Personality, abnormal behaviour, and health: An evaluation of the welfare of police horses

An animal’s welfare depends on an individual’s capacity to adapt to the environment in which it lives. This adaptation is directly associated with the quality of the environment and to the possibility of expressing natural behaviours. Horses kept in stables often display a range of abnormal behaviours related to lack of control over their environment, which can lead to behavioural and health problems. An individual’s personality also plays an important role in its susceptibility and resilience to the development of diseases and abnormal behaviour; thus, an evaluation of horses’ personalities could be crucial to selecting individuals best able to cope with different work activities. This study aimed to assess the well-being of police horses maintained in a semi-confinement regime in Brazil by associating their personalities to the occurrence of abnormal behaviours and disease. Using a non-invasive approach, different tests were performed to investigate the horses’ behaviour, personality and welfare. A frustration test and a novel object test were conducted with 46 horses and the individuals’ personalities were assessed using questionnaires and behaviour tests. In addition, we evaluated their physical health through a survey of their veterinary records. The data for horses were evaluated individually. The results demonstrated the occurrence of multiple abnormal behaviours motivated by factors such as diet and lack of social contact. Moreover, the personality tests indicated this is an important component when evaluating welfare, since correlations were found between personality traits and abnormal behaviour expression, and between personality traits and health problems. According to our results, passive, stubborn, and confident horses are better suited to be selected as police horses. The ability to classify horses according to their personalities could help in selecting horses most suitable for patrolling, thereby helping to reduce behavioural problems and increasing animal well-being.


Introduction
Tests can also be applied to evaluate the behaviour of animals when experiencing frustration, which could be linked to the animal's personality. A sham-feeding test is used to evaluate satiety and is represented by the presence of mastication and saliva without food in the oral cavity) [34][35]. The presence of sham feeding, in some cases, is related to frustration due to the absence of food and can lead to abnormal behaviours [36][37][38]. In horses, food deprivation is linked to the increase of acidification in the digestive system, which causes abnormal behaviours such as crib-biting [13]. Therefore, a sham-feeding test could be used to evaluate the potential expression of abnormal behaviours in horses.
Currently in the state of Minas Gerais Brazil, horses are used by the police as the main source of labour for urban patrolling and previous studies with these animals have shown high incidence of abnormal behaviour, colic and other diseases [39][40][41] Further research conducted with police horses also demonstrated that this specific type of work could significantly affect the physical and behavioural responses of these animals [42][43].
Considering the welfare of police horses, this study aimed to evaluate the personality and wellbeing of Brazilian police horses using behavioural and physical health variables. To investigate if the horses' personalities predicted the expression of abnormal behaviours and the incidence of disease.

Ethical disclaimer
All the behavioural tests with the horses were considered non-invasive and followed the internal regulations and horse management practices of the police. The tests were accompanied by either a veterinarian or a veterinary nurse at all times. At the time this study was conducted, no ethical approval was required for non-invasive research by the University's Ethics Panel (PUC Minas) or the Police (PMMG) (S1 Fig).

Horses, housing and maintenance
For this study, 46 adult horses, 20 neutered males and 26 females used for urban patrolling, with regular working shifts, were tested. The horses belonged to the Brazilian Sport Horse breed, with a mean age of 10.37 (± 0.67 SD) years and were being used for urban patrolling for at least one year prior to the start of the study. The normal work load for the horses consisted of an 8-hour shift every other day: in cycles of 45 minutes patrolling followed by 15 minutes of rest. When not patrolling the horses were maintained in stalls and did not perform other activities.
The horses were individually housed in 2 x 2.5m masonry stalls with wooden doors and cement floors (no shavings/bedding). The animals were fed six times a day, alternating 2kg of hay and 2kg of a concentrate mix of grains, and had access to water ad libitum. The horses had weekly veterinary health check-ups.

Data collection
Data collection was divided in four different phases. All data were collected on days when the horses were not on duty.
The animals were observed for a total time of 10 minutes. Observations were made using focal sampling with instantaneous recordings of the behaviour every 10 seconds [46] and data collection was based on an ethogram for horse behaviour developed from two previous studies [50][51] (Table 2).
Health assessment-veterinary records. The last step in evaluating the welfare of these horses, was a survey of their veterinary records and any pathology that could be associated with behavioural problems (abnormal behaviours) [41] was recorded.

Statistical analyses
Sham-feeding test. The normality of the data was tested through an Anderson-Darling test. The data were classified as non-parametric and Generalized Linear Mixed Models (GLMMs) were built to verify if abnormal behaviours (dependent variables) were affected by the three phases of the sham test or by the individual horses (explanatory variables). All the tests were conducted in Minitab1 18.1 [52] with 95% confidence interval (S1 Table).
Horses personaltity. To evaluate the composition of the horses' personalities, different analyses were conducted combining the data from the personality's questionnaires, the results of the novel object test and the results of each phase of the sham-feeding test.
According to other studies, the great variation in the assessment of an individual's personality through a questionnaire relies on different evaluators [49]. Hence, Kendall's coefficient of concordance (W) [53] was calculated to verify raters' agreement and reliability. This coefficient varies from 0 to 1 and a score closer to one means a higher association [53]. The coefficient was calculated for each horse and each one of the adjectives among the four raters. When the Table 1. Personality traits used in the questionnaires applied to police cavalry officers and veterinarians to evaluate horses' personalities.

Behavioural description Active
Constantly moving, is not observed standing still for long.

Aggressive
Displays signs of aggression towards humans and/or horses, can cause harm to another individual.

Confident
Behaves in an assured manner, does not easily hesitate.

Curious
Explores new situations without hesitation.

Equable
Easy to handle, remains calm around other horses and/or people, behaves gently.

Insecure
Hesitant when alone, is reassured by the presence of others, seems more confident among others.

Irritable
Does not tolerate disturbances, responds negatively if provoked.

Opportunistic
Takes advantage of situations as they arise.

Playful
Initiates and/or takes part in play when requested.

Passive
Behaves in a relaxed manner, not easily disturbed, is slower than other horses when handled.

Sociable
Seeks company of others, behaves in a positive and appropriate way around other horses and/or people.

Stubborn
Does not cooperate easily, takes time when performing tasks, does not easily give in.

Intelligent
Learns new things easily, is faster than others in responding to mental tasks.

Solitary
Prefers to be alone when a group interaction is possible.

Hardworking
Responds promptly and appropriately when given tasks, remains focused on what is required to do.

Fearful
Is startled easily, does not react well to new situations, tries to escape from disturbances.

Reliable
Can be trusted to perform tasks, is consider a safe horse to be around.

Cooperative
Horse is easily conducted, does not demonstrate resistance. Kendall's coefficient of concordance for a horse trait's was non-significant (p>0.05), the data were analysed using a Spearman's rank correlation to identify if any of two evaluators had agreed on their scores. All the adjectives that failed to find any significant correlation among at least any two raters were considered unreliable and discarded from further analysis [54]. After the removal of the non-significant personality traits, the Kendall's coefficient was recalculated and the final (W) value was considered as the concordance rate among the raters. Following this, a mean of the reliable adjectives was calculated for each animal using the scores marked by the raters. The obtained values were then analysed in a Principal Component Analysis (PCA) with a Varimax rotation. The extracted components were determined using the eigenvalue criteria (the value must be greater than one) and the percentage of contribution of each one of the adjectives to the components was verified. Additionally, during the PCA, the scores of the components for each horse were calculated using the adjectives' loadings.
To investigate if the personality traits had an association with the recorded diseases (colic and lameness), Generalized Linear Models (GLMs) were built to verify if the occurrence of diseases (dependent variables) were affected by the horses' personalities (explanatory variables).

Behaviour Ã Description
Standing active Animal is standing still, but shows activity, such as hoof or head movements.

Standing resting
Animal is standing but shows rest positions. E.g. One of the hind legs flexed, ears pointing down and to the side or, lower lip relaxed (dropping), eyes closed or partially closed.

Alert
Animal is standing still with the, head standing high. Ears are erect and pointing forward. The nostrils may or may not be dilated.

Locomotion
Animal moves using its limbs. E.g. straight line or in circles.

Alarm movements
Animal displays an alert posture with members stretched and rigid, open nostrils, ears pointing in the direction of the stimulus. Can accompany abrupt movements of escape attempt, like small jumps and/or alarm vocalizations (Winch).
Interacting with object Any interaction performed by the animal with the object presented in the test E.g.: sniff, lick, bite, etc.

Observing object
Animal carefully observes the object with ears up and pointing to the direction of the object

Interacting with the basket
Any interaction by the animal related to the basket but not with the object E.g.: sniff, lick, bite, etc.

Flehmen
Animal stands still, raises the neck and tilts the head back with the ears upside down, while everting the upper lip, making evident of the incisor teeth and the upper gum. Behaviour associated with olfactory exploration.

Exploration
Olfactory and tactile exploration of the environment. E.g. Horse searches for food on the ground or in different locations of the stall, other than the feeder. Usually touches the environment using the upper and lower lip.

Abnormal behaviour
This category includes any behaviour that derives from a normal pattern for the animal and repetitive abnormal behaviours. E.g.: air swallowing, weaving, pacing, lip smacking, tongue playing, lip-twisting, crib-biting, stomping, pawing, head-shaking.

Social interaction
Interactions with other horses through physical contact. Eg: touch of snouts, mutual grooming, etc.

Agonistic behaviour
Aggressive social interaction or aggressive displays towards another horse and/or person. E.g. bites.

Vocalization
Characteristic sounds emitted by the animal with short or long duration.

Eat
Animal ingests food portions, terminated with the end of mastication and swallowing. It also includes activities such as licking salt blocks.

Drink
Animal sips and swallow water.

Maintenance
Urination, defecation and grooming related behaviours, such as scratching, licking the fur, etc.

Other
Any other behavior displayed by the animal that does not fit in any other categories. https://doi.org/10.1371/journal.pone.0202750.t002 Personality, abnormal behaviour, and health: An evaluation of the welfare of police horses Also, Spearman's rank correlations were calculated among the results to verify if any personality trait varied with the occurrence of the diseases. These analyses were run using both the mean values obtained from the questionnaires and the residual loadings from the PCA with a quadratic transformation (trait measures). Furthermore, the mean frequency of the observed behaviours in the novel object test and during the three phases of sham-feeding test were compared against the adjective loadings for each component obtained in the PCA analysis for each horse. Both PCA analysis and the Spearman's rank correlation tests were carried out in SPSS1 20.0 [55] with 95% confidence interval. The GLM was carried out using Minitab1 18.1 [52] with 95% confidence interval (S2 Table).
Veterinary records. Data regarding colic and lameness cases whose cause was unknown by the veterinary staff were selected from the horses' veterinary records-the records covered any occurrence in the period between the horse arrival to the unit up to six months prior to the study. An individual's health rate was created using the number of colic cases and number of lameness cases divided by the total number of years each horse was in the police. Additionally, a mean rate was calculated for all animals for both pathologies. Spearman's rank correlations were calculated between each individual's health rate (colic and lameness) and the mean frequency of abnormal behaviours observed for each horse during the three phases of the sham-feeding tests.

Sham-feeding tests
During the three phases of the sham-feeding test, it was possible to identify 10 distinctive abnormal behaviours expressed by the horses: Head-shaking, Lip-smacking, Tongue playing, Licking, Crib-biting, Kicking, Pawing, Stomp, Lip-twisting and Weaving. These abnormal behaviours accounted for 16.75% of the behaviours observed during the sham-feeding test, which corresponded to >15% of the total test time. However, the expression of abnormal behaviours was not affected by any phase of the sham-feeding test (P > 0.05), but it was affected by the individual variation ( Table 3).
The horse that exhibited most abnormal behaviours during the tests expressed these behaviours during 14:14:04 minutes (59.31% ± 8.98 SD) of the total test time (18 minutes). The most expressed abnormal behaviour by this individual was lip-twisting (26.59% ± 5.79 SD). When

Personality tests
Twelve of the 18 personality traits showed positive concordances among the raters (W = 0.394, N = 46, p < 0.001). Six personality traits were excluded from further analysis because they did not reach concordance with at least two raters (confident, opportunistic, passive, sociable, fearful and solitary). Four components were extracted from the Principal Component Analysis with the remaining horses' personality traits. These four components explained 77.52% of the total variance and presented eigenvalues greater than one. The traits aggressive (+), confident (-), irritable (+), cooperative (-), equable (-) and stubborn (+) presented their highest loadings in the first component and were associated with the perception of the raters about the horses behaviour during husbandry/patrolling (Fig 1A; Table 4).
The traits curious (+), intelligent (+), playful (+) and active (+) presented their highest loadings in the second component and were associated with the horses usual responses to their environment. Finally, the traits hardworking (+) and insecure (+) presented their highest loadings in the third and fourth components, respectively. The third component related to the horses' obedience to the police commands and the fourth component related to the horses' insecurity in its behavioural responses (Fig 1B; Table 4).
Eight horses presented different personalities when compared to others (individuals 33, 168, 137, 188, 181, 221, 968 and 184); they did not group with the others in the PCA analysis ( Fig 1B). In addition, these were horses that presented the lowest number of veterinary problems and presented the lowest number of abnormal behaviours.
For the novel object test, the sum of abnormal behaviours and a few abnormal behaviours individually (oral abnormal behaviours: lip-smacking, tongue playing, lip-twisting, licking and crib-biting; locomotors abnormal behaviours: head-shaking, pawing, kicking, stomp and weaving) showed correlations with different personality traits. The locomotors behaviours were weakly positively correlated to curious (r s = 0.285, N = 46 p = 0.05) and showed a medium positive correlation with intelligent (r s = 0.400, N = 46, p = 0.002). The oral abnormal behaviours were weakly positively correlated to the cooperative personality trait (r s = 0.285, N = 46, p = 0.05) and the sum of abnormal behaviours showed a medium-strength positive correlation with intelligent (r s = 0.439, N = 46, p = 0.006).
Latency to react to the presence of the novel object was 59.1s (± 19.2 SD) and to interact with the novel object was 120s (±15.1 SD). No correlations were found between the type of novel object presented and the latencies to react and to interact with the objects (p > 0.05).
Nine behaviours expressed during the novel object test were correlated to eight personality traits. The personality traits active and playful correlated positively with three behaviours, irritable correlated positively and negatively with two behaviours, and curious, insecure and equable correlated positively or negatively to only one behaviour (Table 5).

Veterinary records
Only 12 horses out of the 46 individuals did not have any veterinary records of disease (i.e., had never required treatment), 156 veterinary records were analysed and 131 colic cases and 25 lameness cases were found. The maximum number of cases of colic for a single horse was 30; the maximum number of cases of lameness for a single animal was nine. The mean rate of colic per animal was 2.84 (+0.437) and the population rate was 0.29 cases. For lameness occurrences, the mean rate was 0.54 (+0.126) per animal and the population rate was 0.03 cases per animal per year.
Negative weak correlations were found between the abnormal behaviours and the rate of lameness of the horses, both for all recorded abnormal behaviours (r s = -0.189, N = 46, p = 0.027) and for grouped locomotor abnormal behaviours (r s = -0.17, N = 46, p = 0.046), but no correlations were found between the occurrence of lameness and the expression of each behaviour individually.
Personality traits explained the occurrence of lameness in the horses, when analysing both the questionnaires and the PCA loadings. Curious, playful and intelligent horses presented more lameness than the other horses, both in the questionnaire analysis and in the behavioural analysis. The personality traits active, cooperative, irritable, reliable and equable influenced the occurrence of lameness in the horses when only the ratings from the questionnaires were taken into account, and the personality trait aggressive influenced the occurrence of lameness in the horses when only the behavioural measures were taken into account (Table 6).

Discussion
The horses exhibited abnormal behaviours in rates varying from 5 to 15% during the different tests and many of these behaviours were associated with personality traits. The most curious, cooperative and intelligent horses exhibited more abnormal behaviours than the more passive and stubborn ones. Aggressive, insecure, irritable and hardworking horses also presented more abnormal behaviours than the horses with confident and reliable traits. In general, more intelligent, playful and curious horses presented more lameness than horses with other personality traits.
During the sham-feeding tests, the expression of abnormal behaviour was always associated with the presence of food, with horses increasing the expression of abnormal behaviours, between the three phases of the test. Some researchers performed restriction tests and found an increase in crib-biting in horses after the withdrawal of the food reward offered to them [11,[56][57]. Similarly, in another study, an increase in the expression of weaving was observed in horses that received more food items during the day than the controls [58]. The occurrence of abnormal behaviours due to food restriction has been observed in other species, such as pigs [37][38], piglets [59] and cows [60]. For horses, it is suggested that the development of abnormal behaviours responses to food are linked to anxiety and anticipation [61].  Personality, abnormal behaviour, and health: An evaluation of the welfare of police horses Although the incidence of abnormal behaviours was high during the sham feeding tests, more than 16% in only 6 minutes, it is known that these behaviours were not developed in response to the sessions, only stimulated. This means that the conditions in which the horses live are a contributing factor for the development of such behaviours [11,14,[62][63].
Police horses are trained to cope with conflict (e.g., crowds, loud noises, etc.). In the horses studied, this training lasts one year and it is conducted with foals at the age of two. In a natural environment, this is usually the age where the foals are developing their social bonds as they leave their maternal groups [64][65]. Considering these extreme aversive stimuli, the training schedules and the social isolation during training, behavioural problems in these horses are induced; it has been suggested that the suppression of emotions by horses while training contributes to a decreased welfare [20]. These animals are evolutionarily programmed to escape situations like these, not to confront them [65]. Hence, this type of routine could induce the occurrence of abnormal behaviours, especially in foals [20]. The fact that the most curious, cooperative and intelligent horses exhibited more abnormal behaviours reinforces the idea that the demanded obedience during training and daily patrols could cause welfare problems to these individuals. Particularly, in our subjects, along with the non-ideal environment, personality appears to be a key factor determining how the animals' responded to their environment.
For the studied horses, the first Principal Component loaded traits were more related to extrinsic factors, how the rider perceived the horses' behaviours, while the second component loaded traits more related to intrinsic factors or how the animals' naturally behave in the environment (not the perception of the riders and/or during duty). In the first component, the horses were more aggressive, irritable, confident, cooperative, equable and stubborn, and in the second component, the horses were more curious, intelligent, playful, and active. Similar results were found in other studies [47,[66][67]. These results point to a variation in the composition of the horses' personalities, affected by both extrinsic and intrinsic factors. This variation can be due to a high exposure to various stimuli throughout the horses' lives, such as number of riders, number of owners, type of work and husbandry [68][69][70][71]. Age, breed and genetic factors may also be responsible for differences in horses' personalities [15][16]70].
In our study, the expression of abnormal behaviours was also correlated with personality traits. In general, intelligent, cooperative, curious, equable and playful horses were more prone to express oral (crib-biting and lip-twisting) and locomotor (head-shaking, kicking and pawing) abnormal behaviours than others. Personality, abnormal behaviour, and health: An evaluation of the welfare of police horses The reason why intelligent and curious horses tend to develop more abnormal behaviours is yet not clear, but it was suggested that these horses could learn from other individuals that display abnormal behaviours [72], similar occurrences have been found in other species [73][74]. In the same manner, hardworking and cooperative horses are normally more obedient, and these characteristics are linked to the commands received during training, husbandry and work making them more responsive to the environment and more susceptible to develop abnormal behaviours [11]. In fact, there were many anecdotes by the riders in our study, when completing the questionnaires, about the expression of abnormal behaviours by the horses during training sessions.
Lastly, aggressive and irritable horses are known to develop repetitive abnormal behaviours because they are not able to deal with the frustration of a restricted environment, such as stabling [72,[75][76]. In the present study, aggressive and irritable horses showed less expression of licking, lip-smacking and stomp. Maybe these horses did not possess higher levels of anxiety that could culminate in an abnormal response, thereby dealing better with stress [72].
The behaviours expressed by the horses to the novel objects indicated that the objects functioned as environmental enrichment items. They were alert when novel objects were available and interacted with them, which could be interpreted as curiosity towards the objects or as fear. Increased vigilance behaviours during novel object tests were observed in other studies [51,[77][78]. However, due to the short latency time taken to interact with the objects and to the great amount of time spent interacting with the novel stimulus, we interpret their response as curiosity.
In our study, active, equable and curious horses explored more their stalls during the test and ate more, were less alarmed in the presence of novel objects, while horses that were more insecure interacted less with the objects. Corroborating our results, a study showed more insecure horses avoided novel objects [79]. Contrary to this, if a horse does not consider the novel object as a threat, the chance of interaction with the object or of expressing calm behaviours, such as feeding, increases [80]. Active horses showed the propensity to express more abnormal behaviours, whereas irritable and playful horses showed the propensity to express less abnormal behaviours. Usually, active horses maintained in small, barren stalls can redirect their motivation to express abnormal behaviours [81].
The associations found in our novel object tests did not have a strong correlation with the personality assessment. This may be due to the fact that an individual's behaviour is affected by more than one personality trait and varies greatly between individuals [66][67]. To test this fully would require a much larger sample size.
Finally, when comparing the horses' health and clinical records with their individual expression of abnormal behaviour, we were able to identify that the expression of locomotor and abnormal behaviours (grouped) presented a weak negative correlation with the incidence of lameness. Colic cases were weakly negatively correlated with the expression of licking and crib-biting, but positively correlated with lip twisting and tongue playing. Previous studies have shown that the occurrence of colic in horses is directly associated with crib-biting and air swallowing [22][23]. Usually, the horses expressed lip twisting and tongue playing after feeding, however, crib-biting was not recurrent. Why this behaviour did not show any correlations with colic occurrence remains unclear.
The occurrence of lameness but not colic was influenced by certain aspects of the horses' personalities, and this influence was observed both with the questionnaire and behavioural data. Curious, playful and intelligent horses were more prone to develop lameness. The differences in the effects of personality traits on the occurrence of lameness found in the questionnaire and in the behavioural data showed that it is difficult to rely only on one type of personality assessment [82][83].
It is important to state that the results reflected the mean for the entire horse population and only from colic cases with unknown causes. In previous studies conducted within the same horse population, the number of colic cases was above the mean found in the literature [39][40][41]. Since the correlations between lameness and personality traits in the present study were low, results should be treated with caution and further studies with large sample sizes should be conducted.

Conclusion
In this study, the most intelligent, curious and playful horses were the most susceptible horses to express abnormal behaviours and develop lameness. According to our results, passive, stubborn and confident horses were better able to cope with the demands of being a police horse. The classifying of horses according to their personalities could help in choosing the most suitable individuals for patrolling, thereby increasing animal welfare; however, this will need validating through long-term studies.