Laterality as a Predictor of Coping Strategies in Dogs Entering a Rescue Shelter

: It has been reported that during the ﬁrst few days following entry to a kennel environment, shelter dogs may suffer poor welfare. Previous work suggests that motor bias (the preferred use of one limb over the other) can potentially be used as an indicator of emotional reactivity and welfare risk. In this study, we investigate whether paw preference could be used as a predictive indicator of stress coping (measured using cortisol levels and behavioural observation) in a sample of 41 dogs entering a rescue shelter. Cortisol levels and behavioural observations were collected for one week after admission. We scored the dogs’ paw preference during a food-retrieval task. Our results showed that increasing left-pawedness was associated with a higher expression of stress-related behaviours such as frequent change of state, vocalisations and lower body posture. These results are in keeping with previous ﬁndings showing that left-limb biased animals are more vulnerable to stress. Paw preference testing may be a useful tool for detecting different coping strategies in dogs entering a kennel environment and identifying target individuals at risk of reduced welfare.


Introduction
Entering a rescue shelter can be a very stressful experience for dogs; they are separated from any social attachment figures, they are exposed to a novel environment (i.e., unfamiliar noise, smells, disruption of familiar routine) and to daily interactions with unfamiliar people and conspecifics [1][2][3]. Previous studies show how this transition can generate a state of fear, anxiety, and frustration [4,5]. Social and spatial restriction following confinement can be a cause of both acute and chronic stress [6][7][8].
Physiological studies have confirmed that kennelling is perceived by dogs as a psychogenic stressor, with animals displaying peaks in cortisol levels in the first few days after arrival [3,5,9]. Individual dogs, however, may have different coping abilities or stress resilience. Previous research has highlighted two main coping styles in individuals that are environmentally challenged: proactive and reactive [10]. Proactive coping styles may be more typical of 'bold' personality types and are defined by active attempts to counter the stressful stimuli and by low Hypothalamic-Pituitary-Adrenal (HPA)-axis reactivity. Reactive, or passive, coping may be more typical of 'shy' personality types, and involves higher activation of the HPA-axis system, immobility, and low levels of aggression [11,12]. Hiby and colleagues [13] found that dogs showing or not showing physiological adaptation to kennelling had different behavioural styles, that is, dogs with lower HPA-axis activation spent more of their time walking or trotting compared to dogs with high cortisol levels. Effective interventions to minimise signs of poor welfare should be based on the evaluation of dogs' individual ability to cope and adapt to confinement in kennels. Behavioural indicators offer excellent validated measures of welfare as they represent the output of a range of sensory and cognitive experiences and decision-making processes, reflecting the expression of the animal's underlying emotional and physiological state [14].

Animals
Data collection was performed at the Dogs Trust Rehoming Centre in Ballymena, Co. Antrim, UK, over a period of 9 months. All dogs entering the shelter during this period were enrolled in the study; exceptions were made if the dog was pregnant, seriously ill or injured, impossible to handle or walk on the lead due to excessive fear or aggression. A total of 41 dogs were assessed, 22 males (54.5% neutered) and 19 females (58.0% spayed), including a number of different purebreds and crossbreeds. The minimum age for a subject to be enrolled was 12 months; the oldest dog in our sample was 9 years (median = 3 years; mean ± SD 3.7 ± 2.5 years) ( Table 1). Charlie Collie  M  1  yes  32  Tess  PatterdaleX  F  1  yes  33  Zoe  JackRuss  F  2  no  34  Tia  Shihtzu  F  8  no  35  Gizmo  Maltese  M  6  no  36  Lily  Shorky  F  2.5  no  37  Benson  CollieX  M  1.5  yes  38  Honey  Yorkie  F  7  no  39  Coco  TerrierX  F  5  no  40  Enzo  JackRuss  M  2.5  yes  41 Aria JackRuss F 2 yes A portion of dogs (14.6%, unknown history), were rescued from the municipal pound or other shelters and the remainder were surrendered to the Dogs Trust by their owners for a variety of reasons. Main reasons included family health issues (e.g., allergies or illness of a family member), work commitments (i.e., not being able to take care of the dog anymore), owner's death, or moving home as most common (Table 2).

Procedure
On the day of admission to the shelter (Day 0), dogs were enrolled and general information collected from shelter records (sex, age, provenience, date of entrance, etc.). Data collection started on the following morning (Day 1) for one week, with sampling occurring on days 2, 3, 5 and 7. On each of the sampling days, both urine samples and behavioural recordings were collected following the protocols described below. Paw preference was assessed once on Day 3, see later (Section 2.2.3). The day of paw preference assessment was decided during pilot testing: it was noticed that dogs were not interested in the toy during their first two days in the kennels, whereas on day three most dogs would actively interact with, and retrieve food from, the ball.

Urine Cortisol/Creatinine Ratio
Between 08:00 h and 10:00 h of sampling days (1, 2, 3, 5, and 7), dogs were leashed and walked outdoors; a mid-stream sample of naturally voided urine was collected using urine sampling kits (Rocket ® URIPET™, Washington, WA, USA). Urine tubes were labelled and stored at −30 • C for up to 3 months. Samples were shipped in batches to IDEXX laboratories (West Yorkshire, UK) and tested for cortisol/creatinine (C/C) ratio. Cortisol was extracted with chemiluminescent competitive immunoassay using the Siemens Immulite 2000, whereas creatinine with Jaffe (alkaline picrate) reaction using the Beckman AU 5800 analyser.

Behavioural Observations
Behavioural observations started after urine sampling (Days 1, 2, 3, 5, and 7). A digital camera on a tripod was positioned in front of the kennel and recorded the behaviour of each dog over three sessions of 35 min (30 central minutes extracted for analysis). Members of staff prioritised cleaning the kennels and feeding the dogs that were going to be recorded on that day. This allowed no interruptions during recording. Routine activities, however, continued as normal in the adjacent kennels. Observational sessions were distributed as follows: Session 1 (OB1) between 09:00 h and 10:00 h, during morning activities. Session 2 (OB2) between 11:00 h and 12:00 h during staff tea break; Session 3 (OB3) between 13:00 h and 15:00 during visiting hours (the Centre opened to the public between 12:00 and 16:00 h). The sessions were planned this way to allow for an overview of the range of behaviours that dogs performed across a typical day.

Paw Preference Assessment
Dogs' paw preferences were assessed on Day 3 using the 'Kong ball test', one of the most commonly used measures of canine motor bias [33,35,36]. Following previously published protocols, each dog was allowed to sniff a Kong Ball TM (KONG Company, Golden, CO, USA; a hollow conical-shaped toy) filled with dog food. Then the Kong was placed inside the kennel on the floor centrally in front of the animal. The paw used (left, right) by the dog to hold/stabilise the toy was recorded using a smartphone app purposely developed for this study. The test continued until 100 paw uses (left or right) had been made. The use of both paws was recorded but not counted towards this total. The use of the app had several advantages, including not having to interrupt eye contact with the subject to write down the scores, recording would automatically stop when the 100 paw-use target was reached, main statistics for the dog appeared immediately on the screen, and raw data were readily available for download in Microsoft Excel format. A single paw use was recorded regardless of how long the paw stayed on the ball. The animal was required to remove its paw completely from the ball for paw use to be scored as a separate response. Dogs were tested for approximately 30 min (the average length of time taken to collect 100 data points).

Analysis
All analyses were carried out using IBM SPSS Statistics 21.0 (IBM, Armonk, NY, USA).

Paw Preference Assessment
Individual laterality scores were calculated using a binomial test and converted to a z-score using the formula z = (L − 0.5N)/ √ (0.25N), L being the number of left paw uses and N the total of left and right paw uses. A z-score ≥ 1.96 indicates a left paw bias, a z-score ≤ −1.96 indicates a right paw bias; a value between these two scores indicates no lateral bias (ambilateral) [33,36]. A chi-squared test was used to calculate departures from random distribution of left-, right-and ambilateral paw preference groups. Paw-preferent (either to the left or right) vs. ambilateral, and right-vs. left-paw preferent animals were compared using binomial tests to assess any significant group difference. Chi-squared tests were also used to assess whether the pawedness classification was associated with the dogs' sex (male, female) or castration status (neutered, intact).
A directional laterality index (LI) was calculated to quantify each dog's paw preference on a continuum from strongly left-paw preferent (+1) to strongly right-paw-preferent (−1). The LI score was calculated as (L − R)/(L + R), where R represents the number of right paws and L the number of left paws used [30]. A score of 0 indicates no bias, a score of ±1 indicates that the subject used the same paw throughout the trial. In addition to the directional bias of lateral behaviour (i.e., left or right bias), the strength of laterality has also been used as a proxy measure of hemispheric brain activity. Strongly lateralised animals show a greater activity of one hemisphere (irrespective of the side), while weakly lateralised animals do not show a significant dominance of one hemisphere over the other (i.e., ambilateral) [25]. The absolute value of LI (LI_ABS) gives a measure of the strength of laterality, irrespective of the direction of paw use. A Shapiro-Wilk normality test was used to assess the distribution of LI and LI_ABS values to identify any population bias. Any effect of sex or castration status on the direction and strength of laterality was calculated using a Mann-Whitney-U test for independent samples.

Behavioural Analysis
A total of 307.5 h of footage was analysed. Videos were scored using the behavioural recording software The Observer XT13 (Noldus, The Netherlands). Due to time and resource constraints we used an instantaneous sampling method. However, given that some behaviours (e.g., barking) might have been lost or underestimated using this approach, we also scored a subset of videos using continuous sampling (Table 3). Instantaneous sampling was used every 60 s: the observer recorded the behaviour expressed by the focal animal at a given instant and scoring was performed for all the video footage (i.e., 90 data points per observational day, 450 data points for the 5 days). Continuous sampling was performed during the first 15 min of recoding of session OB1 on each of the 5 sampling days. Table 3. Behavioural variables recorded during the behavioural observations either using instantaneous (i) or continuous (c) sampling techniques. During the latter, behaviours could be scored as either duration (d) or frequency (f).

Inactive behaviours
Stand: the dog is still, standing on fours (i) Sit: the dog is sitting on hind legs

Predictive Factors Affecting Behavioural Variations
To investigate if any variation in behaviour could be associated with the observation time, days from shelter entrance, sex, castration status, C/C ratio or laterality (either direction or strength), we performed Linear Mixed Model (LMM) analysis with each behaviour as the dependent variable and sex, castration status, observation times, day, C/C, LI and LI_ABS as predictive factors and dog identity as random.

Laterality as a Predictive Factor of C/C Level Variation
To investigate if any variation in cortisol levels throughout the observation time could be predicted by laterality (either direction or strength), we performed LMM with C/C ratio as the dependent variable and day of observation as the repeated measure, LI and LI_ABS as predictive variables, and dog identity as a random factor.

Paw Preference Assessment
When calculating lateralisation at the individual level, 8 (19.5%) dogs mainly used their left paw to hold the Kong, 18 (43.9%) mainly used their right paw and 14 (34.1%) used both paws equally. Binomial tests showed no significant difference between the distribution of lateralised and ambilateral dogs (p = 0.81) or between the number of left and right pawed dogs (p = 0.80). Paw preference was not successfully recorded for one dog (2.5%).
The distribution of the three pawedness classification groups did not differ significantly from that expected by chance, that is, no population level effect (χ

Behavioural Data Management
Given that the different sampling methods (instantaneous and continuous) were performed for different lengths of time, analyses were kept separate using two different datasets. From an initial inspection, a few behaviours were performed only rarely or by a small number of animals. Only behaviours with a median >0 were included in the analysis. Behaviours excluded were: play, socialising, repetitive behaviours, stress-related behaviours, and all behaviours under 'other'. Walking and trotting were analysed as one variable termed 'moving', while barking, howling/whining, and growling were analysed as one variable termed 'vocalisations'.

Predictive Factors Affecting Behavioural Variations
The dogs' behaviour did not vary greatly over their first week in the rescue shelter (Table 4), with the exception of 'posture', with dogs showing an increase in higher posture over time (D1 = 2.7 ± 5.4; D3 = 4.2 ± 6.9; D7 = 5.6 ± 6.9). The time of observation had a significant effect on the expression of behaviours including standing, resting/sleeping, and maintaining a low posture (Table 4). Descriptive analysis showed that dogs spent less time standing during the second bout of observation t (OB2) (i.e., staff break, 8.1 ± 6.9) compared to OB1 (i.e., cleaning, 12.3 ± 8.1) and OB3 (i.e., visiting hours, 12.0 ± 7.8). On the contrary, dogs spent more time resting/sleeping during the second observation bout (OB2) (7.4 ± 8.4) compared to OB1 (4.4 ± 7.2) and OB3 (4.2 ± 6.1). Finally, dogs were likely to show less low posture during the second observation bout (OB2) (4.1 ± 5.6) compared to OB1 (6.4 ± 7.4) and OB3 (6.4 ± 7.4). These outcomes are rather intuitive as dogs were undisturbed, thus mainly resting during OB2 and scored as having neutral posture. Variations in cortisol levels were associated with variation in behaviour expression (Table 4). Dogs with increasingly higher cortisol levels were more likely to spend time inactive (sitting and lying down) but less time moving, standing, and showing a high posture (Table 4).
LMM analysis highlighted a significant effect of both LI and ABS_LI on a number of behaviours (Table 4). Increasing left pawedness was associated with more frequent change of state, more time spent vocalising (Figure 1a), sitting (Figure 1b), and showing a lower posture (Figure 1c). By contrast, dogs with increasingly strong right-paw bias were likely to spend more time resting/sleeping, lying down ( Figure 1d) and moving.  Variations in cortisol levels were associated with variation in behaviour expression (Table 4). Dogs with increasingly higher cortisol levels were more likely to spend time inactive (sitting and lying down) but less time moving, standing, and showing a high posture (Table 4).
LMM analysis highlighted a significant effect of both LI and ABS_LI on a number of behaviours (Table 4). Increasing left pawedness was associated with more frequent change of state, more time spent vocalising (Figure 1a), sitting (Figure 1b), and showing a lower posture (Figure 1c). By contrast, dogs with increasingly strong right-paw bias were likely to spend more time resting/sleeping, lying down ( Figure 1d) and moving.

Laterality as a Predictive Factor of C/C Level Variation
LMM showed no significant effect of laterality (direction or strength) on the levels of C/C ratio in dogs (p > 0.05).

Discussion
This study provides the first evidence for the potential use of paw preference as an indicator of coping strategy in dogs entering a kennel environment.
At the individual level, results from the Kong tests revealed a significant paw preference bias (either left or right, 63.4%), but no significant population bias emerged. This supports some previous research in this area [23,35,37]. However, results seem to depend greatly on the population of animals under study, with other authors having reported population biases in dogs [36,38]. Neither the direction nor strength of the dogs' paw preferences differed significantly according to canine sex or castration status. Again, the literature is divided on this matter and results appear to be populationdependant [23,32,33,[35][36][37].
Our results showed a relationship between the dogs' paw preferences and certain behaviours. Right-pawedness was associated with a frequent expression of both active and inactive behaviours such as moving or resting/sleeping and lying down. Increasing left-pawedness, by contrast, was associated with a higher expression of stress-related behaviours, including frequent change of state, vocalisations, sitting, and low posture. These results tie in nicely with previous findings of a right hemisphere dominance (left motor bias) in animals being more vulnerable to stress [20]. For example, when stressed by being caged and placed in an unfamiliar environment, cats with higher cortisol levels also had a higher activation of the right brain hemisphere [39].
In a review paper, Rogers [20] suggested that the left hemisphere has a dominant control of proactive/calm behaviours (e.g., approach, exploration) and the right hemisphere of reactive behaviours (e.g., fear, aggression). In marmosets, for example, studies showed that left-handed animals (right hemispheric dominance) were more fearful of alarm calls [40], more likely to show higher cortisol levels [41], and were less reactive to the effects of social facilitation on capturing prey [42] than right-handed marmosets (left hemispheric dominance). Thus, it could be hypothesised that animals with a left-limb bias are less likely to exploit new resources and are more likely to express negative emotional functioning compared to individuals with a right-limb bias [32]. Recent studies

Laterality as a Predictive Factor of C/C Level Variation
LMM showed no significant effect of laterality (direction or strength) on the levels of C/C ratio in dogs (p > 0.05).

Discussion
This study provides the first evidence for the potential use of paw preference as an indicator of coping strategy in dogs entering a kennel environment.
At the individual level, results from the Kong tests revealed a significant paw preference bias (either left or right, 63.4%), but no significant population bias emerged. This supports some previous research in this area [23,35,37]. However, results seem to depend greatly on the population of animals under study, with other authors having reported population biases in dogs [36,38]. Neither the direction nor strength of the dogs' paw preferences differed significantly according to canine sex or castration status. Again, the literature is divided on this matter and results appear to be population-dependant [23,32,33,[35][36][37].
Our results showed a relationship between the dogs' paw preferences and certain behaviours. Right-pawedness was associated with a frequent expression of both active and inactive behaviours such as moving or resting/sleeping and lying down. Increasing left-pawedness, by contrast, was associated with a higher expression of stress-related behaviours, including frequent change of state, vocalisations, sitting, and low posture. These results tie in nicely with previous findings of a right hemisphere dominance (left motor bias) in animals being more vulnerable to stress [20]. For example, when stressed by being caged and placed in an unfamiliar environment, cats with higher cortisol levels also had a higher activation of the right brain hemisphere [39].
In a review paper, Rogers [20] suggested that the left hemisphere has a dominant control of proactive/calm behaviours (e.g., approach, exploration) and the right hemisphere of reactive behaviours (e.g., fear, aggression). In marmosets, for example, studies showed that left-handed animals (right hemispheric dominance) were more fearful of alarm calls [40], more likely to show higher cortisol levels [41], and were less reactive to the effects of social facilitation on capturing prey [42] than right-handed marmosets (left hemispheric dominance). Thus, it could be hypothesised that animals with a left-limb bias are less likely to exploit new resources and are more likely to express negative emotional functioning compared to individuals with a right-limb bias [32]. Recent studies on marmosets [43] and dogs [44] confirmed this link by observing that a stronger left-limb motor lateralisation was associated with a more negative or 'pessimistic' cognitive bias.
Dogs may show different behavioural styles according to their physiological adaptation. In this study, we found that lower cortisol levels were associated with more frequent observations of moving, standing, and high posture. It could be argued that dogs showing these behaviours were better at coping with the kennel environment, showing a more confident posture and being more active. Hiby and colleagues [13] found that on days when dogs were more active, cortisol levels were lower compared to days when dogs were spending more time sitting or lying down. High activity levels do not necessarily indicate lower stress; there is evidence of increased activity following social and spatial restrictions [7]. It could be the case that staying active helps dogs to cope better with confinement, hence the lower cortisol levels [13]. Interestingly, our results showed that a stronger right-paw bias was linked to higher activity. This lends support for the previously suggested association between left-hemispheric bias and exploratory/proactive coping styles [20].
A high cortisol profile was associated with increased observations of sitting and lying down. Once more these results are in line with findings by Hiby et al. [13]. Both behaviours indicate a more vigilant posture (compared to e.g., resting/sleeping) and a more reactive posture. Following the proactive/reactive hypothesis, dogs showing more sitting and lying behaviour are expected to be more strongly left-pawed. This was only partially true, as we found left-biased dogs did indeed spend more of their time sitting (Figure 1), but less time lying down and resting/sleeping. The link between stress, physiology, and behaviour is complex and still debated [3,5,45]. Multiple factors may play a role in modulating the results, including individual variability and past experiences [7,13,45]. Here, we added a new indicator of emotional state (as assessed by motor bias) that could help triangulate our results. If, based on previous work, we consider right-paw biased animals to be less susceptible to environmental stress, we could suggest that animals that either are very active or, on the contrary, very quiet, might cope better overall with novel environmental challenges. Left-pawed dogs, typically reacting to kennelling by showing anxiety and stress-related behaviours such as vocalising, low posture, sitting, and frequently changing state might, on the contrary, find adapting to confinement more challenging.
Strength of laterality (rather than the direction of an animal's motor bias) has been previously associated with increased levels of stress behaviour. In dogs, for example, weaker motor laterality has been linked to higher reactivity and fear of thunderstorms and firework sounds when compared to stronger lateralisation [36]. Our results show that weaker lateralisation was associated with a higher occurrence of standing behaviour and maintenance of a low and high posture. Both high and low posture may be a sign of strong emotional arousal, with different valence (e.g., aggression/extraversion versus fear respectively). Previous work has found that weak limb preference was in fact associated with more fearful, as well as excitable, reactions [31,36]. Having a strongly lateralised brain, by contrast, seems to be advantageous for enhanced cognitive abilities and higher survival fitness [46]. In our study, resting/sleeping and lying down was linked to a higher degree of lateralisation. As mentioned for the right-paw bias, it could be that spending most of the time sleeping helps an individual to cope better with the stress of entering a kennel environment. It is worth mentioning, however, that spending large amounts of time sleeping has also been reported in sheltered dogs as an indicator of learned helplessness and depression-like state [4] and would be more typically associated with a reactive coping style [10].
We found no association between cortisol levels and laterality. The right hemisphere has been found to have a dominant control of endocrine function, especially of the HPA axis [20]. Being associated with more fearful behaviour, left-limb bias should, in theory, be associated with higher corticoid responses than right-limb bias, as found in rhesus macaques and rats [47,48]. In dogs, Siniscalchi and collaborators [49] found a chronic elevation of hair cortisol in those individuals showing a higher reactivity to acoustic stimuli with different emotional valence. However, Batt and colleagues [31] found no correlation between behaviour, laterality, and salivary cortisol in dogs. This is something that should perhaps be explored further.
Overall, it appears that a left-motor bias may be linked to a more negative affective state, a more reactive coping style, and a more challenging adaptation to novel environments. Assessing paw preference may become a useful tool to detect different coping strategies in dogs entering a kennel and reduce stress in target individuals at higher welfare risk. Further work is needed to explore this in greater depth.