect of human gender has been reported [49] and the effects of body size, body weight, age and the behavioural level of dogs on the leash tension, and the relationship between canine sex and behaviour and leash tension were also reported [43,49]. The influence of other human demographics and personality will be reported in a future publication.

#### **2. Materials and Methods**

This study was approved by the Human Research Ethics and Animal Ethics Committees (approval numbers: 2018001570 and SVS/400/18, respectively) of the University of Queensland.

#### *2.1. Study Site*

The research was conducted at the Royal Society for the Prevention of Cruelty to Animals, Queensland (RSPCA, Qld) shelter. The housing schedule and shelter environment have been fully described in a previous paper [49]. This study focused on the human-dog interaction during the walks as described in Shih et al. [49].

#### *2.2. Subjects*

This study investigated 370 walks (370 unique dog-walker pairs), involving 111 shelter dogs and 74 volunteer walkers, with each walker walking 5 different dogs. The classification of the dog behavioural level was based on their performance during the daily walk. Level 1 dogs walked on a loose leash most of the time. Level 2 dogs pulled the leash occasionally. Level 3 dogs tended to pull the leash fiercely. Level 3+ dogs had severe behavioural issues (e.g., aggressiveness or fearfulness) and they might or might not pull the leash harder than level 3 dogs. More details about the behavioural level are described in Shih et al. [49].

#### *2.3. RSPCA Canine Behavioural Assessment*

Each dog entering RSPCA Qld is behaviourally assessed within 5 days of entry using a standardised assessment. The assessment used is fully described in Clay et al. [19] and comprises the following subtests: Socialisation, Tolerance, Toys, Run and Freeze, Resource Guarding, Toddler Doll, Time Alone and Dog to Dog. In the Socialisation subtest, a dog was allowed to freely explore the room and interact with the handler. Its behaviour and body tension were evaluated. In the Tolerance subtest, the dog's behaviour, body position and tension were assessed when it was stroked and touched on the feet by the handler. In the Toys subtest, the dog was allowed to play with the toys by itself and with the handler; its behaviour during the interaction was recorded. In the Run and Freeze subtest, the handler ran around in the room and suddenly froze in the corner. The response of the dog was recorded. In the Resource Guarding subtest, the dog was given wet food, dry food and a real bone, consecutively. The examiner assessed its behaviour, body position and tension when the handler was trying to remove the food with a fake hand. In the Toddler Doll subtest, the dog was presented with a fake toddler which mimicked the physical appearance and behaviour of a real child, and its reaction to the toddler was assessed. In the Time Alone subtest, the dog was left alone in the room without any room light (but some natural light through a window) while being videoed. Its behaviour was later evaluated by reviewing the digital recording. Finally, in the Dog to Dog test, the assessed dog was introduced to another dog, and the behaviour of the assessed dog was recorded. This research focused on the human-dog dyad and thus the Dog to Dog subtest was not discussed here.

#### *2.4. Canine Leash Tension Meter*

The customised canine leash tension (sampling rate: 10 Hz; measuring range: 0–100 kgforce; resolution: 100 g-force) meter was commissioned for this project (RobacScience Australia). The device measures the force exerted on the leash and detects the direction of the pulling (handler versus dog). [43].

#### *2.5. Measurement of the Dog's Body*

Body size was determined using the body height (cm), body length (cm), body weight (kg) and body condition score (BCS). Body height and length were measured from the ground to the dorsal scapular rim and from the cranial aspect of the shoulder joint to the caudal aspect of the sciatic tuberosity, respectively, using a tape measure [50]. Body weight was obtained from the RSPCA database. The BCS was determined according to the 9-point scale BCS system [51]. Cephalic index (CI) was calculated as the ratio of skull width to skull length. The skull length was measured from the occipital crest to the tip of the nose, and the width was the widest part of the dog's head (distance between two zygomatic arches) [11].

#### *2.6. Study Design*

The study design is explained in Shih et al. [49]. Participants were instructed to walk the dog with one hand and only to touch the leash when the dog got tangled. This represented the most common practice in the real world, even though the official advice is to use two hands [52]. After each walk, participants completed a questionnaire about their perspective of the walk (Table 1).

**Table 1.** Exit questionnaire used for volunteer walkers (reprinted from Shih et al. [49] for convenience).


Each description was rated on a 5-point Likert scale from 1 (strongly disagree) to 5 (strongly agree). Reverse scores were used for negative wording for the calculation of mean scores. Human satisfaction factor (Factor H) utilised responses to questions 2, 3, 4, 5, 6, 10, 11. A higher score of factor H represented a higher satisfaction of the walk. Walker's perception of dog factor (Factor D) utilised responses to questions 1, 7, 8, 9, 12, 13. A higher score of factor D indicated that the dog was considered more supportive and better behaved. Details about the statistical justification of the questionnaire are described in [49].

#### *2.7. Data Analysis*

#### 2.7.1. Video Recordings of Dog and Human Behaviours

To record the interaction during the walk, a camera (GoPro Hero 7 Silver, GoPro®, San Mateo, CA, USA) was mounted on the walker's head, and, at the same time, the researcher recorded the walk from 10 m behind with an iPhone 7 (Apple Inc., Cupertino, CA, USA). Videos were coded in their entirety by the researcher, who is a veterinarian and a certified dog trainer, with Boris© behaviour observation software [53], using the same method described in Shih et al. [49]. To blind the coder, video coding was completed prior to any analysis of human and canine characteristics. Ten randomly selected videos were recoded to check intra-observer reliability (Cohen's Kappa = 0.76). Canine behaviours, human verbal cues and human body language were coded using ethograms developed based on previous research [4,45,54–57], as previously described [49] and modified during practice sessions (tables 2 to 4). These tables are reproduced from Shih et al. [49] to help with the understanding of this paper.


#### **Table 2.** Ethogram of canine behaviour [49].

Point event: the number of times the event was observed. State event: the duration of the observed event.

#### **Table 3.** Ethogram of human verbal cues [49].


Point event: the number of times the event was observed. State event: the duration of the observed event.

#### **Table 4.** Ethogram of human body language [49].


Point event: the number of times the event was observed.

#### 2.7.2. Leash Tension Analysis

Leash tension and pulling frequency were analysed with MATLAB® (MATLAB® and Statistics Toolbox Release 2018b, The MathWorks, Inc., Natick, MA, USA) using the same approach as described in Shih et al. [49].

Every pulling episode can be divided into three phases. Phase one is the "initiation phase", when either the dog or the handler initiates the pulling, which is marked by an increase in leash tension and acceleration of the device towards the initiator of the pull. Phase two is the "contest phase", when the other party counteracts the pulling, which is marked by an acceleration in the opposite direction to the initiator and a sharp increase in the leash tension. Finally, phase three is the "losing or winning phase", when either the party that initiated the pulling wins the contest or the opposite party wins (the party that initiated the pulling loses) the contest. This phase is characterised by a decrease in the leash tension; there should be acceleration towards the "winner". The leash tension then either returns to the baseline or a new pulling episode characterised by a change in the gradient of the leash tension occurs.

Net maximal tension (NTmax), maximal tension by dog (DTmax) and handler (HTmax) represent the maximal tension throughout the walk, recorded for the dog and handler, respectively. Net mean tension (NTmean), mean tension by dog (DTmean) and mean tension by handler (HTmean) represent the mean tension throughout the walk, recorded for the dog and handler, respectively. Dog pulling frequency (DPF) and handler pulling frequency (HPF) were the frequency of pulling initiated by the dog and handler, respectively.

#### *2.8. Statistical Analysis*

Statistical analysis was conducted using RStudio Version 1.2.1335 [58] with packages leaps [59], MASS [60], car [61], carData [62], Matrix [63], polycor [64], plyr [65], psych [66], ggpubr [67] and nlme [68].

The canine behavioural assessment was transferred into the scoring system based on the effect of each behavioural presentation on the human-dog interaction and safety (Table S1). A higher score in the subtest indicates that dogs displayed more behaviours that would favour the human-dog interaction and safety. One dog had two behavioural assessment results because the shelter staff wanted to reconfirm or monitor the result. For that dog, an average score was calculated.

This study used the same statistical analysis methods as described in [49]. Bivariate generalised linear models were used for the analysis of each combination of dependent (leash tension, pulling frequency, behaviour and the score of the exit questionnaire) and independent (human and dog demographics, human personality, canine behavioural assessment) variables, followed by generalised linear mixed models for multivariate analyses and repetitions of dogs and walkers. According to the results of bivariate generalised linear models, independent variables with *p* values less than 0.2 and those logically expected to affect the dependent variable, regardless of the *p* value, were included in the generalised linear mixed model. Dependent variables were manually transformed for statistical analysis to meet the assumptions of generalised linear mixed models, including the normality of residual and random effects and homogeneity of variance of residual [49,69].

#### **3. Results**

#### *3.1. Demographics*

This study involved 111 shelter dogs, with 58 (52.3%) females and 53 (47.7%) males and all were gonadectomised. The mean age of dogs was 3.74 (±2.45) years (44.82 ± 29.37 months) old. The mean body height and length were 52.04 (±6.29) cm and 56.05 (±5.93) cm, respectively. The mean body weight was 24.87 (±6.65) kg and the mean body condition score was 4.59 (±1.07). Finally, the mean cephalic index was 0.58 (±0.058).

There were 43 (38.74%) stray dogs, 31 (27.93%) surrendered to the RSPCA by owners, 19 (17.12%) returned to the RSPCA by previous adopters and 18 (16.22%) having other or unknown sources.

#### *3.2. Canine Behavioural Assessment, Demographics and Leash Tension/Frequency*

The median score and interquartile range (IQR) of each behavioural assessment subtest was as follows: socialisation (median = 3.00, IQR= 1.00), tolerance (median = 4.00, IQR = 3.00), toy (median = 1.00, IQR = 3.00), run and freeze (median = 3.50, IQR = 1.00), resource guarding (median = −3.00, IQR = 1.00), toddler (median = 4.00, IQR = 1.00) and time alone (median = −1.00, IQR = 1.00).

Dogs that were more socialised when exploring the room and interacting with humans were correlated with lower maximal net leash tension (*p* = 0.026). However, dogs that were more tolerant of the human's physical contact were associated with higher maximal net leash tension (*p* = 0.048). Dogs that were more engaged in playing with toys by themselves or with humans had less pulling frequency (*p* = 0.043). Dogs being more friendly to a model toddler were related to higher maximal net leash tension (*p* = 0.018) and higher pulling frequency created by dogs (*p* = 0.016) and humans (*p* = 0.0003). Dogs exhibiting more reactions (lower score) to time spent alone were correlated with higher mean net leash tension (*p* = 0.018), higher mean (*p* = 0.039) leash tension and pulling frequency (*p* = 0.038) created by dogs and higher maximal (*p* = 0.026) and mean (0.025) leash tension created by humans (Table 5).



Tension and pulling frequency were analysed in log10 transformation. NTmax: maximal net leash tension. NTmean: mean net leash tension. DTmax: maximal leash tension caused by dog. DTmean: mean leash tension caused by dog. HTmax: maximal leash tension caused by handler. HTmean: mean leash tension caused by handler. DPF: dog pulling frequency. HPF: handler pulling frequency. *β*: regression coefficient. SE: standard error of *β*. *p*: *p* value of the model. <sup>1</sup> Pulling frequency = (Number of pulls) / (walking duration). A pull was defined as a bout of force greater than 0.1% of the dog's body weight force.

Compared to dogs classified as strays, those surrendered by their owners had lower net maximal (*p* = 0.0084) and mean (*p* = 0.0008) leash tension, maximal (*p* = 0.022) and mean (*p* = 0.0003) leash tension created by dogs and maximal (*p* = 0.041) and mean (*p* = 0.0029) leash tension created by humans; dogs returned after a failed adoption were associated with lower maximal leash tension (*p* = 0.047) and pulling frequency by humans (*p* = 0.013) (Table S2).

#### *3.3. Canine Behavioural Assessment, Demographics and Canine Behaviour*

Dogs that were more socialised when exploring the room and interacting with humans spent a higher percentage of time tracking (*p* = 0.0085) and keeping their tails in a high position (*p* = 0.049) and gazed at handlers more frequently (*p* = 0.036). Dogs that were calmer when seeing a person running and suddenly freezing less frequently displayed gazing (*p* = 0.038) and lip-licking (0.0013) behaviours, but they spent a greater percentage of time sniffing (*p* = 0.013). For dogs showing less resource guarding potential, gazing (*p* = 0.0006) and lip-licking (*p* = 0.0001) were less commonly observed, but these dogs eliminated more often (*p* = 0.036). Finally, dogs being calmer and friendlier toward the fake toddler gazed at handlers less frequently (*p* = 0.045) (Table 6).


**Table 6.** Generalised linear mixed model of the effect of canine behavioural assessment on canine behaviour. the dog's head was visible in the Gopro video (s), analysed in power of 0.4. Eliminate-mark (no./sec): number of eliminate-marks/total walking time (s), analysed in power of 0.6. Shake (no./sec): numbershakes/total walking time (s), analysed in power of 0.8. Pant (%): painting time (s)/time when the dog's head was visible in the Gopro video (s) × 100%, analysed in power of 0.5. Sniff (%): sniffing(s)/total walking time (s) × 100%, analysed in power of 0.5.β: regression coefficient. SE: standard error of β. *p*: *p* value of the model.

 time

The age of the dog was negatively correlated with the percentage of time dogs spent tracking (*p* = 0.012) and the frequency of shaking behaviour (*p* = 0.0021). The cephalic index of the dog was positively related to the percentage of time dogs spent tracking (*p* = 0.0025) but negatively associated with the percentage of time dogs spent panting (*p* = 0.0009). Compared to level 3 dogs, level 2 dogs wagged their tails more often (*p* = 0.038); level 3+ dogs displayed higher percentages of tracking (*p* = 0.0015) and sniffing behaviours (*p* = 0.039) but a lower percentage of panting behaviours (*p* = 0.027) (Table S3).

#### *3.4. Canine Behavioural Assessment, Demographics and Human Behaviour*

Less frequent communication (*p* = 0.029), commands (*p* = 0.0079) and giving food treats (*p* = 0.034) were observed when handlers walked dogs that were calmer when seeing a person running and suddenly freezing. When walking dogs with less resource guarding potential, handlers were less likely to communicate (*p* = 0.0025) with them and displayed less body language (*p* = 0.0017) and were less likely to give food treats (*p* = 0.0002) and initiate physical contact (*p* = 0.0008) (tables 7 and 8).

**Table 7.** Generalised linear mixed model of the effect of canine behavioural assessment on human verbal cue.


<sup>1</sup> Analysed after transformation to the power of 0.5. <sup>2</sup> Analysed after transformation to the power of 0.4. β: regression coefficient. SE: standard error of β. *p*: *p* value of the model.

The age of the dog was negatively correlated with the frequency of human communication (*p* = 0.043). Compared to level 3 dogs, handlers used more negative verbal cues when walking level 2 dogs (*p* = 0.042). Compared to stray dogs, handlers were more likely to talk to surrendered dogs using a high-pitched voice (*p* = 0.028). The size of the dog was positively related to the frequency of communication (*p* = 0.011), total body language (*p* = 0.02) and physical contact initiated by the handler (*p* = 0.0058) (Tables S4 and S5).

#### *3.5. Canine Behavioural Assessment, Demographics and Walking Experience*

No significant relationship was observed between the canine behavioural assessment and the walking experience (*p* > 0.05) (Table 9). The age of the dog was positively associated with the score of factor H (*p* = 0.041) and factor D (*p* = 0.0096). Compared to level 3 dogs, level 3+ dogs were correlated with lower scores of factor H (*p* = 0.016) and factor D (*p* = 0.026). Mean leash tension created by humans was negatively related to the factor H score (*p* = 0.0062) (Table S6).


**Table 8.** Generalised linear mixed model of the effect of canine behavioural assessment on human body language.

<sup>1</sup> Analysed after transformation to the power of 0.5. <sup>2</sup> Analysed after transformation to the power of 0.4. β: regression coefficient. SE: standard error of β. *p*: *p* value of the model.

**Table 9.** Generalised linear mixed model of the effect of canine behavioural assessment on volunteers' walking experience. Factor H represented human satisfaction factor and factor D represented walker's perception of dog factor.


<sup>1</sup> Analysed after transformation to the power of 10. β: regression coefficient. SE: standard error of β.

*p*: *p* value of the model.

#### **4. Discussion**

#### *4.1. Canine Behavioural Assessment, Demographics and Leash Tension/Frequency*

Dogs being more engaged in playing with toys by themselves or with humans and releasing the toy on command or when traded with treats were less likely to pull on the leash. This finding may indicate that these dogs were more relaxed when interacting with humans and they were more likely to understand and obey human signals when walking on the leash. Dogs that were more friendly and relaxed when facing the fake toddler had higher maximal net leash tension. In addition, these dogs pulled on the leash more frequently and so did the handlers. Since dogs did not encounter any other human being, except the handler, during the walk, it was unlikely that these dogs were attracted by a person or a toddler. Moreover, a previous study on the canine behavioural assessment using a fake toddler or fake cat has shown that instead of testing the dog's response to a real toddler or a cat, the test is likely testing its response to a foreign object [70–72]. Therefore, in our result, it was more likely that these dogs pulled more frequently on the leash because they were more interested in exploring objects around them and handlers were simply responding to the higher pulling frequency by pulling on the leash more frequently too.

Finally, dogs that were more anxious and reactive when left alone in an unfamiliar environment pulled harder and more frequently, which supports our hypothesis about the negative correlation between the relaxation of dogs in a new environment and leash tension. It might also suggest that these dogs were anxious and attempted to escape from the environment. A better image quality is needed to allow a better examination of the dogs and therefore permit a better understanding of the underlying emotion or motivation of the dog. For handlers, during such interactions, handlers also pulled harder on the leash. This subtest was intended to identify the dog's potential to develop separation-related issues. Nevertheless, it has been shown that such tests fail to reliably predict the future development of separation-related behaviours after adoption [23]. It might be, however, that, in the subtest, dogs were demonstrating behaviours responding to the foreign environment, which explains why dogs showing more anxious and stress-related signs in the subtest were less relaxed during the walk, even though handlers were around them.

Compared to dogs found as strays, those relinquished by owners had lower leash tension and those returned by adopters had lower pulling frequency created by handlers. This result satisfies the hypothesis that owner relinquished and returned dogs are more familiar with human interaction and thus are more manageable on the leash. However, this result differs from a previous study showing that the prevalence of unruly behaviour (e.g., jumping up, pulling the lead, poor command responding, lack of concentration) between stray and relinquished dogs was not significantly different [73].

Dogs that were more socialised when exploring the room and interacting with humans had lower maximal net leash tension, potentially because these dogs were generally more relaxed. In line with our hypothesis, dogs that were more tolerant of human physical contact were associated with higher maximal net leash tension, probably because they tended to ignore the pressure exerted on their bodies when pulling on the leash or were simply less concerned about the human and just eager to go for their walk. However, these results should be interpreted with caution because only net tension was measured.

#### *4.2. Canine Behavioural Assessment, Demographics and Canine Behaviour*

Dogs that were more socialised when exploring the room and interacting with humans spent a higher percentage of time tracking and keeping their tails high and they more frequently gazed at the handler. Such results satisfy the hypothesis regarding the positive association between canine socialisation and positive body language of dogs. Dogs that explored the exam room were also more likely to explore in other environments, supporting the finding that the RSPCA Qld socialisation test predicts their friendless

and sociability in a new environment after adoption [23]. The high tail position and frequent gazing behaviour show that socialised dogs were more confident and more engaged in their interactions with handlers [74,75]. Dogs that were calmer and more relaxed when seeing a person running and freezing less frequently gazed at the handler and displayed lip-licking behaviours, but they spent a higher percentage of time sniffing. Since human-directed gazing and lip-licking can also be interpreted as signs of anxiety or anticipation [76,77], it was possible that these dogs were generally less stressed or aroused when walking on a leash and preferred spending more time exploring the environment through sniffing [76,77]. The sniffing behaviour may also be a displacement signal, indicating conflicting emotions during environmental exploration and mild stress resulting from the novel environment or interacting with the handler [78]. Dogs that were more relaxed in the resource guarding subtest less frequently licked their lips and gazed at the handler but eliminated more frequently. Dogs that were calmer and friendlier toward the fake toddler also gazed at handlers less often. These results might indicate that dogs that were more relaxed when approached during eating and when encountering a foreign object were less defensive and stressed during the walk [76,77]. However, again, results should be interpreted with care as this result might only represent dogs' in-shelter behaviours [19] because shelter assessment of resource guarding has been shown to unreliably predict their behaviours outside of the shelter post-adoption [79].

The age of the dog was negatively correlated with the time spent tracking during the walk, which supports the previous assumption that older dogs accumulate more experiences and thus they naturally become less engaged in their surroundings, showing less interest in exploration and a reduction in excitement [80]. Older dogs also shook their bodies less frequently when on leash. Body shaking is a recognised displacement behaviour linked to stressful situations involving anxiety and excitement, which is believed to be an attempt to relieve the accumulated stress [81]. In addition, older dogs were found to pull less frequently on the leash [43]; therefore, it might be concluded that compared to younger dogs, older dogs were generally calmer and less stressed and excited, showing less interest in the environment during the walk.

Dogs with a wider head (higher cephalic index) spent more time tracking but less time panting, which seems to contradict the fact that brachycephalic breeds are more likely to pant due to brachycephalic obstructive airway syndrome [82]. The result should be interpreted cautiously because most of the dogs in this study were mesocephalic [83]. Therefore, it is more reasonable to conclude that within mesocephalic breeds, wider-headed dogs may spend more time tracking but less time panting when walking on the leash. The cephalic index of dogs is fully explained by neither the breed groups nor the genetic clusters [84]. Consequently, the finding is less likely due to the breed effect but more likely due to the structural differences, which requires further investigation.

Compared to level 3 dogs, level 2 dogs wagged their tails more often, which was more commonly seen when interacting with handlers or waiting for handlers to open the gate to enter the next walking section. This may indicate that level 2 dogs enjoyed being around humans more, which is supported by the lower leash tension created by level 2 dogs [43]. Level 3+ dogs showed more tracking and sniffing but less panting behaviours, potentially because they were actively searching for stimuli.

#### *4.3. Canine Behavioural Assessment, Demographics and Human Behaviour*

Less frequent communication, commands and giving food treats were observed when handlers walked dogs that were calmer upon seeing a person running and freezing. Similarly, handlers were less likely to communicate and displayed less body language, including giving food treats and initiating physical contact when walking dogs that were less defensive and anxious in the resource guarding test. These findings support the hypothesis that handlers would use fewer verbal and gestural cues if dogs were less aroused in the shelter assessment. A possible explanation may be that fewer verbal and physical

guides are needed during the walk, when dogs are calm and relaxed, and also such results support previous findings that behaviours associated with fear, anxiety and arousal in the RSPCA Qld assessment predict dogs' in-shelter behavioural presentations [19].

Our hypothesis that fewer human verbal or physical signals would be observed when walking older dogs was supported. Handlers less frequently used communication and negative verbal cues when interacting with older dogs, showing that older dogs were generally more stable [85], which aligns with the lower pulling frequency observed in older dogs [43]. Handlers were more likely to use negative verbal cues when talking to level 2 dogs compared to level 3 dogs. A possible explanation may be that level 2 dogs were more often involved in interactions with humans and thus were more likely to get excited and needed to be stopped by handlers. Voice pitch was a key factor modulating the behaviour of younger dogs, and humans often communicate with them using a high-pitched voice [86]. In this study, dog age was not found to influence the tendency of humans to use a high-pitched voice for communication. One possible explanation is that most of the dogs in our study were older than 6 months of age. Therefore, these dogs were less likely to be viewed as puppies. Handlers verbally communicated with larger dogs and initiated physical contact more frequently, which may be because handlers could more easily interact and have physical contact with larger dogs due to the shorter physical distance between them.

#### *4.4. Canine Demographics and Walking Experience*

Handlers were less satisfied with the interaction and dogs were perceived as less obedient when dogs were younger [87,88], supporting the hypothesis that higher satisfaction would be found when the walk involved an older dog. The adolescent-phase disobedient behaviour has been reported in younger dogs, which corresponds with the peak age at which dogs are relinquished to shelters [88,89]. Interestingly, the reduced obedience of dogs observed in the adolescent phase is only found with respect to the carer who has developed attachment with the dog [88]. The development of disinhibited attachment has been reported in shelter dogs, which characterises quickly forming bonds to new humans after short interactions [90,91]. Future study is encouraged to explore the relationship between adolescent-phase disobedient behaviour and human-dog attachment in the animal shelter setting.

Compared to level 3 dogs, handlers reported a lower level of satisfaction and rated the dogs as less obedient and supportive when walking level 3+ dogs. However, there was no difference between level 3 and level 2 and 1 dogs, and no significant correlation was observed between the canine behavioural assessment and walking experience. In this study, volunteers were matched with dogs based on training experience and the behaviour of the dog due to safety and welfare concerns, and thus the exit questionnaire might fail to reflect the differences. In addition, the level of each dog was rated by RSPCA Qld staff during the daily walk, which included both walk inside and outside of the research area. To standardise the research, in this study, volunteers were asked to answer the exit questionnaire based on their experiences inside the research area. Differences between level 3 and level 2 and 1 dogs might be more detectable when considering the behavioural presentation of the dog outside of the research area, where more stimuli (e.g., other dogs and humans) are present. Another possibility may be that volunteers were generally more tolerant or preferred not to reveal their true thoughts in the questionnaire despite being told that the survey would be de-identified. In spite of this, the satisfaction score was negatively correlated with the mean leash tension created by the handler, supporting the hypothesis that handlers would be less satisfied with the interaction when they needed to pull the leash harder during the walk.

#### *4.5. Limitations*

In this study, there were difficulties in accurately identifying the underlying emotions and motivation of the dog. Similar to the cortisol level and heart rate variability, our leash tension meter may only be able to differentiate the relaxed and aroused states of the dog but cannot specify whether the arousal is due to excitement, anxiety or fear [81,92]. To precisely interpret the underlying emotion and motivation of the dog, a better video quality that records the face and the entire body of the dog, environment and the context of the interaction is needed. Other limitations for this study are described in [49].

#### **5. Conclusions**

This research explored human-dog interactions with two features. Different from many human-dog interaction studies with participants being owners and dogs being pets [28,56], this study was conducted in a shelter setting, where volunteers generally shared a short-term relationship and a weaker bond with shelter dogs. Additionally, since on-leash walking is an important activity for dogs in a shelter and postadoption, the other feature of this study is that volunteers interacted with dogs when on a leash.

RSPCA canine behavioural assessment may be useful in predicting the behaviour of shelter dogs when walking on-leash with volunteers. Dogs that were more relaxed during the assessment (e.g., when socialising with humans or being left alone in a new environment) were less reactive on the leash, with lower tension and pulling frequency. Additionally, socialised and relaxed dogs displayed more positive body languages, such as tail in a high position, gazing and exploring the environment, and humans showed fewer verbal cues and body languages during the walk with these dogs.

In addition, we found correlations between canine demographics and the behavioural interaction and human perception. Demographics included age, skull shape, body size and previous ownership history. Finally, the tension of the leash was related to human perception, with the walk perceived as less satisfactory when volunteers needed to pull the leash harder during the walk. This study may help to enhance volunteers' experiences when walking shelter dogs on a leash and improve canine welfare in shelters. Matching shelter dogs with potential adopters, considering the demographics of dogs and humans and canine behaviour to achieve a higher rate of successful adoption, may be another possibility [43,49,93–95].

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2076-2615/11/1/26/s1, Table S1: Scoring of canine behavioural assessment. Details of each subtest are described in [19], Table S2: Generalised linear mixed model of the effect of canine demographics on leash tension and pulling frequency, Table S3: Generalised linear mixed model of the effect of canine demographics on canine behaviour, Table S4: Generalised linear mixed model of the effect of canine demographics on human verbal cue, Table S5: Generalised linear mixed model of the effect of canine demographics on human body language, Table S6: Generalised linear mixed model of the effect of canine demographics and leash tension caused by humans on volunteers' walking experience.

**Author Contributions:** Conceptualization, H.-Y.S., C.J.C.P. and M.B.A.P.; methodology, H.-Y.S., C.J.C.P., F.G. and M.B.A.P.; software, F.G. and H.-Y.S.; validation, H.-Y.S., F.G. and C.J.C.P.; formal analysis, H.-Y.S., F.G. and C.J.C.P.; investigation, H.-Y.S.; resources, M.B.A.P. and C.J.C.P.; data curation, H.-Y.S., F.G. and C.J.C.P.; writing—original draft preparation, H.-Y.S. and F.G.; writing—review and editing, H.-Y.S., C.J.C.P. and M.B.A.P.; visualization, H.-Y.S. and C.J.C.P.; supervision, C.J.C.P. and M.B.A.P.; project administration, H.-Y.S., C.J.C.P. and M.B.A.P.; funding acquisition, C.J.C.P. and H.-Y.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Higher Degree Research Fund of the University of Queensland and the RSPCA Fund donated by philanthropists.

**Acknowledgments:** We thank RSPCA, Qld for allowing us to conduct this research at the shelter. We also appreciate all dogs, volunteers and all RSPCA staff for playing important roles in this study. Finally, we acknowledge Solomon Woldeyohannes for the statistical advice.

**Conflicts of Interest:** Mandy B.A. Paterson is employed as the principal scientist by RSPCA, Qld. None of the authors receive any interest or financial support from people or organisations who can bias the research.

#### **References**






© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Do Behaviour Assessments in a Shelter Predict the Behaviour of Dogs Post-Adoption?**

#### **Liam Clay 1,\*, Mandy B. A. Paterson 1,2 , Pauleen Bennett <sup>3</sup> , Gaille Perry <sup>4</sup> and Clive J. C. Phillips <sup>1</sup>**


Received: 11 June 2020; Accepted: 16 July 2020; Published: 18 July 2020

**Simple Summary:** In shelters it is usual to conduct standardised behaviour assessments on all incoming dogs. The information gathered from the assessment is used to identify dogs that are suitable for adoption and assist in matching dogs with suitable adopters. We investigated the predictive value of the standardised behaviour assessment protocol currently used in an Australian shelter for dog behaviour post-adoption. A total of 123 dogs, aged 1–10 years and housed in an animal care shelter, were assessed before they were adopted. The new owners of the dogs took part in a post-adoption survey conducted 1 month after adoption, which explored the behaviour of their dog in its new home. Regression analyses identified that friendly/social, fear and anxiousness identified in the shelter assessment significantly predicted corresponding behaviours post-adoption. However, behaviour problems, such as aggression, food guarding and separation-related behaviours, were not reliably predicted by the standardised behaviour assessment. We recommend that dog behaviour assessments in shelters are used only in conjunction with other monitoring tools to assess behaviour over the whole shelter stay, thus facilitating increased safety/welfare standards for dogs, shelters and the wider community.

**Abstract:** In shelters it is usual to conduct standardised behaviour assessments on admitted dogs. The information gathered from the assessment is used to identify dogs that are suitable for adoption and assist in matching the dog with suitable adopters. These assessments are also used to guide behaviour modification programs for dogs that display some unwanted behaviours. For some dogs, the results may indicate that they are unsuitable either for re-training or for adoption. In these circumstances the dogs may be euthanised. We investigated the predictive value of a standardised behaviour assessment protocol currently used in an Australian shelter for dog behaviour post-adoption. A total of 123 dogs, aged 1–10 years and housed in an animal care shelter, were assessed before they were adopted. The new owners of the dogs took part in a post-adoption survey conducted 1 month after adoption, which explored the behaviour of their dog after adoption. Ordinal regression analyses identified that friendly/social, fear and anxiousness identified in the shelter assessment significantly predicted corresponding behaviours post-adoption. However, behaviour problems, such as aggression, food guarding and separation-related behaviours, were not reliably predicted by the standardised behaviour assessment. The results suggest that further research is required to improve the predictability of behaviour assessment protocols for more specific behaviour problems, including different categories of aggression and separation-related problems. We recommend that dog behaviour assessments in shelters are used only in conjunction with other monitoring tools to assess behaviour over the whole shelter stay, thus facilitating increased safety/welfare standards for dogs, shelters and the wider community.

**Keywords:** dog behaviour prediction; dog behaviour problems; dog behaviour assessment; canines; animal shelters; dog post-adoption behaviour

#### **1. Introduction**

In Australia, the Royal Society for the Prevention of Cruelty to Animals (RSPCA) is a National, not-for-profit organisation that accepts approximately 46,000 dogs per year [1]. A 2014 study [2] found that these dogs, most of which were adult, were most commonly admitted after being collected by local council officers as strays (34%). Others were presented by members of the public as strays (24%), owner surrenders (19%), or euthanasia requests (4%), with a small number being brought in by Humane Officers, employees of the RSPCA tasked with rescuing animals from situations where their welfare may be compromised (6%). Other studies have shown that relinquishment reasons are usually human-related (unwanted, changed circumstances, financial, owner's health, household problems) but medical issues and behavioural problems also lead people to relinquish their dog [3–10]. In the Australian study [2] most dogs were either reclaimed (32%) or adopted (43%), with 14% euthanised. Reasons for euthanasia were dog behaviour (53%), dog health (23%), and owner requested (20%). If euthanased for behavioural reasons, it is likely that the dog displayed severe aggression, fearfulness and/or escaping behaviour.

Many shelters attempt to identify behavioural problems by continually monitoring behaviour and by formal behaviour assessments (BAs) while dogs are in care [11–13]. The behaviour assessments aim to identify behaviours that may cause problems in the dog's future home, and to give an overview of the dog for potential adopters [14]. However, their ability to predict future behaviour or behavioural issues is questioned [15]. There is a concern that dogs that appear aggressive during a BA are being unnecessarily euthanased because they would not necessarily be aggressive in a home environment, and that non-aggressive dogs may be adopted out only to become aggressive at a later stage in the new home.

Life in a shelter is stressful and traumatic for dogs due to sensory overstimulation, social isolation, change/loss of control of daily routines and the novelty of the environment [14,16,17]. Stress has wide-ranging impacts, including on cognitive ability, behaviour and the dogs' emotional state [18–20]. Therefore, a standardised BA conducted in shelters may not provide an accurate representation of the normal behaviour of the dog in a more stable and settled home environment.

Research conducted by Mornement et al. [14] in Australia compared the results of a Behaviour Assessment for Re-homing, K9 s (B.A.R.K.), administered in shelters, with results of a post-adoption survey. They reported that the only predictable outcomes were friendliness and fear-related behaviours. However, other behaviours, in particular aggression and food guarding, are rare post-adoption; Mohan-Gibbons [21] found that only six out of 96 adopted dogs were reported to display at least one incident of food guarding in the first 3 weeks, and at 3 months the adopters reported no food guarding behaviours at all. There was no evidence in this study, or a subsequent study [22], that food guarding increased return of the dogs to the shelter. In addition, injuries to staff, volunteers and adopters were rare and did not change if the food guarding test was omitted from the assessment.

'Time alone' tests have been used to identify dogs with separation-related behaviours [23]. Separation causes dogs to exhibit anxiety when away from owners or people in general; it is expressed as vocalisation, destruction of their environment, excretion, drooling, attempting to escape and depression-like responses [24,25]. Most shelters include a time alone test in their BA, during which the dog is placed alone in an unfamiliar room and observed for up to 10 min [23]. Dogs with separation-related anxiety spend the majority of the time vocalising, orienting to escape, panting and engaging in destructive behaviour.

Despite the current controversy about the use of BAs in shelters to gain an understanding of a dog's behaviour and to identify any major or minor behavioural problems, we consider that assessments still have a role to play [26]. They can be used to identify stable behaviours. To further our understanding of how well BAs can predict dog behaviour in adoptees' homes, we aimed to identify whether the standard BA protocol conducted at a Queensland shelter 5 days after admission predicted behaviour in adopters' home environment, as assessed 1 month post-adoption.

#### **2. Materials and Methods**

#### *2.1. Ethical Approval*

This study was conducted with the approval of The University of Queensland Human Ethics Committee (2017000044). The RSPCA Animal Welfare and Ethics committee approved the use of data from the RSPCA Queensland survey of adoptees and behaviour assessment data.

#### *2.2. Subjects*

The dogs used in the study were housed at the RSPCA Queensland Animal Shelter at Wacol. Before inclusion in the experiment, dogs were assessed by a veterinarian and identified as having no apparent medical problems. Upon admission to the RSPCA, behaviour profiles were completed by the owners for owner-surrendered dogs (these were not available for stray dogs). Each dog was then evaluated by an RSPCA behaviour assessor using the RSPCA Qld. behaviour assessment 5 days after admission [13]. Data were collected from 955 dogs. Of the 955 owners that adopted these dogs, 125 were successfully contacted later and completed a post-adoption survey (14% response rate). Two owners initially agreed to participate in the study when contacted but later declined to take part.

#### *2.3. Behaviour Assessment*

A standardised behavioural assessment (Supplementary Materials) was conducted on all dogs during their stay at RSPCA Queensland by two staff (one Handler and one Observer/Rater) responsible for evaluating the dogs' suitability for re-homing. These assessments were not able to be repeated due to staffing changes, therefore intra-rater and inter-rater reliability assessments were not possible. The assessments monitored the following behaviours: room exploration, behaviour when on a leash, sociability, tolerance, play behaviour with toys, tag (run and freeze), possessive behaviours, toddler and stranger interaction, time alone and social interactions with other dogs (RSPCA, 2012) [13]. The assessment comprised 11 different tests performed over a 15 min period, 10 have previously been described in detail [13]. The additional test 'Response to a fake cat" is outlined in Supplementary Materials. The equipment used followed RSPCA Queensland's protocol and included a 1.8 m leash, tennis ball, squeaky toy, rope, plastic hand on an extended pole, bowl, raw hide or bone, and combination of wet and dry dog food. At the conclusion of the behavioural assessment, animals were either deemed suitable for re-homing (*n* = 772), enrolled in a behaviour modification program (*n* = 133) or scheduled for euthanasia (*n* = 50). Decisions for behaviour modification and/or euthanasia were made by a professional review panel.

#### *2.4. Behaviour Scoring by RSPCA Assessors*

In each test, one RSPCA assessor rated the behaviour of the dog using binary occurrence of behavioural states (present or absent), except for the resource guarding test, which relied on a score by the assessor on an 8 point scale (Table 1). An overall score using the 11 tests was determined. All behaviours were assessed in each test using binary scoring (present or not) (Table 2)


**Table 1.** Resource guarding scoring system aimed at identifying possessive aggression by the dogs in defence of food.


#### *2.5. Post-Adoption Phone Interview*

Participants were asked when adopting a dog if they would agree to be included in a post-adoption phone survey. The survey was conducted by RSPCA customer service staff 1 month after adoption of the dog. The phone survey asked about the dog's behaviour in the home environment and in different everyday situations (Supplementary Materials). It took approximately 10 min to complete and consisted of 36 multi-choice questions with the option to add additional information.

Participants rated the frequency of socialisation to owners and children, and behaviour with run and freeze play, an unfamiliar person, unfamiliar children, an existing dog, an unfamiliar dog, and interactions with cats, on a 5 point scale (1: moves towards you in a playful manner, 2: moves, leans, or looks away, 3: no response, 4: moves or leans away in a manner that concerns you, 5: moves towards you in a way that concerns you).

#### *2.6. Statistical Analysis*

Statistical analysis was conducted using Minitab 18. Behaviour data were first screened for errors and then transposed into percentage of occurrence in tests for descriptive analyses. Ordinal logistic regression analysis using a logit model was used to identify behaviours in the assessment that best predicted dog behaviour post-adoption.

#### **3. Results**

#### *3.1. Descriptive Details*

The sample included 123 companion dogs (males: 61, females: 62) over the age of 1 year and under 10 years. The sources for the 125 dogs were as follows: owner surrender (45%); transfer (17%); RSPCA officer intake (13%); stray (12%); return (6%); lost (5%); emergency ambulance intake (3%); and pound (1%). The majority of dogs in the study were mixed breeds (45%). Median time of stay in shelter was 55.5 days (range 3–114 days).

#### *3.2. Behaviour Assessment (Table 3)*

The number of dogs displaying the different behaviours during each test is presented in Table 3.

In Test 1, "Exploring the Room", in the Exploration and Upon Call phases, dogs had a high occurrence of Friendly behaviour, with low occurrences of Anxious, Fear, and Arousal behaviours (Table 3). In Test 2, "Tolerance to Handling", in all components the majority of dogs displayed friendly interactions with the assessor, with increases in Anxious behaviours in Stroke and Foot Sensitivity (Table 3). In Test 3, "Startle Response", there was higher Avoidance, Fear, and Arousal in the Startle component, compared to the Recovery period, with a high occurrence of dogs displaying Friendly behaviours (Table 3). Recovery times varied between dogs, with 68% recovering within 5 s, 22% within 6–10 s and 3% taking over 10 s (7% of dogs did not exhibit as startle response).

In Test 4, "Toy Interactions", there was a high occurrence of Play in all components of the test, with low instances of Fear and Anxious behaviour (Table 3). The component with the greatest number of dogs exhibiting Arousal was Rope interactions. In Test 5, "Response to Unusual/Predictable Stimulus", there were high occurrences of Friendly behaviour in the Run and Freeze components but low levels of Anxious, Arousal and Fear behaviours (Table 3). In Test 6 (data not shown), "Resource Guarding", dogs displayed a high occurrence of levels between 2 and 3 with wet (68.2%) and dry food (80%). There were low occurrences of levels 4–6 with bone (9.9%) or pig's ear (7.43%).

**Table 3.** Number of dogs (and %) exhibiting behaviour's in the various test components in the behavioural assessment of shelter dogs (*<sup>n</sup>* = 123).


In Tests 7 and 8 "Stranger Interactions" and "Toddler Interactions", there were high occurrences of dogs displaying Friendly behaviour, with under 10% displaying Anxious or Displacement behaviours, Fear, or No Response towards the stranger (Table 3). Furthermore, there was only one dog that displayed Aggressive behaviour in each test. In Test 9, "Fake Cat", there were high occurrences of Friendly behaviour towards the fake cat, with minimal dogs displaying other behaviours (Table 3). In Time Alone (Test 10), 51% of dogs displayed Separation-Related behaviours, 31.4% displayed no problematic behaviours and 18% displayed Anxious behaviours.

Finally, in Test 11, "Behaviour with Another Dog", Friendly behaviours had the highest occurrence in dogs in all components of the test, with low levels of all other behaviours (Table 3). One interesting finding was the higher instance of Reactivity towards the opposing dog during the Walking component, which did not occur in the Circling or Nose to Nose components (Table 3).

#### *3.3. Post-Adoption Behaviour*

Only three participants no longer had the dog they had adopted. The remaining 120 participants still had their dog. With regard to the dogs' living arrangements, 49% were indoor/outdoor dogs, 29% mainly indoors and 23% mainly outdoors.

Participants were asked how the dog responded to different situations (Table 4) with most owners outlining that the dog "moves towards the stimulus in a playful manner" and a low occurrence of the opposite response. In situations related to unfamiliar visitors and unfamiliar dogs, there were higher levels of "moves, leans or looks away", "moves or leans away in a manner that concerns you", and "moves towards in a way that concerns you" (Table 4).


**Table 4.** The percentage (%) of dogs (*n* = 120) displaying specific behaviours post-adoption.

In terms of interactions with cats, 93 (74%) participants did not answer, with 32 participants answering that their dogs interact with cats with 19% of dogs moving towards them in a playful/friendly manner, and under 3% displaying other behaviours. With respect to resource guarding, participants were asked whether they were concerned about their dog's behaviour around food, treats, toys, and human food; over 90% reporting that there were no issues and under 10% saying there were issues (Table 5).

**Table 5.** The percentage (%) of dogs (*n* = 120) displaying possessive behaviour post-adoption.


Participants were asked how their dog reacts to a loud noise or something else startling the dog. 37% ignored the question, 25% reported a mild startle response from their dog, 9% of dogs ran and hid, and 4% displayed a pronounced startle response. With dogs that were startled, participants were asked how long it took them to recover; 45% recovered immediately, 29% recovered within a few

seconds, 15% recovered between 5 and 10 s, and 11% took longer than 10 s, avoided the situation and did not settle.

Participants were asked if they had ever left the dog alone, with 114 saying yes, and only nine saying no. Of the 114 participants that responded yes, 59% of dogs were left outside, 24% were left inside, 14% were allowed a combination of inside and outside, and 3% were left in a laundry or garage. Time spent alone ranged from 5 to 12 h (55%), 1–4 h (36%) and less than an hour (9%). Participants were asked whether their dog's behaviour changed when they were preparing to leave, with 72% reporting no change and 28% some changes in behaviour. Participants were asked if any behaviours were of concern, with 80% saying no, and 21% saying yes.

#### *3.4. Standardised Assessment Scores Verses Owner Surveys*

Ordinal regression analyses were conducted to determine whether scores derived from the behaviour scores in assessment tests could predict behavioural traits in the new home using reported behaviour in the home environment as the dependent variable. Questions from the survey that called for a response along a 5-point scale were related to relevant tests in the assessment that measured interactions with the handler, children, strangers and dogs, as well as the startle response, response to usual stimulus, food items and time alone situations. The regression analyses found that friendly/social behaviours (scored in tests: Interaction with Assessor in exploration of room, Response to unusual/unpredictable stimulus, Stranger interactions, Behaviour with another dog) significantly predicted 'playful/friendly manner' behaviour post-adoption in interactions with owners, children, strangers, existing dogs and unfamiliar dogs (Table 6). Anxious behaviour (scored in the tests: Assessor in exploration of room, Response to unusual/unpredictable stimulus, Fake toddler doll and Behaviour with another dog) significantly predicted 'Moving towards owner/children/stranger in a way that concerns you' behaviour post-adoption with interactions with owners, unfamiliar child, running and freezing, and unfamiliar dog (Table 6). Fear (scored in the tests: Assessor in exploration of room, and Fake toddler doll) significantly predicted 'Moves or leans away in a manner that concerns you' post-adoption with interactions with owners, and children (Table 6). The remaining 13 post-adoption behaviours were not predicted by the standardised behaviour assessment protocol conducted at the shelter.


**Table 6.** Significant or trend level (*p* < 0.10) relationships between behaviours scored from the shelter behaviour assessment and responses in the post-adoption survey, analysed by ordinal logistic regression.

#### **4. Discussion**

The aim of this paper was to evaluate how well the standardised behaviour assessment (BA) protocol currently used in a Queensland RSPCA shelter predicted post-adoption behaviours. In general, the ability of the standardised BA protocol to predict specific behaviours post-adoption was only somewhat effective. It appears, then, that the standardised BA may, as previous authors have outlined [16], be useful as a tool for providing an overall measure of dog behaviour, particularly with respect to friendly, fearful, and anxious behaviour, but that it requires supplementation with other sources of information. However, our study was unable to adequately assess whether behavioural

problems, specifically the identification of different categories of aggression, possessive behaviour (resource guarding), or separation anxiety, can be predicted from shelter assessments, since dogs displaying these behaviours were not rehomed.

There are several possible explanations for why the assessment was not more strongly predictive of our outcome measures. One constraint is that we cannot predict how an owner's behaviour or personality, and other animals/individuals in the household, can influence/affect the dog's behaviour post-adoption. Such effects may be substantial. Due to this, it may not be realistic to expect to be able to predict with accuracy behaviour over time.

A further explanation is that the standardised protocol may be inadequate as a tool to assess complex canine behaviours and behavioural problems either because of the structure of the assessment and/or its administration or due to the complex nature of such behavioural problems. We argue that the instrument is unlikely to be inadequately designed as it draws upon countless research studies and has been used and modified over many years [14,27–30]. The administration is also unlikely to have been inadequate, due to the standardised nature of the tests. Staff were trained and evaluated in the shelter, with the majority of the dogs in the large sample being assessed by the same individuals.

Another possible explanation is that due to the nature of canine behaviour, only some aspects of behaviour are stable [31,32]. Some aspects of canine behaviour may not be predictive in a single test, including aggression or other behaviour problems. Consistent with this idea was the number of new owners who reported their dog moving towards an individual in a way that concerned them, even though these dogs did not show these behaviours in the shelter assessment, or were not identified by shelter staff as displaying aggressive tendencies outside of the assessment. Dogs that displayed aggressive tendencies in the BA, or at other times during their stay at the shelter in the Queensland facility, were reviewed by a consultant for further testing. Such dogs were either then enrolled in a behaviour modification program or deemed to be unsuitable for adoption. Indeed, this study is similar to other studies in the area of canine behaviour assessment in shelters [12,21,22], where only dogs that did not show signs of aggression were made available for adoption and therefore included in the sample.

This suggests that there is a high possibility of a number of false negatives in the initial BA, which therefore is not offering a valid index of aggression. As seen in numerous studies, to reliably identify aggression and diagnose its causation is difficult, due to its infrequency and the nature of behavioural problems. Canine aggression is complex, and may be context specific [33]. The belief that one can assess a dog and diagnose it as aggressive is incorrect and should not be done. A specialist trained to identify and classify canine aggression would be in a better position to have a comprehensive understanding of physiology, behaviour and neurology, thus allowing a more nuanced diagnosis to be drawn [34]. Even in an assessment used primarily for identification of aggression, for example, the Dutch Socially Acceptable Behaviour (SAB) test, a portion of aggressive dogs remained undetected and the test was substandard for the assessment of types of aggression unrelated to fear [35]. This leads to the idea that fearful and anxious behaviours may be more stable and easier to detect than forms of aggression that can be motivated by numerous factors [17].

The final possibility is that canine behaviour may be predictable and the standard BA protocol used may be adequate at measuring certain categories of common/prominent canine behaviours (Friendly, Fearful, Arousal, Anxious), due to the common occurrence of these behaviour in everyday populations. However, due to the administration of the assessment after 5 days in the new environment, the tests may produce deceptive results. While many shelters maintain the highest standards of animal welfare, dogs still suffer from social isolation, abnormal sleep patterns, auditory pollution, olfactory overstimulation, and emotional stress, especially if individuals have no prior experience in shelters and do not habituate using positive coping mechanisms. The stressors that are inherent in any shelter may force some dogs to employ negative coping mechanisms (avoidance, inhibition or appeasement) as an outlet rather than displaying aggression [36,37]. This may especially be the case after surrender and over the first few days of entering the shelter, with some dogs likely to experience acute stress and social isolation [17]. Research into this area has found that shelter dogs showed more aggression when tested 2 weeks

after being admitted to a shelter in comparison to 1–2 days after surrender [38]. Furthermore, only a few studies have studied the relationship of aggression with welfare standards for dogs [17,20] and whether the behaviour is due to environment stressors. Evidence in the literature suggests that stress can have an effect on cognitive function, negative emotional state and behaviour [18–20]. This implies that standardised canine BAs, timed incorrectly and used to make decisions about dogs (rehomed, trained or euthanised), may give false information to shelter staff.

Consistent with this possibility, recent studies into the test used to identify food resource guarding found the prevalence of issues post-adoption were low and that removal of the test did not increase the likelihood of food guarding in the new home [21,22]. The reason for this result can be identified in the complex aetiology behind food resource guarding. It is defined as the use of avoidance, threatening or aggressive behaviours by a dog to retain control of food or non-food items in the presence of a person or other animal [39]. It is not surprising that many dogs are so labelled in a shelter environment, due to the high occurrence of acute stress from sensory overload causing dogs to feel threatened and in turn aggressive. However, outside of the shelter environment, in a non-threatening and predictable environment, this reaction decreases. In addition, other types of aggression, such as territorial and maternal, remain very difficult to assess in shelters [33,40].

We advocate that shelters must look for a new approach that allows an improved ability to identify behaviour problems in a more stable environment. One such solution currently implemented in RSPCA Queensland shelters is the use of a foster care system, in which dogs that are unable to cope in the shelter are housed with foster carers until they are able to be adopted. This solution allows dogs to live in a stable environment with minimal exposure to stressors that may otherwise lead to the deterioration of the dog's behaviour thus leading to behaviour problems. Furthermore, it allows shelters to house more dogs able to cope in the shelter environment, as well as individuals requiring behaviour modification and further testing of behaviour problems. In addition, RSPCA Queensland uses a qualified behaviourist to help to understand dogs that are identified in the behaviour assessment as having behavioural issues. The consultant conducts further tests to better identify the behavioural problems and implement behaviour modification programs with the use of qualified dog trainers. The dogs are constantly reviewed and evaluated to monitor progress over time.

However, implementing these solutions requires resources that most shelters do not have. Most shelters have financial, time and staff constraints that hinder them utilising such techniques. The authors understand that no one BA protocol has the ability to accurately predict every future behaviour, but these assessments can be used as one tool in conjunction with continual monitoring of behaviour and health of dogs in shelters, to gain an overview of the dog's behaviour and identify dogs that require further testing or behaviour modification. Additionally, BAs can be used as monitoring tools to identify dogs not coping in the novel shelter environment. This, in conjunction with surrender information, veterinary monitoring and evaluations, in-kennel scoring from staff and volunteers, and behaviour modification should help develop a better system for shelters. To achieve this, continuous improvement and studies into dog behaviour in shelters are required.

#### **5. Conclusions**

Findings from this study suggest that a standardised behaviour assessment protocol used at an Australia shelter is a useful tool to predict some behaviours, mainly, friendly, fearful, arousal and anxious behaviours. However, in the predictability of behaviour problems, such as different categories of aggression or separation anxiety, it appears largely ineffective. This may be a result of the assessments being conducted in a highly stressful/novel environment where dogs experience many stressors in addition to lack of a human–animal bond, and then trying to use that information to predict home behaviour in a stable environment where supportive social bonds have formed. A thorough review of the protocol is recommended to identify any possible improvements, and care should be taken if the BA is the only tool used to identify a dog's adoption suitability. However, using the BA as one tool in a toolbox of many others, including pre-surrender information, veterinary clinical

assessments, monitoring in kennel and responses to training, may provide a more comprehensive picture of behaviour. Behaviour is multifactorial, requiring an in-depth understanding of multiple neurological and physiological processes. Therefore, continuous research and training in shelters together with ongoing support may help gain a better understanding of canine behaviour.

**Supplementary Materials:** Supplementary Materials are available online at http://www.mdpi.com/2076-2615/10/7/ 1225/s1.

**Author Contributions:** L.C., M.B.A.P., P.B., G.P., and C.C.J.P. conceived the project. L.C. drafted the paper and all authors had input into modifying it into the present format. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** The authors acknowledge the assistance of RSPCA Queensland.

**Conflicts of Interest:** Mandy Paterson declares that she works for the RSCPA Qld. Liam Clay declares that RSPCA Qld funded his studentship. Apart from this, no other author has any conflict.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Characteristics and Adoption Success of Shelter Dogs Assessed as Resource Guarders**

#### **Betty McGuire**

Department of Ecology and Evolutionary Biology, Cornell University, Ithaca, NY 14853, USA; bam65@cornell.edu

Received: 16 October 2019; Accepted: 14 November 2019; Published: 17 November 2019

**Simple Summary:** Dogs that aggressively guard resources, such as food, toys, and sleeping sites, can pose risk to people unfamiliar with canine communication. Such dogs also present challenges to animal shelters, which typically screen for food-related guarding during behavioral evaluations. Some shelters euthanize dogs that aggressively guard food, whereas others restrict adoptions. However, few studies have examined the characteristics and adoption success of dogs that guard food in shelters. I analyzed demographic data and adoption success of dogs assessed as resource guarders at a shelter in New York (NY) over a nearly five-year period. Fifteen percent of the dog population was identified as resource guarders during shelter behavioral evaluations. Resource guarding was more common in adults and seniors than in juveniles, and it was more common in small and large dogs than medium-sized dogs. While spayed females were more likely than intact females to guard food, neutered males and intact males did not differ in their propensity to guard food. Dogs that showed severe guarding were more likely to be returned by adopters, but almost all were successfully re-adopted. These findings provide a detailed description of food guarders in a shelter dog population and show that most such dogs were successfully re-homed.

**Abstract:** Some domestic dogs aggressively guard resources. Canine resource guarding impacts public health through dog bites and affects dog welfare through adoption and euthanasia policies at animal shelters. However, little is known about the demographic characteristics and adoption success of dogs assessed as resource guarders during shelter behavioral evaluations. I reviewed nearly five years of records from a New York (NY) SPCA and categorized 1016 dogs by sex; age; size; reproductive status; and resource guarding. I then examined how these characteristics influenced the returns of dogs by adopters. The prevalence of resource guarding in this shelter dog population was 15%. Resource guarding was more common in adult and senior dogs than in juvenile dogs; and it was more common in small and large dogs than medium-sized dogs. Spayed females were more likely than intact females to guard food; neutered males and intact males did not differ in their likelihood of food guarding. Most dogs identified as resource guarders showed mild to moderate guarding. Severe guarders were more likely to be returned by adopters; although almost all were eventually re-adopted and not returned to the shelter. Data presented here provide the most comprehensive description of resource guarders in a shelter dog population and show the successful re-homing of most.

**Keywords:** dog; food aggression; food guarding; resource guarding; shelter; behavior; adoption; return rate

#### **1. Introduction**

Some domestic dogs are possessive of resources such as food, toys, and sleeping sites, and they display threatening or aggressive behavior when a person approaches or attempts to take control of the resource. Such resource guarding occurs in homes and in animal shelters. One survey of animal shelters in the United States found that most shelters test for food guarding as part of their canine behavioral

evaluations, and about half do not make available for adoption dogs assessed as food aggressive [1]. Shelters that make food guarding dogs available for adoption often place restrictions on who can adopt them (e.g., experienced dog owners with no young children in the household), which can prolong the time these dogs remain in shelters [2,3]. Thus, canine resource guarding can impact not only public health through dog bites, which are the most extreme form of guarding behavior [4], but also dog welfare via shelter policies on adoption and euthanasia [1–3]. The need for informed re-evaluation of shelter policies whereby all dogs classified as food aggressive are euthanized is especially critical given evidence that dogs assessed as food aggressive in a shelter do not necessarily guard food in their adoptive home [1,3]. In addition to these findings specific to food aggression and the predictive utility of behavioral evaluations [1,3], other research has more generally revealed the inadequacies of behavioral evaluations [5–7].

Few studies have examined characteristics of dogs that guard food. Most such studies have been based on owner reports of food guarding by dogs in the home [4,8,9] rather than observations of dogs during behavioral evaluations at shelters; one study used both shelter evaluations and reports from adopters [3]. For studies based on owner reports, two found that mixed breed dogs were more likely than purebred dogs to guard food [4,8]. Owner reports also identified increasing age of dog at acquisition as a predictor of food guarding [8]. A dog's body size, as estimated by height at withers, was found to be negatively correlated with owner-directed aggression, a category that included resource guarding [9]. Conflicting results have been obtained regarding the influence of the sex of dogs on the likelihood of resource guarding. One study, based on owner reports, indicated that males were more likely than females to guard resources, and this was particularly true for neutered males [4]. In contrast, Marder et al. [3] found no sex difference in incidence of food aggression based on either shelter behavioral evaluations or subsequent reports by adopters. To date, no study has examined multiple demographic characteristics of dogs assessed as displaying food aggression during behavioral evaluations at shelters; current information is limited to one study that examined the influence of a single characteristic, the sex of shelter dogs, on the likelihood of food guarding during behavioral testing [3]. Understanding the characteristics associated with the expression of food guarding could serve as the basis for future research on the causation of guarding behavior [4]. Additionally, given that some shelters do not behaviorally evaluate all dogs made available for adoption [2], information on additional characteristics that might be associated with food aggression, such as age, reproductive status, and body size, could be useful to staff making decisions concerning which dogs to evaluate.

Two studies have examined the adoption success of dogs assessed as food aggressive in shelters. Mohan-Gibbons et al. [1] identified 96 food aggressive dogs at one shelter, placed them on a behavior modification program, and contacted their adopters three times in the months following adoption (adopters were asked to continue the behavior modification program that had begun in the shelter). Marder et al. [3] followed 97 shelter dogs, some of which were food aggressive and others not, and contacted adopters at least three months after adoption. Both studies found that dogs assessed as food aggressive during shelter behavioral evaluations did not necessarily guard food in their new homes, although the percentages of adopted dogs that continued food guarding in the home varied considerably, ranging from less than 10% [1] to 55% [3]. Results from both studies indicated that even if dogs displayed food aggression in the home, adopters did not consider the behavior to be a major challenge [1,3]. Mohan-Gibbons et al. [1] also found that the return rate for dogs assessed as food aggressive at their study shelter was lower than that for dogs assessed as not food aggressive; Marder et al. [3] did not provide return rates. No study has examined how food guarding, when considered with demographic characteristics such as sex, age, and body size, influences return rates. Additionally, no study has examined how the severity of food aggression (mild to moderate versus severe) influences return rates.

To further inform shelter policies regarding resource guarding dogs, additional information is needed on the demographic characteristics and adoption success of dogs identified as resource guarders during shelter behavioral evaluations. This paper considers food-related guarding; it does not consider other forms of resource guarding, such as the guarding of toys or sleeping sites. I reviewed nearly five years of records from a New York (NY) SPCA to develop a demographic profile for dogs assessed as resource guarders at the shelter and to determine the success of these dogs once adopted. I first examined whether sex, age class, reproductive status (intact versus spayed or neutered), or body size could be used to predict resource guarding, and then I assessed how these demographic characteristics, along with resource guarding, influenced the returns of dogs by adopters.

I predicted that likelihood of resource guarding would increase with the age of dogs, given the association found between food guarding and increasing age of dogs at acquisition [8]. Based on data indicating that behavioral problems are more common in small dogs than in large dogs [9], I predicted that small dogs would be more likely than medium and large dogs to display guarding behavior. I did not expect likelihood of resource guarding to vary by sex or reproductive status, given findings of no sex difference in the incidence of food guarding during shelter evaluations [3] and little or no effect of gonadectomy on aggression directed by dogs to people [10]. Adopters often cite behavioral problems as their reason for returning dogs to shelters [11–14], so I predicted that dogs assessed as resource guarders in the shelter would have higher return rates than dogs assessed as non-resource guarders and that severe resource guarders would be returned more frequently than dogs that showed mild to moderate guarding or no guarding.

#### **2. Materials and Methods**

I analyzed records of dogs at the Tompkins County SPCA in Ithaca, NY, USA. These records included data input by shelter staff into the PetPoint data management system from 1 September 2014 through 31 May 2019. Records included information on dog intakes (including returns), behavioral evaluations, and adoptions (*n* = 1016 adopted dogs; puppies excluded, see Section 2.2). Tompkins County SPCA is a no-kill, open-admission shelter with scheduled intake. The shelter has a small set of dog foster parents and allows for overnight fostering with dog volunteers. Additional programs to promote dog adoptability include: volunteer dog walking, volunteer in-kennel companionship, volunteer day-trips with dogs, playgroups for suitable pairs of dogs, nightly stuffed Kong enrichment, adoption promotion in local print and social media, off-site events to advertise dogs, and a volunteer group independently promoting dogs that are hard-to-place or have been at the shelter a long time. All procedures were carried out under protocol 2012-0150, which was approved by Cornell University's Institutional Animal Care and Use Committee.

#### *2.1. Dogs and Housing*

Original sources were available for 1015 of the 1016 adopted dogs whose records I reviewed: owner surrendered, 473 (46.6%); transferred from other shelters, 343 (33.8%); picked up as strays, 166 (16.4%); and seized by animal control officers, 33 (3.2%). Most dogs at the Tompkins shelter were mixed breeds; the number of purebred dogs was unknown due to lack of pedigrees and DNA analyses. A brief description of housing and care of dogs is provided here because details have been presented elsewhere [15].

At intake, dogs were housed in the rescue building in chain link cages with an indoor space (2.2 m2) and outdoor run (3.5 m2). Veterinary staff examined each dog at intake or within a few hours of intake and performed vaccinations, flea control, fecal exam, deworming, and a heartworm test. Following the veterinary exam, each dog was scheduled for behavioral evaluation (see Section 2.2). Within a few days of the completion of the behavioral evaluation, dogs were moved to the pet adoption center, adjacent to the rescue building. Once on the adoption floor, dogs were housed in one of 13 cubicles, which ranged in size from 5.2 to 7.3 m2. Almost all dogs were housed individually; only dogs that came in together and staff deemed needed to stay together shared the same cubicle. Each cubicle contained a water bowl, a raised bed, a blanket, and a toy. Volunteers or staff either walked the dogs or brought them to an outdoor enclosure several times a day. Staff fed the dogs each day between 08:00

and 09:00 h and again between 15:00 and 16:00 h. Intact dogs were spayed or neutered when housed in either the rescue building or the pet adoption center; all dogs were spayed or neutered before adoption.

#### *2.2. Behavioral Evaluations*

Shelter staff evaluated each dog's behavior using a series of tests based on Sternberg's Assess-a-Pet [16], with modifications described by Bollen and Horowitz [17]; these modifications were made as part of the shelter's standard operating procedures and were in place well in advance of the present study. Behavioral evaluations were performed approximately 3 days after intake and included nine tests in the following sequence: cage presentation; sociability; teeth exam; handling; arousal; food bowl (using a mix of kibble and canned food); possession (using a raw hide chew, pig ear, etc.); human stranger; and dog-to-dog. Behavioral responses on the food bowl test were organized into seven levels, listed in order of increasing intensity of response: (1) stopped eating and backed away from the dish; (2) continued eating without showing any signs of uneasiness; (3) moved muzzle deeper into the dish and ate faster; (4) stiffened slightly; (5) moved muzzle toward the Assess-a-Hand; (6) stiffened, exhibited whale eye, and snarled; and (7) froze, growled, lunged, snapped, and bit the Assess-a-Hand. Behavioral responses on the possession test were organized into five levels, also listed in order of increasing intensity of response: (1) readily dropped the item; (2) allowed the Assess-a-Hand to take the item; (3) resisted letting go of the item but did not show outward aggression; (4) stiffened, exhibited whale eye, and snarled; and (5) froze, growled, lunged, snapped, and bit the Assess-a-Hand. When a dog was very uncomfortable with the Assess-a-Hand, the evaluator used her own hand to remove the food bowl and chew. Dogs were assessed as resource guarders if they exhibited at least one of the following behaviors during either the food bowl test, possession test, or both tests: stiffened, exhibited whale eye, snarled, froze, growled, lunged, snapped, or bit the Assess-a-Hand. For one analysis, I classified resource guarding dogs as exhibiting either mild to moderate resource guarding (stiffened, exhibited whale eye, snarled, froze, or growled) or severe resource guarding (lunged, snapped, or bit the Assess-a-Hand) during either the food bowl test, possession test, or both tests. This categorization was based on that described by Mohan-Gibbons et al. [2].

Though the behavior of puppies was formally evaluated by staff, the tests differed somewhat from those of older dogs (e.g., recent tests were conducted in the cubicle in which puppies were housed on the adoption floor rather than in the conference room where tests were conducted for dogs in older age classes). Additionally, puppy results were not input into the PetPoint database. For these reasons, puppies were not included in the present study.

#### *2.3. Statistical Analyses*

I classified dogs by sex, age class, body size, and reproductive status. The ages of dogs were estimated by shelter veterinarians. For the purpose of this study, the following age classes were defined based on those used in previous studies [18,19]: juveniles, from 4 months to <1 year; adults, from 1 year to <8 years; and seniors, ≥8 years. The number of dogs in each sex and age class during the study period was as follows: males, 100 juveniles, 348 adults, 66 seniors; females, 99 juveniles, 340 adults, and 63 seniors. I used the body mass recorded at veterinary intake exams to classify adult and senior dogs into the following size classes: small, <11 kg; medium, 11–24 kg, and large, ≥25 kg (categories modified from those used by Taylor et al. [20]). I did not assign juveniles a size class because they were still growing; thus, juveniles were excluded from data analyses in which body size was a variable. Mature dogs (adults and seniors) fell into the following size classes: small, 32.1%; medium, 37.7%; and large, 30.2% (body mass was not available for one adult female out of the combined 817 adults and seniors). The following percentages of dogs by sex and age class were intact at the time of behavioral evaluation: males, 83.0% of juveniles, 53.7% of adults, 30.3% of seniors; females, 85.9% of juveniles, 54.4% of adults, and 25.4% of seniors. The final dispositions of returned dogs were classified as adopted again and not returned, euthanized for either behavioral or medical reasons, transferred to a rescue group, or returned to the original owner (i.e., the person who originally surrendered the dog to the

shelter experienced a change in living situation such that he or she was able to take the dog back). The final dispositions of dogs returned toward the end of the study period (May 2019) were followed for an additional 4 months.

I used logistic regression to determine significant predictors of resource guarding. Fixed factors in the first model for resource guarding were sex, reproductive status, and age class (juveniles, adults, and seniors). I then excluded juveniles from the data set so that body size could be added as a fixed factor in the second model for resource guarding. I also used logistic regression to determine significant predictors of a dog being returned to the shelter by adopters. Fixed factors in the first model for likelihood of return were sex, age class (juveniles, adults, and seniors), and resource guarding status. In the second model for likelihood of return, I excluded juveniles from the data set so that body size could be added as a fixed factor. Finally, in the third model for likelihood of return, I considered the level of resource guarding and categorized dogs as non-resource guarders, mild to moderate resource guarders, or severe resource guarders, as defined in Section 2.2. For all models, I examined the main effects and two-way interactions; I dropped two-way interactions that were not significant from final models. All dogs were spayed or neutered prior to adoption, so reproductive status was not a fixed factor in any of the models for likelihood of return. Statistical analyses were completed in JMP Pro (version 13.1.0).

#### **3. Results**

#### *3.1. Resource Guarding*

Over the nearly five-year study period, staff evaluated the behavior of 1051 individual dogs (juveniles, adults, and seniors); 161 dogs were assessed as resource guarders, resulting in a prevalence of 15.3% of dogs evaluated. Fifteen of the resource guarding dogs were not made available for adoption: 10 were euthanized for behavioral reasons and one for medical reasons; three were transferred to rescue groups; and one was returned to her original owner. All of the results that follow pertain to the 1016 dogs that were behaviorally evaluated and made available for adoption.

Overall, 14.4% of dogs moved to the adoption floor were classified as resource guarders based on behavioral evaluations (146/1016; juveniles, adults, and seniors). Of the dogs assessed as resource guarders, 30.8% (45/146) guarded on the food bowl test, 83.6% (122/146) guarded on the possession test, and 17.1% (25/146) guarded on both tests. On both the food bowl test and the possession test, freezing was the most common behavior displayed by resource guarding dogs, and lunging was the least common (Table 1). The two most extreme behaviors, snapping and biting the Assess-a-Hand, occurred in less than 14% of resource guarding dogs (Table 1).

**Table 1.** The percentages of resource guarding dogs that displayed specific behaviors during the food bowl test and possession test. The number of dogs that displayed the behavior/number of dogs assessed as resource guarding on the particular test are in parentheses.


<sup>1</sup> Mild to moderate resource guarding included the behaviors from stiffened through growled; severe resource guarding included the behaviors lunged, snapped, and bit the Assess-a-Hand.

The percentages of dogs assessed as resource guarders in relation to main effects of sex, age class, reproductive status, and body size are shown in Table 2. Age class was a significant predictor of

resource guarding (*X*<sup>2</sup> = 13.53, *d.f.* = 2, *p* = 0.001), with adults and seniors more likely than juveniles to show guarding behavior (Table 2, second column). Seniors tended to be more likely than adults to guard resources (*p* = 0.08; Table 2, second column). There was a significant sex by reproductive status interaction for likelihood of resource guarding (*X*<sup>2</sup> = 5.24, *d.f.* = 1, *p* = 0.022). While spayed females (17.1%; 37/216) were more likely than intact females (9.1%; 26/286) to guard food, neutered males (15.2%; 34/224) and intact males (16.9%; 49/290) did not differ in their propensity to guard food. Neutered males and intact males also were more likely than intact females to guard food.

**Table 2.** The percentages of dogs assessed as resource guarders in relation to sex, age class, reproductive status, and body size. The number of dogs assessed as resource guarders/number of dogs evaluated and made available for adoption shown in parentheses. Within columns and specific variables, values with different superscript letters are significantly different (*p* ≤ 0.05).


<sup>1</sup> Age classes included in analyses.

When juveniles were excluded from the data set to allow for the inclusion of body size as a fixed factor in the model, body size was a significant predictor of resource guarding (*X*<sup>2</sup> = 7.05, *d.f.* = 2, *p* = 0.03), with small dogs and large dogs more likely than medium dogs to display guarding (Table 2, third column). Small and large dogs did not differ from one another in propensity to guard. With juveniles excluded, age class did not predict food guarding (*X*<sup>2</sup> = 1.63, *d.f.* = 1, *p* = 0.20; Table 2, third column). As before, there was a significant sex by reproductive status interaction for likelihood of food guarding (*X*<sup>2</sup> = 5.45, *d.f.* = 1, *p* = 0.02). While spayed females (18.3%; 37/202) were more likely than intact females (10.4%; 21/201) to guard food, neutered males (15.5%; 32/207) and intact males (20.3%; 42/207) did not differ in their propensity to guard food. Intact males were more likely than intact females to guard food.

#### *3.2. Returns of Dogs by Adopters*

Of the 1016 dogs adopted during the nearly five-year study period (juveniles, adults, and seniors), 181 (17.8%) were returned to the shelter at least once. The number of returns per dog ranged from one to six, with one being most common: one return, 80.7% (146/181); two returns, 17.1% (31/181); three returns, 1.7% (3/181); and six returns, 0.5% (1/181). The percentages of adopted dogs returned to the shelter in relation to main effects of sex, age class, resource guarding status, and body size are shown in Table 3. Age class did not predict likelihood of return (*X*<sup>2</sup> = 2.94, *d.f.* = 2, *p* = 0.23; Table 3, second column). There was a borderline significant sex by resource guarding status interaction for likelihood of return (*X*<sup>2</sup> = 3.80, *d.f.* = 1, *p* = 0.0514). While food aggressive males (27.7%; 23/83) were more likely than non-food aggressive males (17.2%; 74/431) to be returned, food aggressive females (14.3%; 9/63) and non-food aggressive females (17.5%; 77/439) did not differ in their likelihood of return. Food aggressive males also were more likely to be returned than food aggressive females and non-food aggressive females.


**Table 3.** The percentages of adopted dogs returned to the shelter in relation to sex, age class, resource guarding status, and body size. The number of dogs returned/number of dogs adopted shown in parentheses. Within columns and specific variables, values with different superscript letters are significantly different (*p* ≤ 0.05).

<sup>1</sup> Age classes included in analyses.

Body size was a significant predictor of a dog being returned (*X*<sup>2</sup> = 15.38, *d.f.* = 2, *p* = 0.0005), with large dogs more likely than small and medium dogs to be returned (Table 3, third column; juveniles excluded). Small and medium dogs did not differ from one another in their likelihood of return. Age class did not predict likelihood of return when the data set was restricted to adults and seniors (*X*<sup>2</sup> = 0.76, *d.f.* = 1, *p* = 0.38; Table 3, third column). As before, logistic regression revealed a sex by resource guarding status interaction for likelihood of return (*X*<sup>2</sup> = 4.47, *d.f.* = 1, *p* = 0.034). While food aggressive males (27.0%; 20/74) were more likely than non-food aggressive males (17.9%; 61/340) to be returned, food aggressive females (13.8%; 8/58) and non-food aggressive females (19.1%; 66/345) did not differ in their likelihood of return. Note, however, that the subsequent pairwise comparison between percentages of food aggressive males and non-food aggressive males returned by adopters fell short of statistical significance (*p* = 0.08). There was a tendency for food aggressive males to be more likely than food aggressive females to be returned (*p* = 0.06). The interaction between body size and resource guarding status was not significant (*X*<sup>2</sup> = 0.74, *d.f.* = 2, *p* = 0.69), indicating that returns of food aggressive dogs did not vary by size of dog.

Of the 146 dogs assessed as resource guarders at the shelter, 121 (82.9%) showed mild to moderate guarding, and 25 (17.1%) showed severe guarding. (Note: 25 does not equal the sum of number of dogs shown in Table 1 that lunged, snapped, and bit the Assess-a-Hand, because some dogs exhibited more than one of these behaviors during either the food bowl test, possession test, or both tests). The 25 dogs that showed severe guarding included two juveniles, 18 adults, and five seniors. Given the small numbers of juveniles and seniors in the severe guarding group, I did not include age class as a fixed factor in the third model for likelihood of return. When resource guarding was differentiated by level, guarding was a significant predictor of a dog being returned (*X*<sup>2</sup> = 6.72, *d.f.* = 2, *p* = 0.035), with severe guarders more likely to be returned (40.0%; 10/25) than mild to moderate guarders (18.2%; 22/121) and dogs classified as non-resource guarders (17.5%; 152/870). Dogs showing mild to moderate guarding and dogs classified as non-resource guarders did not differ from one another in likelihood of return. Sex did not predict likelihood of return when resource guarding was differentiated by level (*X*<sup>2</sup> = 0.40, *d.f.* = 1, *p* = 0.53).

Fifteen of the 25 dogs that exhibited severe guarding during behavioral evaluations were adopted and not returned to the shelter. For the remaining 10 dogs in the severe group, seven were returned once and then adopted without return; one was returned twice and then adopted without return; one was returned three times and then adopted without return; and one was returned twice and euthanized (this dog bit an adult in its second adoptive home). Thus, of the 25 dogs classified as

severe resource guarders at the shelter, 24 (96%) were eventually placed in a home and not returned to the shelter.

The canine surrender profile form of the Tompkins shelter includes the statement, "Please explain why you need to relinquish your dog in your own words." Reasons given for returns of the nine severe guarders that were eventually successfully re-homed included elimination in the house, owner allergies, unforeseen personal reasons, moving, aggression directed at another dog in the home, and over-arousal; one small dog bit the adopter's grandson. None of the adopters completing the form described aggression around food; one adopter, who chose to provide a lengthy written explanation rather than completing the surrender form, described over-excitement around food, but stated the reason for surrender was unforeseen personal reasons. The surrender form also includes the statement, "Please check all that apply to your dog's personality" and lists the following options: friendly, shy, independent, fearful, playful, affectionate, aloof, aggressive, and overly reactive. Two adopters listed shy, fearful, and overly reactive; one listed aggressive but to another dog; some combination of friendly, independent, playful, and affectionate were checked by remaining adopters.

Of the 181 dogs returned during the study period, one was brought to a shelter located in a different state, and his final disposition was unknown. The final dispositions for the remaining 180 dogs returned at least once to the Tompkins shelter were as follows: re-adopted and not returned to the shelter, 87.2% (157/180); euthanized for either behavioral or medical reasons, 8.9% (16/180); transferred to a rescue group, 2.2% (4/180); and returned to the original owner, 1.7% (3/180).

#### **4. Discussion**

Measures of prevalence and severity of resource guarding in dogs at the Tompkins County SPCA, as well as overall return rate, are similar to those reported previously for dogs at other shelters. Fifteen percent of dogs behaviorally evaluated at the Tompkins shelter were assessed as resource guarders. This measure of guarding prevalence is similar to those reported by Mohan-Gibbons et al. [1], who surveyed 77 shelters in the United States and found that percent of dog populations exhibiting food guarding ranged from 7–30%, with an average of 14%. In a later study involving nine shelters, Mohan-Gibbons et al. [2] reported that 17% of behaviorally evaluated dogs were classified as food guarders, and Marder et al. [3] found that 21% of dogs at one shelter exhibited aggression around food. Additionally, 83% of the dogs assessed as resource guarders at the Tompkins shelter showed mild to moderate guarding, and 17% showed severe guarding; these same percentages were obtained by Mohan-Gibbons et al. [2] for their sample of shelter dogs assessed as food guarders. Finally, the overall return rate at the Tompkins shelter (18%) was similar to the average return rate reported for shelters in the United States, United Kingdom, and Italy (15%; see review by Protopopova and Gunther [21]). The consistency in both the prevalence and degree of severity of resource guarding across shelters despite the use of different behavioral assessments (e.g., Assess-a-Pet, SAFERTM, blends and modifications of these and other assessments, as well as assessments developed by individual shelters) and the similarity in overall return rates suggest that the present findings on characteristics and returns of resource guarding dogs at the Tompkins shelter might generalize to other shelters.

Age class was a significant predictor of resource guarding in dogs at the Tompkins shelter, with adults and seniors more likely than juveniles to show food-related guarding during behavioral evaluations. Additionally, there was a tendency for seniors to be more likely than adults to guard food. This is the first study based on direct observations of dogs during behavioral evaluations at a shelter to examine the relationship between resource guarding and age. Using owner responses to a questionnaire distributed by an Australian dog magazine, McGreevy and Masters [8] reported that food-related aggression was associated with increasing age of dogs at acquisition (485 respondent households and a total of 690 dogs obtained from a variety of sources, including pet shops, breeders, pounds, shelters, friends, and family). Dogs included in the survey ranged from eight weeks to over 11 years old [22]; thus, this study differed from the present study not only in its method of obtaining information on dogs with respect to resource guarding but also in its inclusion of data on puppies

(results from evaluations of puppies were not available in the present study). Guarding behaviors have been described in puppies only a few weeks old [23]. It would be useful in future studies with shelter dogs to include data from puppies to provide a more complete picture of age-related patterns in resource guarding.

Body size, based on body mass, was a significant predictor of resource guarding in the dog population at the Tompkins shelter, with small dogs and large dogs more likely than medium dogs to display guarding behavior. Small and large dogs did not differ from one another in propensity to guard resources. Using behavioral and body mass data collected from dog owners who completed the Canine Behavioral Assessment and Research Questionnaire (C-BARQ; 49 common breeds were represented in the study sample) and height data drawn from breed standards, McGreevy et al. [9] found that height was negatively correlated with owner-directed aggression, a category that included resource guarding. More specifically, McGreevy et al. [9] found that shorter dogs were more likely than taller dogs to display threatening or aggressive responses to household members in a variety of situations, which included being roughly handled, stared at, challenged, stepped over, or approached when possessing food or objects; body mass, however, did not predict owner-directed aggression. It is possible that absence of a relationship between body mass and owner-directed aggression in the study by McGreevy et al. [9] reflected the same pattern found here, i.e., despite differing from medium dogs, small (light) and large (heavy) dogs did not differ from one another in their tendency to guard resources. However, direct comparison of the present results with those of McGreevy et al. [9] are difficult given major differences between the two studies in methods of classifying dogs with respect to guarding behavior (direct observations by shelter staff during behavioral evaluations versus owner reports), dog populations (primarily mixed breed dogs at a shelter versus purebred dogs in homes), and scope of behavioral categories (restricted to food-related guarding versus owner-directed aggression, which included resource guarding and several other situations involving dogs and household members). Nevertheless, the present finding that small dogs were more likely than medium dogs to show resource guarding is consistent with the general pattern that problem behaviors are more common in small dogs [9]. Factors underlying the present finding that large dogs were more likely than medium dogs to guard resources remain to be determined.

The effect of reproductive status on propensity to guard resources varied by sex at the Tompkins shelter, with spayed females more likely to guard than intact females, and no difference in guarding propensity between neutered males and intact males. Neutered males and intact males were more likely to guard than intact females and did not differ from spayed females. To my knowledge, the present study is the first based on shelter behavioral evaluations to examine the relationship between resource guarding and both sex and reproductive status in dogs. The present findings differ from those of Jacobs et al. [4], who surveyed dog owners and found that dogs showing aggressive resource guarding in the home were more likely to be male and neutered. Jacobs et al. [4] acknowledged that dogs in their study might have been neutered after showing resource guarding aggression, in which case neutering might be considered a consequence of aggression rather than a cause (age at castration and age at first display of resource guarding aggression were not obtained from owners). The present findings are consistent with the general conclusions of Farhoody et al. [10] that gonadectomy does not result in predictable decreases in aggression in all male and female dogs.

Initial analyses that coded dogs at the Tompkins shelter as either resource guarders or non-resource guarders indicated that the effect of resource guarding on likelihood of a dog being returned to the shelter varied by sex. More specifically, whereas food aggressive males were more likely to be returned than non-food aggressive males, food aggressive females and non-food aggressive females did not differ in their likelihood of return. In other words, food aggression either increased returns (in the case of males) or had no effect on returns (in the case of females). The only other data available comparing return rates of food aggressive and non-food aggressive dogs to shelters are those of Mohan-Gibbons et al. [1], who reported slightly lower return rates for dogs identified as food aggressive (5%) when compared to dogs assessed as not food aggressive at one shelter (9%); the sex of dogs was not considered with resource guarding. Possible explanations for these different patterns include the following aspects of the study design used by Mohan-Gibbons et al. [1], which differ from the present study: pit bulls and Rottweilers were excluded, the inclusion criteria focused on dogs showing highly adoptable behavior except on the food bowl test, and food aggressive dogs were in a behavior modification program while in the shelter and later in their adoptive home (although many adopters did not comply). Finally, the study by Mohan-Gibbons et al. [1] used results from the food bowl test, whereas results from the food bowl test and possession test were used here.

When resource guarding was differentiated by level of severity in the present study, guarding was a significant predictor of a dog being returned, with severe resource guarders more likely to be returned than mild to moderate guarders and dogs classified as non-guarders. Dogs showing mild to moderate guarding did not differ from dogs classified as non-guarders in their likelihood of return. The reasons for return of dogs identified in the shelter as severe guarders typically did not involve aggression to humans; instead, the reasons given were those commonly provided by adopters returning dogs to shelters (e.g., allergies, moving, personal reasons, not getting along with other pets, and behavioral problems such as elimination in the house and over-arousal; [13,24]). Importantly, despite the greater likelihood of return of severe resource guarding dogs to the shelter, almost all of these dogs (24 of 25) were eventually placed in a home. Adopter surveys have revealed that many dogs assessed as food aggressive in shelters do not guard food in their adoptive homes, and, even when dogs continue to display food guarding in the home, adopters do not consider it to be a major problem [1,3]. Taken together, the present results on adoption success and published results from adopter surveys [1,3] strongly suggest that shelter staff consider adoption rather than euthanasia for most dogs identified as resource guarders during behavioral evaluations in shelters.

Body size also influenced likelihood of return at the Tompkins shelter. Using body mass as the measure of body size, I found that large dogs were more likely to be returned than small and medium dogs. Similar results have been obtained by Marston et al. [12], Diesel et al. [14], and Posage et al. [25]; suggested explanations for the observed pattern include the greater costs, space needs, and exercise requirements of large dogs, as well as the increased challenges of managing any behavioral issues. Interestingly, in the present study, the body size by resource guarding status interaction was not significant in the analysis of factors affecting likelihood of return, indicating that returns of food aggressive dogs to the Tompkins shelter did not vary by size of dog (e.g., adopters were not more likely to return large food aggressive dogs than small food aggressive dogs).

Most dogs (87%) returned to the Tompkins shelter were subsequently re-adopted and not returned to the shelter; 9% of returned dogs were euthanized, and the remaining 4% of dogs were either transferred to a rescue organization or returned to the original owner. Lower rates of re-adoption and higher rates of euthanasia have been noted for returned dogs at other shelters. Patronek et al. [26] reported that 50% of returned dogs were subsequently adopted; these authors also found a 33% euthanasia rate for all potentially adoptable dogs at the study shelter, although this value likely represented an upper limit (the percentage of returned dogs euthanized was not described). Across three Australian shelters, Marston et al. [12] reported 57% of returned dogs were subsequently re-adopted, 38% were euthanized, and fates were unknown for the remaining 5%. I cannot definitively state that all re-adopted dogs remained in the home; I can only state that the dogs were not returned to the Tompkins shelter. However, several policies at the Tompkins shelter encourage people who do not wish to keep their adopted dog to return it to the shelter rather than give the dog to someone else. First, all adopters must sign a contract stating that they will return the dog to the Tompkins shelter if the dog is not a good fit for their household. Second, all dogs receive a microchip, which is registered before leaving the shelter, so dogs can be identified if brought elsewhere, such as to a different shelter. Finally, if a dog is returned within two weeks of adoption, then the shelter refunds 75% of the adoption fee. For these reasons, I expect that most, if not all, re-adopted dogs remained in their new homes.

#### **5. Conclusions**

The prevalence of resource guarding during behavioral evaluations was 15% in the population of dogs at the Tompkins shelter, which is comparable to that observed at other shelters in the United States [1]. The demographic profile developed for dogs identified as resource guarders at the Tompkins shelter indicated they were more likely to be adults and seniors than juveniles, and when fully grown, more likely to be either small or large than medium with respect to body size based on body mass. Spayed females, intact males, and neutered males were more likely than intact females to guard resources. Ideally, shelters should conduct behavioral evaluations of all dogs made available for adoption. However, some shelters do not follow this procedure, especially with dogs considered highly adoptable at intake [2]. The profile provided here may help such shelters make informed decisions about which dogs should be evaluated for resource guarding. For example, shelters might be less likely to assess small dogs than large dogs, but the data presented here show that small dogs are just as likely as large dogs to display food-related guarding during behavioral evaluations, and those that do are just as likely as large dogs assessed as food guarders to be returned by adopters. The ability to generalize results presented here to other shelters will depend on how similar other shelters are to the Tompkins shelter with respect to dog populations and shelter policies.

Most dogs assessed as resource guarders at the Tompkins shelter showed mild to moderate guarding. Dogs assessed as severe guarders were more likely to be returned by adopters than dogs assessed as mild to moderate guarders or non-guarders. However, almost all severe guarders that were returned to the shelter were eventually re-adopted and not returned. Thus, results from this population of shelter dogs indicate that most dogs identified as resource guarders during behavioral evaluations can be successfully re-homed, although it might take more than one effort at adoption. These data on adoption success, together with data showing that dogs assessed as food aggressive at shelters do not necessarily display food-related guarding in their adoptive homes [1,3], strongly suggest that shelter staff consider adoption rather than euthanasia for most dogs identified as resource guarders during behavioral evaluations in shelters.

**Funding:** This research received no external funding.

**Acknowledgments:** I thank Jim Bouderau, Executive Director of the Tompkins County SPCA, for permission to analyze dog records. Emme Hones, Behavior Program Manager at the shelter, provided PetPoint files and body mass data and shared her knowledge and expertise concerning behavioral evaluations and resource guarding. Samantha Rubio was instrumental in initiating this project. Willy Bemis and Kate Bemis read an earlier version of this manuscript and three anonymous reviewers also provided helpful comments.

**Conflicts of Interest:** The author declares no conflict of interest.

#### **References**


© 2019 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
