Behavioural Repertoire of Working Donkeys and Consistency of Behaviour over Time, as a Preliminary Step towards Identifying Pain-Related Behaviours

Background The donkey has a reputation for stoicism and its behavioural repertoire in clinical contexts is under-reported. Lack of understanding of the norms of donkey behaviour and how it may vary over time can compromise use of behavioural measures as indicators of pain or emotional state. The objective of this study was to find out whether the behaviour of working donkeys was influenced by gender, the time of day or differed between days with a view to assessing how robust these measures are for inclusion in a working donkey ethogram. Methodology/Principal Findings Frequency and consistency of postural and event behaviours were measured in 21 adult working donkeys (12 females; 9 males). Instantaneous (scan) and focal sampling were used to measure maintenance, lying, ingestive and investigative behaviours at hourly intervals for ten sessions on each of two consecutive days. High head carriage and biting were seen more frequently in male donkeys than females (P<0.001). Level head carriage, licking/chewing and head-shaking were observed more frequently in female donkeys (P<0.001). Tail position, ear orientation, foot stamping, rolling/lying and head-shaking behaviours were affected by time of day (P<0.001). However, only two variations in ear orientation were found to be significantly different over the two days of observations (P<0.001). Tail swishing, head shaking, foot stamping, and ears held sideways and downwards were significantly correlated (P<0.001) and are assumed to be behaviours to discourage flies. Conclusions/Significance All donkeys expressed an extensive behavioural repertoire, although some differences in behaviour were evident between genders. While most behaviours were consistent over time, some behaviours were influenced by time of day. Few behaviours differed between the two test days. The findings can be used to inform the development of a robust, evidence-based ethogram for working donkeys.


Introduction
Donkeys working in developing countries are used for transporting both goods and people, often in extreme heat and humidity, in urban working environments where there are hazards such as heavy traffic, noise, pollution and debris. Working donkeys are usually owned by very poor members of society and seldom receive adequate resources and management to maintain good welfare. During rest periods, donkeys often remain tethered or harnessed alone or in pairs. Consequently, they have little opportunity to engage in truly restorative resting behaviour or to socialise freely.
Domestic donkeys are descended from the African wild ass [1]. Observations of wild ass and populations of feral donkeys provide an insight into donkey behaviour outside of the working environment. The donkey digestive system is very similar to that of the horse, and if unimpeded donkeys will spend a comparable amount of their time grazing, typically 14-16 hours a day [2]. The social organisation of wild ass and feral donkeys can vary with their environment and the resources available [1,3,4], with some populations operating resource or territory guarding systems and others guarding females or harems [3]. Although the social lives of working donkeys have yet to be studied, the propensity for donkeys to form pair-bonds or situation -specific groupings demonstrates their social tendencies [5] and lack of opportunity to engage in social behaviour may therefore compromise their welfare further.
Behavioural observation gives the most reliable and immediate insight into the animal's perception of and interaction with its environment [6]. Behavioural divergence from the norm may be clinically significant, for example may indicate pain or pathology. It is likely that many working donkeys suffer from pain associated with multiple acute and/or chronic clinical conditions. Donkeys behave differently to horses [7], and while a number of equid ethograms exist, e.g. [8], they typically focus exclusively on horse behaviour. Little is published regarding the behavioural repertoire of the domestic donkey, particularly under working conditions. As a result, it is challenging to identify when donkey behaviour changes from the norm, compromising the use of behaviour as a welfare indicator in this species.
Research investigating consistency of behavioural parameters has examined natural behaviour and treatment effects in a range of species including horses [9], dairy cows [10], red deer [11] and lambs [12]. Consistency of behaviour can be defined as a percentage agreement between behavioural observations over time. Studying the influence of time on behaviour enables researchers to identify the behaviours most suitable for measurement during short observation periods, making them practical for use in clinical situations.
Diurnal variations in the behaviour of different groups of horses and ponies have previously been investigated, for example see [13] and [14], studies of Przewalski's horse and New Forest ponies respectively. Few peer reviewed studies could be found on diurnal patterns or consistency of measured behaviours in domestic donkeys, although detailed studies of diurnal variations in activity [15] grazing behaviour [16] and physiological measures [17] in small samples of donkeys have been published. Posture and event behaviours have been used in livestock research to assess pain [12,18], fear [19,20], distress [21], anxiety and adaptation to novel environments or stimuli [22]. The lack of information regarding the normal behaviour of donkeys hinders progress towards recognition of abnormal behaviour indicative of pain, fear or distress. Improving our understanding of what constitutes normal postural and event behaviours for donkeys will enable us to detect more readily when behaviour diverges from the norm, and facilitate the use of behaviour as an indicator of pain and emotional state in this species.
The study objective was to find out whether the posture and event behaviours recorded in working donkeys were influenced by gender or time of day and whether they differed between consecutive days. In addition, an avoidance test was used to gauge reaction towards approaching humans [23].

Ethics statement
The study was carried out in Lahore, Pakistan during November 2005, under ethical approval from the University of Bristol Faculty of Medical and Veterinary Sciences Ethics Committee (Investigation number UB/05/017) and was compliant with Pakistan law regarding ethical use of animals in science.

Donkey selection and acclimatisation to study environment
Twenty-one working donkeys were recruited from within a 15 km catchment area of Shadara fodder market, Lahore. Inclusion criteria for the study included: non-pregnant females, entire males, minimum body condition score of 1.5 (1 (very thin) -5 (very fat) [24], donkeys over 3 yrs of age but under 10 years, no acute traumatic injury or severe chronic injury, no wounds or minor body wounds only, no obvious lameness, and intact ears and nostrils. Bodyweight was recorded using an electronic weighbridge (Ezi-weigh, Tru-test Ltd, New Zealand). Donkeys were loose-housed individually in 3 m63 m shaded pen enclosures with both physical and visual contact with another donkey. They were bedded on dry straw, offered water ad libitum and a food ration twice daily of crushed grains and chopped roughage formulated to aid malnourished animals. An acclimatisation period of 18 h enabled animals to recover from any acute stress, fear, hunger or thirst resulting from their daily working lives and familiarised them with the environment prior to starting behavioural observations.

Behavioural observation period
The ethogram used in this study (see Table 1) was developed through pilot studies with working and non-working donkeys and included all possible behaviour that a donkey might display [25].
Behavioural observations were made for ten minutes at hourly intervals (excluding 12:00) for ten sessions on two consecutive days. Behaviours were recorded directly onto a check sheet by a single observer (FHR) seated 3.5 metres outside the pen. Observations consisted of: (i) instantaneous (scan) sampling at one minute intervals throughout each ten minute observation period, recording the anatomical components of posture (ear orientation, head carriage, feet position, and tail carriage), general activity category (standing, walking, lying, rolling, and eating) and the donkey's location in the pen (front or back; standing at the front of the pen would allow visual contact with other donkeys) at the scan time point. (ii) focal sampling of pre-defined event behaviours, such as snorting, braying and leg lifting, recording all occurrences throughout the 10 minute period. Some event behaviours were recorded if a single occurrence was observed (for example cough, snort, leg-lift), and others if the behaviour continued for a minimum of 3 s (such as licking/chewing, body-rubbing, walking).
The ethogram (Table 1) is organised to reflect the two types of observations made in the study.

Avoidance tests
An avoidance test was developed by adapting measures of fear in livestock [26,27]. Following pilot testing, score system [27] was modified so that the observer (FHR) approached each donkey, loose within its pen, from a distance of 3 meters by walking slowly towards the side (shoulder area) of the donkey avoiding direct eye contact. The stage at which the donkey withdrew was then scored (for description of scores see Table 2). The avoidance test was carried out once during the acclimatisation period and twice a day at 12 hour intervals (before and after behavioural observations) on the two consecutive test days.

Statistical Analysis
All data were converted into total number of observations of each posture or event behaviour per 10 minute observation period and per day for each donkey. Data were not normally distributed, so non-parametric tests were used for all analyses. Data were tested for significant gender differences using the Mann-Whitney U test. Friedman analysis of variance (ANOVA) with chi-squared statistic was used to investigate the effect of time of day on observed behaviours. The Wilcoxon Signed Rank test was used to test for differences between matched observation periods on day 1 and day 2. Kendalls coefficient of concordance was used to look for associations within groups of behaviours hypothesised to be related to each other. Statistical analysis was performed using SPSS version 19.0 (SPSS Inc., Illinois, USA) and the significance level was set at P,0.05. Holm's sequentially rejective multiple test procedure [28] was used to reduce the risk of Type I errors through the multiple tests conducted, with only the variables meeting the adjusted P value denoted by the procedure retaining their significance and being reported as such.

Behaviour
During the observation period the donkeys stood for a median (min-max) of 78.4% (48.2-94.5%) of the sampling points. Walking was observed for a median of 5.7% (1.6-13.4%) and lying for 12.3% (0-40%) with some donkeys not seen lying at all. A median of 56.6 (20.0-84.0) occurrences of eating behaviour were observed throughout the period. Maintenance behaviours, such as rolling, self-grooming and stretching were rarely observed with medians of 0.9 (0-6.8%), 3.5 (2-10 occurrences) and 0 (0-1.5 occurrences) respectively. Table 3 shows behaviours which differed significantly between male and female donkeys. Standing, walking, lying, rolling, head turning, tail position and location in pen were not influenced by gender. Some patterns were found with head carriage: female donkeys held their heads in the level position (poll level with withers) (Mann-Whitney U = 339.5; P,0.001) more frequently than male donkeys. Male donkeys more frequently held their heads in the high position (poll above withers) (Mann-Whitney U = 39.0; P,0.001), indicating a more vigilant posture. Lying and maintenance behaviours were unaffected by gender. Biting (Mann-Whitney U = 120.0; P,0.001) was more frequently demonstrated by male donkeys while head-shaking (Mann-Whitney  Behaviour: effect of time Table 4 shows the behaviours which were significantly affected by the time of day. Tail position (relaxed, tucked and swishing) and ear orientation (one ear forwards, one sideways; one sideways, one backwards; and both ears sideways and facing down) differed significantly across the day (Friedman x 2 = from 33.8 to 204.3; P,0.001). Lying down/rolling and foot stamping were also significantly affected by time of day. The majority of event behaviours were consistent over time, with the exception of headshaking which was influenced by time of day (Friedman x 2 = 91.7; P,0.001).

Behaviour: effect of gender
Only two measures were found to significantly differ across the two test days: both ears sideways (Wilcoxon 23.945; median Day 1 = 18.18, median Day 2 = 13.64; P,0.001) and both ears sideways and down (Wilcoxon 3.602; median Day 1 = 18.18, median Day 2 = 22.73; P,0.001). All other postural behaviours and event behaviours were consistent across the test days.
Using the Kendall's coefficient test, foot-stamping, tail swishing, ears held sideways facing down and headshaking were all positively associated with each other (R = 0.780, P,0.001).

Avoidance scores
Three female donkeys did not let the observer enter the pen to start the test on each test session and so scored 21 for each test. For those that did let the observer start the test, avoidance scores did not differ significantly with time of day (Wilcoxon 1242.0; median am = 3, median pm = 3; P = 0.242) or the day of testing (Wilcoxon 1714.0; median day 1 = 3, median day 2 = 3; P = 0.184). There was a significant difference in avoidance test scores between genders (Mann-Whitney U = 1162.5; median male = 3, median female = 2; P = 0.003)

Discussion
The results of this study confirm that donkeys have an extensive behavioural repertoire, as found in previous studies [25], although Table 3. Postural and event behaviours significantly more frequently observed in male (n = 9) or female (n = 12) working donkeys (bold cells denote gender where the behaviour was most frequently observed).

Behaviour
Mann-Whitney U statistic (significance, p) Male donkeys Median (min-max) Female donkeys Median (min-max) Postural behaviours:  some behaviours are observed much more frequently than others. Several behaviours varied between male and female donkeys and this should be accounted for when assessing behavioural signs of pain in working donkeys. Some behaviours of working donkeys are not consistent over time, which may make them unsuitable for use in a tool for recognising pain behaviour in clinical situations. The donkeys spent the majority of the observational period standing with very little time spent on maintenance behaviours. Gender-specific patterns for a number of postural and event behaviours were found, with the strongest differences seen in head carriage. In male donkeys high head carriage and biting were seen, suggesting a focus on the external environment. These are components of a normal social and sexual behaviour repertoire [29,30] and are indicative of the male donkeys' strong motivation to investigate surroundings and interact with con-specifics. Female donkeys appeared to be less focused on the external environment than the males, expressed by their level head carriage. Headshaking was significantly more frequent in females and may suggest that they react more vigorously than male donkeys to the presence of flies. No comparable published behavioural research could be found, therefore the results of this study can only be considered in the context of this study group and environment.
This study has also helped refine the working donkey ethogram by investigating changes in behaviours over time, enabling inclusion of the behaviours which are most robust and useable in the field at any time of day. The majority of foot and limb positions, head carriage positions and event behaviours were found to be consistent over time. Strong time-dependent effects on headshaking, foot stamping and tail swishing were hypothesised to be due to the presence of flies, which was greatest in the middle of the day, although in this study fly levels were not recorded directly. This supports the finding that free ranging Przewalski's horses increase their activity and feeding budgets at night compared to daytime levels in the summer months, due to high temperatures and disturbance by flying insects [13]. Lying down/rolling varied significantly across the eleven hour observation period. In a study of adult feral donkeys in California lying posture during daylight hours in hot months was never observed, suggesting that donkeys may change their lying behaviour to reduce the risk of heat stress [30]. The donkeys in the present study also spent a relatively short time lying down during the daylight observation periods; although they were observed informally to be recumbent during hours of darkness. Working donkeys are rarely allowed to rest un-harnessed and have little or no opportunity to lie down during a typical day, which may contribute to chronic fatigue.
Only two variants of ear orientation were found to differ significantly across test days, with the remaining postural and all event behaviours proving to be consistent. This provides strong evidence for the robustness of these behaviours over time and their inclusion in a working donkey ethogram.
Behavioural observations in the current study were limited to daylight hours; pilot studies using video-recording of behaviour at night were unsuccessful due to an erratic power supply.
Observation of behaviour over only two consecutive days was a further limitation caused by the need to return donkeys to their owners after a relatively short period of time. However, the behaviour observed would not be expected to change markedly given an extended observation period.
The avoidance scoring method generated results comparable with previous research. The present study found that 57% of donkeys achieved the top score (allowing their head to be touched by an approaching observer), in agreement with the 56% of working donkeys rated in a previous study [24] as either nonresponsive or eliciting a friendly response. The present study found that 24% of donkeys showed a negative reaction (aggression towards or avoidance of the test observer), fewer than the 44% of donkeys reported by [24] although still a cause for concern. This could be due to the difference in test environments: donkeys in the current study were loose in a pen, having rested, eaten food and received gentle handling (i.e. avoiding practices commonly seen during the donkey's working day such as shouting, slapping and hitting the donkeys with sticks) prior to testing, while the donkeys in the previous study described above [24] were scored as found during their working day. A consistent negative avoidance test score (21) was exhibited by three of the 21 donkeys in the present study and the observer could not safely enter the pen to start the test. Prolonged or persistent fear can be a serious compromise to an animal's welfare and can result in depression, chronic fatigue, restlessness and anorexia [31]. Female donkeys scored significantly lower in the test than males and the three donkeys that consistently had a negative score in the avoidance tests were all female. Gender differences in emotional reactivity have been identified in horses, with mares being more easily panicked and aggressive than geldings [32]; our findings suggest similar gender differences in donkeys.
This study has confirmed that working donkeys have a large behavioural repertoire and identified some behaviours which lack diurnal consistency in their expression. Few behaviours varied significantly across the test days indicating that time of day has a greater impact on behaviour observed than the day itself. This should be taken into consideration in future observational studies.