1
Applied Animal Behaviour Science
2
In press
3
Final Revision – NOT EDITED by the journal
4
5
Relationships between behaviour and health in
6
working horses, donkeys, and mules in developing
7
countries
8 Charlotte C. Burn1*, Tania L. Dennison2 & Helen R. Whay3
9 10 11
1
Veterinary Clinical Sciences, The Royal Veterinary College, UK; 2Brooke Hospital
12
for Animals, London, UK; 3Clinical Veterinary Sciences, University of Bristol, UK
13 14
Short title: Behaviour and health of working equids
15 16 17 18 19 20 21
*Corresponding author: Dr C. C. Burn, Centre for Animal Welfare, Department of
22
Clinical Veterinary Sciences, The Royal Veterinary College, Hawkshead Lane,
23
Hertfordshire, AL9 7TA, UK 1
24 25
Abstract Recent studies raise serious welfare concerns regarding the estimated 93.6
26
million horses, donkeys and mules in developing countries. Most equids are used for
27
work in poor communities, and are commonly afflicted with wounds, poor body
28
condition, respiratory diseases, parasites, dental problems, and lameness. Non-
29
physical welfare problems, such as fear of humans, are also of concern. Interventions
30
to improve working equine welfare aim to prioritise the conditions that cause the most
31
severe impositions on the animals’ subjectively experienced welfare, but data
32
identifying which conditions these may be, are lacking. Here we describe a stage in
33
the validation of behavioural welfare indicators that form part of a working equine
34
welfare assessment protocol. Over four years, behavioural and physical data were
35
collected from 5481 donkeys, 4504 horses, and 858 mules across nine developing
36
countries. Behaviours included the animals’ general alertness, and their responses to
37
four human-interaction tests, using the unfamiliar observer as the human stimulus.
38
Avoidance behaviours correlated significantly with each other across the human-
39
interaction tests, with 21% of animals avoiding the observer, but they showed no
40
associations with likely anthropogenic injuries. Over 13% of equids appeared
41
‘apathetic’: lethargic rather than alert. Measures of unresponsiveness correlated with
42
each other across the five tests, and were associated with poor body condition,
43
abnormal mucous membrane colour, faecal soiling, eye abnormalities, more severe
44
wounds, and older age, depending on the equine species. This suggests that working
45
equids in poor physical health show an unresponsive behavioural profile, consistent
46
with sickness behaviour, exhaustion, chronic pain, or depression-like states.
47 48
Keywords: Animal welfare; Developing countries; Equine; Human-animal
49
relationships; Inactivity; Sickness behaviour
2
50 51
Introduction An estimated 39 million donkeys, 40.5 million horses and 12.3 million mules
52
live in developing countries, constituting over 85% of the world’s equids (FAOSTAT,
53
2006). In developing countries, equids are mostly used as working animals, often
54
carrying out tasks under harsh and impoverished conditions for long hours each day.
55
Consequently, previous research has shown that they have many physical and clinical
56
problems, such as wounds, poor body condition, respiratory diseases, high parasite
57
burdens, dental problems, and lameness (de Aluja, 1998; Pritchard et al., 2005;
58
Tesfaye & Curran, 2005; Regan, 2009a; Burden et al., in press; Burn et al., in press;
59
Saul et al., in press). These problems are likely to reduce the work efficiency of the
60
animals, indirectly reducing the income of the often very poor people who rely on
61
them. The physical/clinical problems are also likely to cause poor welfare for the
62
animals themselves, since similar conditions in humans are associated with pain,
63
weakness, exhaustion, and depression (Kelley et al., 2003). Hence, charities, such as
64
the Brooke Hospital for Animals (‘the Brooke’) who provided data for the current
65
study, work to attempt to improve working equine health and welfare.
66
To date, despite growing information on working equine health, little is known
67
about the animal welfare implications (in the sense of the animals’ subjective
68
experiences) of the myriad physical conditions these animals accumulate. For
69
example, some physical conditions might be associated with pain (Broster et al., 2009;
70
Regan, 2009b) or exhaustion (the high creatine kinase concentrations in working
71
equids indicates that muscle damage from overwork is prevalent: Tadich et al., 1997;
72
Pritchard et al., 2009), while others may barely be perceived by the animals even if
73
they harm their health and longevity. This is of applied importance because funding
74
and resources to improve animal welfare should ideally be targeted towards problems
75
likely to cause the most suffering as perceived by the animals themselves, which may
76
be those causing deviations from normal behaviour. It would also be of value to have 3
77
easily observed behavioural welfare indicators, so that the individual animals likely to
78
be in most need of welfare improvement can be identified rapidly before a full
79
assessment is done. Here we describe a stage in the validation of behavioural welfare
80
indicators in working equids to assess their potential for providing information
81
towards these purposes.
82
The behavioural indicators here formed part of a non-invasive welfare
83
assessment that was developed in 2003 by the University of Bristol in collaboration
84
with the Brooke. It should be noted that the data were not collected for the purposes of
85
this study, but instead for the internal monitoring purposes of the Brooke. The data
86
thus take the form of a series of standardised surveys of working equine populations,
87
and we use an epidemiological approach, exploiting the natural variation within these
88
populations, to investigate relationships between variables of interest. The welfare
89
assessment protocol was intended to be brief and appropriate for field conditions,
90
where observers were often interrupting the animals’ work. The welfare indicators
91
were chosen to be simple, to minimise assessment time and to facilitate repeatability
92
between observers. The behaviours included an observation of general alertness
93
versus unresponsiveness to the environment. Unresponsiveness can be a component of
94
chronic pain (Ashley et al., 2005), sickness behaviour (Aubert, 1999; Millman, 2007;
95
Weary et al., 2009), depression (Vollmayr & Henn, 2003; Dunn et al., 2005) and
96
exhaustion; but equally it can be associated with neutral or even good welfare if an
97
animal perceives its situation to be secure enough to allow reduced vigilance (Paul et
98
al., 2005). The human-interaction tests, which were proposed as measures of fear,
99
aversion, or friendliness towards humans, incorporated brief assessments of the
100
animals’ avoidance and other responses when a human approached, and their
101
acceptance of human contact (e.g. for farm animals de Passille & Rushen, 2005;
102
Waiblinger et al., 2006, and for sports and companion horses Hausberger et al., 2008).
103
Pain behaviours were not explicitly included in the assessment, since they are highly
4
104
diverse, differing with the source, nature, and time-scale of the pain (Ashley et al.,
105
2005; Regan, 2009a).
106
Like many large-scale, multi-centre epidemiological studies, data were
107
collected by a number of trained observers (e.g. Waters et al., 2002; Dawkins et al.,
108
2004; Rutherford et al., 2009). The observers in the current study were only those
109
attaining ≥80% agreement with the trainer for all indicators, but the more general
110
inter- and intra-observer reliability of the physical and behavioural indicators in the
111
welfare assessment have also been tested (Burn et al., 2009): the population was too
112
homogenous to conclusively test all the indicators, but while alertness showed
113
acceptable intra-observer reliability, it showed poor reliability between observers.
114
This suggests that some observers have different thresholds or use different cues for
115
deciding when equids are alert or not, despite having been trained using the same
116
guidance notes and photographs. These differences will have added unsystematic
117
noise to the data analysed here, but the variable was included because so little is
118
currently known about working equine behaviour, and – with modification – a
119
measure of general alertness could be important in future equine welfare assessments.
120
The responses to the four human-interaction tests did show acceptable reliability
121
between observers. The only physical indicator in the current study that attained poor
122
reliability was mucous membrane colour, but it was included here because of its
123
clinical relevance as a field test for endotoxaemia or gastrointestinal illness (Hailat et
124
al., 1997; Thoefner et al., 2001; Hillyer, 2004).
125
In the current study, we investigated how the behaviours related to each other,
126
and how they related to potentially relevant physical measures. Our hypotheses were
127
broadly that (a) unresponsiveness should correlate across the tests and (b)
128
unresponsiveness should be associated with an overall reduced prevalence of physical
129
problems if it signifies good welfare, or with an increased prevalence of physical
130
problems if it indicates negative welfare (i.e. we tested whether responsiveness in 5
131
these animals had a relationship in either direction with physical welfare). Also, if
132
aversion to humans is consistent within individual animals, animals that show
133
aversion to humans in one of the tests should (c) show aversion to humans in the three
134
other human-interaction tests, and (d) have more physical signs of anthropogenic
135
harm e.g. wounds on the hindquarters, potentially indicating beating.
136
Methods
137
Animals and observers
138
Non-invasive behavioural and physical data were collected from 5481
139
donkeys, 4504 horses, and 858 mules across 60 locations in nine developing
140
countries: Afghanistan, Egypt, Ethiopia, Guatemala, India, Jordan, Kenya, Pakistan
141
and The Gambia. The data were collected over a 4-year period (March 2003-
142
February 2007). The locations were areas where the Brooke was working or
143
considering working in the future, or where a collaborating organisation was working
144
(see Acknowledgements for collaborating organisations). Details of the locations and
145
dates are reported by Burn and colleagues (in press).
146
Welfare assessments were carried out by 42 trained observers, the vast
147
majority of whom were veterinarians, but who were otherwise animal behaviour
148
scientists, agricultural experts, or social scientists. All observers underwent a 6-10 day
149
training course (described in Burn et al., 2009), the length being determined by the
150
previous experience and English-language skills of participants; the course was based
151
around a detailed, 102-page, photographic guide (available upon request from the
152
authors: Pritchard & Whay, 2003 (unpublished), 2004 (unpublished)), which the
153
participants kept. Only those participants who attained ≥80% agreement with the
154
trainer on all measures were selected for data collection in this study. A welfare
155
assessment trainer assessed about 5-30% of the animals in each location to help
156
ensure consistency between the locations (the percentage of animals varied due to the 6
157
logistical constraints of visiting developing countries, gaining owner permission to
158
assess each animal, and a paucity of data available on equine populations at the
159
planning stages of the study). Animals were selected in locations that the observers
160
had experience of working in, and between 5% and 100% of animals were assessed in
161
each location depending on the size of the population, with it being possible to
162
examine a higher proportion of animals in smaller populations within a single visit.
163
Until 2004, observers were requested to sample systematically from known animal
164
populations, e.g. every third donkey in the market or on the roadside; thereafter,
165
observers were instructed to sample randomly by writing ‘encounter numbers’ (e.g.
166
1st, 2nd, 3rd etc animal encountered) or owner names on pieces of paper, which were
167
folded, mixed, and then blindly selected until the required sample size had been drawn
168
for that day.
169
Behaviour and health measurement
170
The data were collected using a welfare assessment protocol that included 41
171
animal-based measures potentially relevant to welfare (Pritchard & Whay, 2003
172
(unpublished), 2004 (unpublished)); but only a subset of the measurements were
173
analysed in the current study and they are summarised and briefly described in Table
174
1. The welfare indicators included physical health measures, as well as the animals’
175
general alertness (versus the opposite, labelled ‘apathy’ but while acknowledging that
176
this is not necessarily the associated emotional state; Table 1) and their behavioural
177
responses to human presence and human contact. The welfare assessment protocol
178
was modified slightly for some of the measurements from 2004 onwards, as described
179
in detail by Burn and colleagues (2009).
180 181
Broadly, the sequence and procedure for taking the measurements, including those described in Table 1, was as follows:
7
182
1. Without disturbing the animal (and where possible, before asking the owner’s
183
permission to observe the animal), the observer assessed the animal’s general
184
alertness from a distance of at least 3 m away, and for up to 10 s.
185
2. After gaining permission from the owner, the observer approached the animal
186
from 3 m away at a normal pace, looking at the animal’s neck or breast. The
187
observer approached at an angle of about 20o, rather than from directly ahead of
188
the animal, and then stopped 30 cm from the animal’s head and recorded the
189
animal’s response at the moment they stopped.
190
3. The observer walked alongside the animal towards its rear and back again,
191
maintaining a distance of about 30 cm from its body, and recorded any signs of
192
attention to them and whether donkeys showed a tail-tuck response (tail-tucking
193
was not observed in horses or mules).
194
4. The observer gently placed their hand under the animal’s chin, contacting it
195
enough to take some weight but not so as to lift the head. If the animal moved its
196
head away from the hand, the observer would not pursue it. This was the first
197
point of physical contact between the observer and the animal, unless the animal
198
itself had already initiated contact.
199
5. The observer then recorded physical health, starting with observing the lips and
200
head, and any eye abnormalities. They inserted a thumb or finger into the corner
201
of the animal’s mouth and lifted the top lip until the gums and teeth were easily
202
visible. The observer stood back to assign a body condition score (BCS), before
203
walking to the rear to record faecal soiling. The observer tested skin tent duration
204
on the animal’s neck and observed any behavioural signs of heat stress (discussed
205
elsewhere in Pritchard et al., 2006, 2007; Pritchard et al., 2008). Ectoparasites and
8
206
lesions were assessed across the body and limbs. Visible swellings of the flexor
207
tendon or fetlock joints were recorded, followed by hoof health, including picking
208
up the right fore-foot to examine the sole surface. Finally, gait was assessed by
209
watching the animal walk as the owner led it for approximately six paces in a
210
straight line away from, and then back towards, the observer; if the animal was
211
hitched to a cart, time-constraints required that its gait was assessed while pulling
212
the cart, once the cart had gained momentum (i.e. not while the animal was
213
starting from stationary).
214
Statistical analyses
215
Data were analysed using generalised linear mixed models for binary
216
outcomes (glmmML and glmmPQL, R, version 2.58). The three species were
217
analysed separately, and the precise sample sizes depended on how many animals fell
218
into each of the categories comprising the predictors in each model. In all models the
219
Location nested within Country was included as a random effect (this took into
220
account the location effects themselves as well as systematic ‘noise’ between
221
locations: changes over the 4-year period, the two assessment versions, and the
222
different observers). Predictors always included characteristics of the animals
223
themselves (species, age, sex, and BCS) to control known variation. They also
224
included the animal’s main work-type, but in some cases this caused multicollinearity
225
(identified from inflated standard errors in the models). When this occurred, an
226
urbanisation score was included instead of work-type (since it predicted the work type
227
to some extent); urbanisation and work-type could not be included together in the
228
models. Model fit was assessed using the deviance and Akaike's information criterion.
229
The models were selected to test: (1) interrelationships between the
230
behaviours, (2) relationships between general alertness/ responsiveness to human
231
interaction and measures of physical health, and (3) relationships between human 9
232
avoidance behaviours and physical signs of anthropogenic harm (firing lesions, slit
233
nostrils, and lesions on the hindquarters and tail were chosen here as proxies to reflect
234
anthropogenic harm).
235
The number of tests carried out was limited to those that were biologically
236
relevant according to the scientific literature and working equine experts. Even so,
237
because 59 tests were carried out for each species, the false discovery rate was
238
controlled for (Benjamini & Hochberg, 1995), which determined that the appropriate
239
significance level was P ≤ 0.016. In addition, because this is an exploratory study,
240
intended to suggest relationships worthy of future research rather than only to test
241
existing hypotheses (Bender & Lange, 2001), P-values between 0.016 and the
242
traditional significance level of 0.05 are reported as ‘trends’; these will require
243
confirmation through further research before they can be treated as significant.
244
Results
245
Correlations between behaviours
246
The prevalences of each behaviour for the three species are given in Table 2,
247
showing that over 13% of equids appeared apathetic or depressed (the two categories
248
are combined in the current study because only 0.6% of the animals (70 individuals)
249
appeared depressed). Measures of unresponsiveness correlated positively with each
250
other across the behavioural tests (Table 3). For example, in all three species, apathy
251
correlated with a lack of response to the observer walking beside the animal, and
252
unresponsiveness to the observer approach correlated with unresponsiveness both to
253
the observer walking beside the animal and to chin-contact. Also, in horses and
254
donkeys, apathy was associated with unresponsiveness to the observer approach. It
255
should be noted that because all these results are correlational, the effect direction can
256
be stated either way around, i.e. alert individuals were more responsive in the human-
257
interaction tests. There were a total of 300 highly unresponsive individuals who 10
258
appeared apathetic and responded to none of the human-interaction tests, equating to
259
2.8% of all the animals (2.5% of donkeys, 3.1% of horses, and 2.9% of mules).
260
Behaviours that were proposed to test aversion to humans also correlated
261
positively with each other across the tests (Table 3). In particular, avoidance of chin-
262
contact was associated, in all three species, with avoidance of observer approach, and
263
in donkeys and horses, with an aggressive response to observer approach. Tail-tucking
264
in donkeys was similarly associated with avoidance of the observer and avoidance of
265
chin-contact. Avoidance of the observer approach was seen in about 26% of donkeys,
266
14% of horses, and 23% of mules (Table 2). Across all species, 6.6% of animals either
267
avoided or showed aggression to the observer and avoided chin-contact, and 2.6% of
268
donkeys showed both those responses as well as tail-tucking.
269
Even the species-specific associations showed similar patterns to the above,
270
indicating either unresponsiveness or aversion to humans (Table 3).
271
Correlations between behaviours and physical welfare problems
272
Apathy was associated with a number of indicators of poor health, including –
273
in all three species – a low BCS and abnormal mucous membrane colour (and
274
possibly more numerous and severe skin lesions, but this was only a trend in horses
275
and mules) (Table 4). The odds ratios show that the effect sizes could be large, such
276
as a one-point decrease in BCS approximately doubling the chance that an animal
277
would appear apathetic. In general, apathy and unresponsiveness to the observer were
278
associated with lower BCSs, older age, and other health problems that depended on
279
the species. It is also worth specifically noting that limb and foot problems (swollen
280
tendons and/or joints in donkeys, and abnormal gaits in horses) reduced the chance
281
that animals would show an avoidance of the observer approach. Conversely,
282
proactive behaviours – including alertness, human avoidance behaviours, and
11
283
aggressive or friendly responses to the observer – were usually associated with
284
measures such as higher BCSs and younger ages.
285
Exceptions to this pattern were that abnormal soles and abnormal hoof shapes
286
were associated with alertness. There were also non-significant trends indicating that
287
older horses might be more likely to avoid chin-contact than younger ones were; and
288
that faecal soiling in donkeys might be associated with a friendly response to the
289
observer.
290
We found no significant relationships or trends between responses to humans
291
and anthropogenic injuries, between avoidance of chin-contact and lesions on the lips/
292
head, or between tail-tuck behaviour and lesions on the hindquarters/ tail.
293
Discussion
294
The results suggest that the behavioural tests incorporated into this welfare
295
assessment have potential as welfare indicators for identifying individual animals with
296
negative welfare, in an unresponsive state, or that consistently show behavioural
297
aversion to unfamiliar humans.
298
Unresponsiveness as a working equine welfare indicator
299
It appears that equids with more severe and numerous physical problems enter
300
a state of behavioural unresponsiveness, consistent across the five measurements
301
taken here. The association with physical problems makes it highly unlikely that the
302
behavioural unresponsiveness we have observed in the context of working equids
303
reflects neutral or positive welfare, such as resting or relaxed states. Instead, it implies
304
that the animals’ resources are being stretched to their limits and their fitness is
305
compromised; the animals – of prey species – are conserving ‘energy’, possibly even
306
at the risk of not responding appropriately to potentially threatening stimuli. In terms
307
of the sentient experiences of these animals, this behaviour could indicate any of 12
308
several different negative welfare states, such as malaise (Kelley et al., 2003;
309
Millman, 2007), exhaustion from overwork (Tadich et al., 1997; Pritchard et al.,
310
2009), chronic pain (Ashley et al., 2005), apathy, or depression (Dunn et al., 2005),
311
which may differ depending on the physical problem in question.
312
Most of these states can be associated with sickness behaviour, which is
313
underlain by a ubiquitous, non-specific proinflammatory cytokine response to a range
314
of problems (Hart, 1988; Kelley et al., 2003; Dantzer & Kelley, 2007). The generality
315
of the response means that it is useful for identifying vulnerable animals initially, but
316
that further diagnostic tests are required to elucidate the specific problem in each case.
317
Sickness behaviour is adaptive in that it allows the body’s resources to be diverted
318
towards the immune system, but the costs include reduced vigilance and a lack of
319
maintenance behaviours, such as eating and grooming: as Hart (1988) states, “The
320
sick individual is viewed as being at a life or death juncture and its behavior is an all-
321
out effort to overcome the disease”. In farm animals, recommendations suggest that
322
individuals exhibiting sickness behaviour should be rested in a quiet area to aid their
323
recovery (e.g. Millman, 2007), but working equids in developing countries can rarely
324
be rested, because of the extreme dependency of their owners on the daily income and
325
subsistence provided by their animals.
326
Approximately 2.8% of equids appeared apathetic and failed to respond to any
327
of the human-interaction tests, even when the observer made physical contact with
328
their chins. While a relatively small percentage of animals, this may signify a severe
329
welfare problem because of the association we have found between unresponsiveness
330
and many physical problems, and it suggests that if we scale up from the 10,843
331
equids here to the 93.6 million equids in developing countries as a whole,
332
approximately 2.6 million working equids may be in this unresponsive state
333
worldwide.
13
334
The unresponsive measure that was associated with the most physical
335
problems in the three species was an apathetic general attitude, in which state
336
approximately 13% of the equids were classified. This attained acceptable reliability
337
within observers, but poor reliability between observers (Burn et al., 2009)
338
(remembering that the observers in the current study were only those who attained at
339
least 80% agreement with the trainer). This suggests that some observers used
340
different thresholds or differing cues for ascribing the more ambiguous animals to one
341
category of alertness or another. However, the biologically plausible associations
342
between apathy, unresponsiveness in the other tests, and physical problems, suggest
343
that the alertness/apathy distinction describes an underlying biological construct that
344
was strong enough to be observed here despite this. Alertness/apathy in working
345
equids could therefore constitute a sensitive, if non-specific, marker of underlying
346
problems that would otherwise only be identified on closer inspection; another such
347
indicator in these animals is BCS, which correlates with many other diverse physical
348
problems (Burn et al., in press).
349
Nevertheless, assessment of alertness/apathy should be refined, particularly for
350
studies where this brief, broad-brush assessment protocol could be replaced with a
351
more focussed and in-depth assessment. The behavioural assessment could be
352
improved by breaking alertness down into its observable components (e.g. ear
353
movement and position, eye closure, head position, and foot and tail movement, etc),
354
or by adding a greater number of subjective ‘whole animal’ descriptors (e.g. rigid
355
posture, restless, relaxed, asleep/resting or eating, etc). A refined system could not
356
only improve inter-observer reliability, but would also allow greater discrimination
357
between different negative, unresponsive welfare states; for example, a rigid stance,
358
inattentive to the external environment, is often associated with chronic pain (Ashley
359
et al., 2005), and thus might be distinguishable from the more ‘slumped’ stance
360
associated with depression/exhaustion. Similarly, the associations observed here
361
between foot abnormalities and alertness in donkeys, might be attributed to foot pain 14
362
causing weight-shifting behaviour (Ashley et al., 2005) (foot movement constitutes
363
part of the ‘alert’ descriptor here), rather than donkeys with abnormal feet being alert
364
per se. Other suggestions of pain in the animals observed here were the associations
365
between reduced observer avoidance and presence of limb swellings/ gait
366
abnormalities in donkeys and horses, respectively.
367
Human-equine interactions
368
Aversion to the observer was suggested through the significant associations
369
between avoidance or aggression towards the observer approach, a tail-tuck response
370
in donkeys, and avoidance of chin-contact. The trend associating a friendly response
371
to the observer with acceptance of chin-contact in donkeys lends some weight towards
372
this being more than just an active/ passive distinction between responses to the
373
observer. The acceptable inter-observer reliability of these tests further suggests that
374
individual animals are consistent in how they respond to different unfamiliar strangers
375
(Burn et al., 2009).
376
These tests did not actually correlate with any likely anthropogenic
377
pathologies, so we have no evidence to suggest that aversion to the observer indicates
378
fear of being physically injured by them. Of course, a lack of significance does not
379
mean that no relationship exists, but with the large total number of animals sampled
380
here, we should have had sufficient power to at least detect strong associations within
381
the current dataset. More sensitive and diverse measurements might reveal more
382
subtle relationships, as the measurements here were not recorded for the purposes of
383
the current study, so they are quite limited. Another limitation was that the response to
384
the (unfamiliar) observer may not reflect the response to the animals’ regular
385
handlers, but the animal attendants would not have been able to carry out the human-
386
interaction tests in a standardised manner; hence trained observers were used as the
387
stimuli instead.
15
388
To confirm whether the aversion to the observer assessed here reflects fear or
389
aversion to humans, rather than, say, aversion to any novel stimuli, more focused
390
experimental studies will be required. Well validated indicators of fear should
391
correlate with the proposed fear or aversion behaviours. For example, opportunistic
392
observations of owner/ user behaviours that are known to cause poor welfare and/or
393
aversions to humans in farm animals, such as slapping, beating or shouting (Rushen et
394
al., 1999; Hemsworth, 2003), could be informative. Such observations might be
395
expected to show that animals that are regularly slapped, beaten or shouted at by their
396
users would generally respond with more behaviours or physiological responses that
397
indicate fear or aversion to humans (possibly including learned helplessness in some
398
extreme cases of chronic stress: Vollmayr & Henn, 2003).
399
Nevertheless, the correlations between tests again suggest that the behaviours
400
are rooted in a biological construct that is consistent within individual equids, and
401
which generalises to different unfamiliar humans. If this represents fear of humans or
402
of novel stimuli generally, it is of concern because some of the behaviours were fairly
403
prevalent: for example, approximately a quarter of donkeys and mules attempted to
404
avoid the observer approach. Clearly this is a topic that merits further research, and
405
interventions may be necessary to facilitate an improvement in the relationships
406
between working equids and their human handlers.
407 408 409
Physical problems that were most strongly associated with behavioural effects A final aspect of welfare significance that these results bring to light, is the
410
relative importance of particular physical problems in terms of the behavioural
411
associations seen here. The physical problems that correlated most strongly with
412
unresponsiveness can be summarised as low BCS, abnormal mucous membrane
413
colour, and to a lesser extent, numerous and severe skin lesions, where the association
16
414
with apathy and at least one other measure of unresponsiveness was observed across
415
all three species. As mentioned previously, lower BCS correlates with many other
416
physical problems in working equids, which is of concern since over 29% of working
417
equids were scored as ‘very thin’ (≤1.5 on a scale of 1 – 5: very thin – very fat) (Burn
418
et al., in press). The causes of low BCS are multifactorial, and likely to include
419
malnutrition, overwork, parasitism, and disease, which could simultaneously cause
420
behavioural unresponsiveness. Conversely, behavioural unresponsiveness may
421
include a reduced appetite, as in sickness (Hart, 1988; Dantzer & Kelley, 2007; Weary
422
et al., 2009) and depression (Vollmayr & Henn, 2003), which would further reduce
423
BCS. Skin lesions in these animals are also known to be severe and numerous, with
424
over 25% of equids having at least one moderate or deep lesion, and 62% having 1 –
425
13 lesions larger than 2x2 cm (Burn et al., in press). The lesions could cause apathy
426
through pain and/or infection, or conversely an apathetic animal may be more prone
427
to lesions if it stumbles or collides with objects. It is well known that thinner equids
428
have more lesions (e.g. Pritchard et al., 2005; Burn et al., in press), but this correlation
429
does not explain the association between apathy and lesion severity, because the
430
association remained significant even when BCS was included in the model.
431
The relationship between apathy and abnormal mucous membrane colour was
432
more unexpected, because mucous membrane colour is subjective, and it attained poor
433
inter-observer reliability (Burn et al., 2009). Nevertheless, in the current study, the
434
relationship was significant in all three species, and abnormal mucous membrane
435
colour does appear to give a clinically useful indication of anaemia, including
436
illnesses, such as colic (e.g. Thoefner et al., 2001), diarrhoea (Hillyer, 2004), or
437
babesiosis (Hailat et al., 1997). The current study therefore suggests that illnesses like
438
these are associated with an apathetic behavioural state in working equids, consistent
439
with sickness induced lethargy and/or pain (Aubert, 1999; Millman, 2007; Weary et
440
al., 2009). In previous haematological studies, even apparently healthy equids
441
working in Pakistan, had lower concentrations of haemoglobin, erythrocytes and 17
442
haematocrit than those seen in horses in developed countries, indicating that sub-
443
clinical anaemia is prevalent in these working equids (Gul et al., 2007; Pritchard et al.,
444
2009).
445
The observer reliability of mucous membrane colour assessment requires
446
improvement by including more categories than simply ‘normal’ or ‘abnormal’. For
447
example, it could include standardised colour descriptors as in the study by Thoefner
448
and colleagues (2001), as well as an ordinal score indicating whether the abnormality
449
was mild, moderate or pronounced.
450
Other physical problems that were significantly associated with apathy or
451
unresponsiveness to the human-interaction tests in more than one species were faecal
452
soiling (indicative of diarrhoea) and older age (Table 4). Eye abnormalities also more
453
than doubled the chance of a horse appearing apathetic; eye closure formed part of the
454
classification of an animal as apathetic, but this result may suggest that many eye
455
abnormalities could be painful, in horses at least. It should be noted that other
456
physical problems not reaching significance in the current study might also have
457
detrimental effects on equine welfare, such as acute, localised pain, or general
458
anxiety, which we did not attempt to measure here.
459
The only positive welfare indicator in the current study was a friendly
460
response to the observer approach. This shows promise, as it was significantly
461
associated with a younger age and better BCS. More ambiguously, however, there
462
was a non-significant trend towards a higher likelihood of faecal soiling, indicating
463
diarrhoea, in friendlier donkeys. This may be a Type I error, but is worth mentioning
464
speculatively because some owners or tourists may hand-feed the animals certain
465
inappropriate foods, making the animals more friendly but also inadvertently giving
466
them diarrhoea. Other positive welfare indicators could help identify additional subtle
467
effects of the physical conditions on welfare. Few positive welfare indicators have yet
468
been validated in animal welfare science (Boissy et al., 2007; Yeates & Main, 2008), 18
469
but it would be worth incorporating such measures into a field assessment protocol
470
once that becomes practicable in the future.
471
Summary
472
In summary, behavioural unresponsiveness correlated across the behavioural
473
measures in the current study, and was associated with more numerous or severe
474
physical conditions. The strongest associations were found between an apathetic
475
general attitude and lower BCS, abnormal mucous membrane colour, older age, and
476
(in horses) eye abnormalities; apathy/alertness may therefore provide a useful – if
477
non-specific –indicator for rapidly identifying animals most in need of welfare
478
interventions. Behaviours that were proposed to indicate aversion to humans
479
correlated across the four human-interaction tests, but showed no significant
480
relationships with physical problems that might suggest injury by humans. Further
481
research will be necessary to investigate fear of humans in these animals, since 21%
482
of equids avoided the observer. Many of the measurements taken here should be
483
refined for future, more in-depth welfare assessments, especially alertness and
484
mucous membrane colour, which showed poor inter-observer agreement.
485
Nevertheless, our results suggest that the behaviours included in this welfare
486
assessment show promise as working equine welfare indicators, and deepen our
487
understanding of the relationships between behaviour and physical health in these
488
animals. The associations found here are correlational, so experimental research will
489
be required to understand their causal relationships, and to develop effective
490
interventions to improve working equine welfare.
491 492 493
Acknowledgements This study was supported and funded by the Brooke Hospital for Animals.
494
Many thanks to all the owners who kindly permitted their animals to be used in this
495
study, and to the observers who collected the data. We are grateful for Dr Joy C. 19
496
Pritchard’s constructive comments on the manuscript. We would also like to thank the
497
Kenya Network for Dissemination of Agricultural Technologies, Equinos Sano para
498
El Pueblo (ESAP, Guatemala), Gambia Horse and Donkey Trust, and the Aga Khan
499
Rural support programme (Chitral, Pakistan) for their collaboration; the Afghanistan
500
data were collected in association with the Committee for Rehabilitation of Aid to
501
Afghanistan; also, in 2003 Help in Suffering and The Blue Cross assisted with the
502
Jaipur and the Hyderabad welfare assessments in India, respectively.
503
References
504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535
Ashley, F.H., Waterman-Pearson, A.E., Whay, H.R., 2005. Behavioural assessment of pain in horses and donkeys: application to clinical practice and future studies. Equine Vet. J. 37, 565-575. Aubert, A., 1999. Sickness and behaviour in animals: a motivational perspective. Neurosci. Biobehav. Rev. 23, 1029-1036. Bender, R., Lange, S., 2001. Adjusting for multiple testing--when and how? J. Clin. Epidemiol. 54, 343-349. Benjamini, Y., Hochberg, Y., 1995. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. Roy. Stat. Soc. B. 57, 289-300. Boissy, A., Manteuffel, G., Jensen, M.B., Moe, R.O., Spruijt, B., Keeling, L.J., Winckler, C., Forkman, B., Dimitrov, I., Langbein, J., Bakken, M., Veissier, I., Aubert, A., 2007. Assessment of positive emotions in animals to improve their welfare. Physiol. Behav. 92, 375-397. Broster, C.E., Burn, C.C., Barr, A.R.S., Whay, H.R., 2009. The range and prevalence of pathological abnormalities associated with lameness in working horses from developing countries. Equine Vet. J. 41, 474-481. Burden, F.A., Du Toit, N., Hernandez-Gil, M., Prado-Ortiz, O., Trawford, A.F., in press. Selected health and management issues facing working donkeys presented for veterinary treatment in rural Mexico: some possible risk factors and potential intervention strategies. Trop. Anim. Health Prod. Burn, C.C., Pritchard, J.C., Whay, H.R., 2009. Observer reliability for working equine welfare assessment: problems with high prevalences of certain results. Anim. Welf. 18, 177-187. Burn, C.C., Dennison, T.L., Whay, H.R., in press. Environmental and demographic risk factors for poor welfare in working horses, donkeys, and mules in developing countries. Vet. J. Dantzer, R., Kelley, K.W., 2007. Twenty years of research on cytokine-induced sickness behavior. Brain. Behav. Immun. 21, 153-160. Dawkins, M.S., Donnelly, C.A., Jones, T.A., 2004. Chicken welfare is influenced more by housing conditions than by stocking density. Nature 427, 342-344. de Aluja, A.S., 1998. The welfare of working equids in Mexico. Appl. Anim. Behav. Sci. 59, 19-29.
20
536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584
de Passille, A.M., Rushen, J., 2005. Can we measure human-animal interactions in on-farm animal welfare assessment?: Some unresolved issues. Appl. Anim. Behav. Sci. 92, 193-209. Dunn, A.J., Swiergiel, A.H., Beaurepaire, R.d., 2005. Cytokines as mediators of depression: What can we learn from animal studies? Neurosci. Biobehav. Rev. 29, 891-909. FAOSTAT. (2006). FAO statistical database website. Food and Agricultural Organisation of the United Nations. Retrieved 7 July 2008, from http://faostat.fao.org/site/409/default.aspx Gul, S.T., Ahmad, M., Khan, A., Hussain, I., 2007. Haemato-biochemical observations in apparently healthy equine species. Pak. Vet. J. 27, 155-158. Hailat, N.Q., Lafi, S.Q., Al-Darraji, A.M., Al-Ani, F.K., 1997. Equine babesiosis associated with strenuous exercise: clinical and pathological studies in Jordan. Vet. Parasitol. 69, 1-8. Hart, B.L., 1988. Biological basis of the behavior of sick animals. Neurosci. Biobehav. Rev. 12, 123-137. Hausberger, M., Roche, H., Henry, S., Visser, E.K., 2008. A review of the humanhorse relationship. Appl. Anim. Behav. Sci. 109, 1-24. Hemsworth, P.H., 2003. Human-animal interactions in livestock production. Appl. Anim. Behav. Sci. 81, 185-198. Hillyer, M., 2004. A practical approach to diarrhoea in the adult horse. In Pract. 26, 211. Kelley, K.W., Bluthé, R.-M., Dantzer, R., Zhou, J.-H., Shen, W.-H., Johnson, R.W., Broussard, S.R., 2003. Cytokine-induced sickness behavior. Brain. Behav. Immun. 17, S112-S118. Millman, S.T., 2007. Sickness behaviour and its relevance to animal welfare assessment at the group level. Anim. Welf. 16, 123-125. Paul, E.S., Harding, E.J., Mendl, M., 2005. Measuring emotional processes in animals: the utility of a cognitive approach. Neurosci. Biobehav. Rev. 29, 469491. Pritchard, J.C., Whay, H.R., 2003 (unpublished). Guidance notes to accompany working equine welfare assessment. Bristol: University of Bristol. Pritchard, J.C., Whay, H.R., 2004 (unpublished). Guidance notes to accompany working equine welfare assessment. Bristol: University of Bristol. Pritchard, J.C., Lindberg, A.C., Main, D.C.J., Whay, H.R., 2005. Assessment of the welfare of working horses, mules and donkeys, using health and behaviour parameters. Prev. Vet. Med. 69, 265-283. Pritchard, J.C., Barr, A.R.S., Whay, H.R., 2006. Validity of a behavioural measure of heat stress and a skin tent test for dehydration in working horses and donkeys. Equine Vet. J. 38, 433-438. Pritchard, J.C., Barr, A.R.S., Whay, H.R., 2007. Repeatability of a skin tent test for dehydration in working horses and donkeys. Anim. Welf. 16, 181-183. Pritchard, J.C., Burn, C.C., Barr, A.R.S., Whay, H.R., 2008. Validity of indicators of dehydration in working horses: a longitudinal study of changes in skin tent duration, mucous membrane dryness and drinking behaviour. Equine Vet. J. 40, 558-564. Pritchard, J.C., Burn, C.C., Whay, H.R., 2009. Haematological and serum biochemical reference values for apparently healthy working horses in Pakistan. Res. Vet. Sci. 87, 389-395.
21
585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619
Regan, F.H., 2009a. Behavioural indicators of pain in working donkeys. Unpublished PhD thesis, University of Bristol, Bristol. Regan, F.H., 2009b. Behavioural responses to the provision of a non-steroidal antiinflammatory drug (meloxicam) (Chapter 4) In: Behavioural Indicators of Pain in Working Donkeys. Unpublished PhD thesis, University of Bristol, Bristol. Rushen, J., Taylor, A.A., de-Passille, A.-M., 1999. Domestic animals' fear of humans and its effect on their welfare. Appl. Anim. Behav. Sci. 65, 285-303. Rutherford, K.M.D., Langford, F.M., Jack, M.C., Sherwood, L., Lawrence, A.B., Haskell, M.J., 2009. Lameness prevalence and risk factors in organic and nonorganic dairy herds in the United Kingdom. Vet. J. 180, 95-105. Saul, C., Siefert, L., Opuda-Asibo, J., in press. Disease and health problems of donkeys: a case study from eastern Uganda. In Donkeys and donkey technology: Animal Traction Network for Eastern and Southern Africa. Tadich, N., Mendez, G., Wittwer, F., Meyer, K., 1997. Valores bioquímicos sanguíneos de equinos que tiran carretones en la ciudad de Valdivia (Chile) (Blood biochemical values of loadcart draught horses in the city of Valdivia (Chile)). Arch. Med. Vet. 29, 45-53. Tesfaye, A., Curran, M.M., 2005. A longitudinal survey of market donkeys in Ethiopia. Trop. Anim. Health Prod. 37, 87-100. Thoefner, M.B., Ersbøll, A.K., Jensen, A.L., Hesselholt, M., 2001. Factor analysis of the interrelationships between clinical variables in horses with colic. Prev. Vet. Med. 48, 201-214. Vollmayr, B., Henn, F.A., 2003. Stress models of depression. Clin. Neurosci. Res. 3, 245-251. Waiblinger, S., Boivin, X., Pedersen, V., Tosi, M.-V., Janczak, A.M., Visser, E.K., Jones, R.B., 2006. Assessing the human-animal relationship in farmed species: A critical review. Appl. Anim. Behav. Sci. 101, 185-242. Waters, A.J., Nicol, C.J., French, N.P., 2002. Factors influencing the development of stereotypic and redirected behaviours in young horses: findings of a four year prospective epidemiological study. Equine Vet. J. 34, 572-579. Weary, D.M., Huzzey, J.M., von Keyserlingk, M.A.G., 2009. Using behavior to predict and identify ill health in animals. J. Anim. Sci. 87, 770-777. Yeates, J.W., Main, D.C.J., 2008. Assessment of positive welfare: A review. Vet. J. 175, 293-300.
22
620
Table 1
621 Variable
Possible categorisations
Brief definition
Age
<5 / 5-15 / >15
(Assessed by observing the teeth)
Sex
stallion / gelding / mare
N/A
Work type
Tourism (riding)/ Tourism (carriage)/ Human transport (carriage)/ Human transport (riding)/ Goods transport (cart)/ Goods transport (pack)/ Agriculture/ Brick kiln (cart)/ Brick kiln (pack)/ Ceremonial/ Foal of working mother/ or Other
N/A
General alertness
alert / apathetic / depressed (apathetic and depressed were combined in the current study)
Responding to surroundings e.g. ears moving and often forward, eyes open, feet may be moving, tail swishing, head up unless sniffing or eating/ passive response to surroundings e.g. small ear movements, some tail swishing, feet may be moving, eyes may be half-closed, head may be lowered/ Unresponsive to surroundings, e.g. ears still and lowered, eyes closed or half-closed, no tail-swishing or foot movement, head lowered.
Observer approach
moves away / turns head away / no response / friendly/ aggressive (moves away and turns away were combined in the current study)
Moves or attempts to move away/ turns head away/ no obvious response/ Turns head towards, ears forward / Attempts to bite, rear, kick or strike with foreleg; ears held back or flattened
Walk-beside
no response / signs of attention
No obvious response/ signs of attention e.g. ears turn towards, head turns towards, moves towards or away, attempts to kick
Tail-tuck (donkeys only)
no response/ tail-tuck
Tail remains relaxed/ clamps down tail or tucks in hindquarters
Chin-contact
accepts / avoids
Shows no response chin-contact/ Moves its head to avoid or reject contact
Body condition
1 – 5 (including half-scores)
Ribs, spine and hips prominent, neck topline concave, hollow pelvis/ Ribs, spine and hips visible, flat pelvis/ Spine just visible, neck topline straight, slightly rounded pelvis/ Spine not visible, neck topline slightly convex, rounded pelvis/ neck topline distinctly convex, rounded pelvis with ‘gutter’ along spine
Faecal soiling
faecal soiling present / absent
Spattered faecal matter, dried or fresh, on inner thighs or back of hocks/ absent e.g. no staining or staining from mud, grass or other materials
Ectoparasites
present / absent
Ticks, mites, bot eggs, lice, or lice eggs anywhere on
General
Behaviour
General health
23
the body/ none observed Eyes
no abnormalities / abnormal
Healthy eyes/ At least one eye with wet eyelashes, discharge, redness, swelling, opacity or injury
Mucous membranes
normal colour / abnormal
Pink/ Abnormal e.g. yellowish, pale, reddish, greyish or purpleish
Teeth missing
yes / no (only assessed during 2003-2004)
At least one tooth missing/ All teeth present
General lesions
Severity score (0, 1, 2, 3) x number of affected locations on body
Severity scores: <2x2 cm / superficial / broken skin / deep Locations of lesions: breast,and shoulders, ears, forelegs, girth and belly, head, hindlegs, hindquarters, knees, lips, neck, point-of -hock, ribs, flank, tail and tailbase, withers and spine
Firing lesions
Severity score (0, 1, 2, 3)
Severity scores: 0 - 2x2 cm / superficial / broken skin / deep
Swollen tendons and joints
yes / no
Cannon bone, suspensory ligament, flexor tendons and fetlock joint are clearly visible and distinct from each other in all four legs/ In one or more legs, the distinction between the suspensory ligament, tendons and joint become unclear; the lower part of the tendon and the back of the fetlock joint may look bulgy, lumpy or square in shape
Hoof horn quality
normal / abnormal
Healthy / abnormalities e.g. broken hoof wall, cracks, ridges, wavy or non-parallel rings
Hoof shape
normal / abnormal
Medio-laterally symmetrical, coronary band horizontal, and toe and heel wall slopes parallel / Medio-laterally asymmetrical, coronary band slanted, and toe and heel not parallel (e.g. toe overgrown or excessively short)
Sole shape and structure
normal / abnormal
Round in horses; keyhole-shaped in donkeys; symmetrical, with intact frog, distinct bars and slightly cupped sole / asymmetrical, cracked, oval or pearshaped; frog narrow, hard, atrophied or missing; bars missing; sole flat; or sheared heels
Gait
normal / abnormal
Normal and even / any reluctance to put weight on a limb, asymmetrical dropping or raising of hip, uneven nodding of head, short or uneven stride length
Skin lesions
Limb and foot pathology
622 623 624 625 626 627
Brief descriptions of the behavioural and physical measures taken as part of a working equine welfare assessment. More detailed descriptions can be found in a 102-page, illustrated guidance booklet (Pritchard & Whay, 2003 (unpublished); 2004 (unpublished), available upon request from the authors). The protocol and sequence in which these measurements were taken are described in the text.
24
628
629 630 631 632 633 634
Table 2 Behavioural response
Donkeys (%)
Horses (%)
Mules (%)
General alertness - Apathetic/ depressed
13.1
13.7
13.4
Observer approach - No response
64.6
68.7
61.5
Observer approach - Avoidance
25.9
14.2
23.0
Observer approach - Aggression
0.3
1.7
0.7
Observer approach - Friendly
11.3
15.5
14.8
Walk-beside - Signs of attention
91.2
86.6
90.0
Tail-tuck
21.2
N/A
N/A
Chin-contact - Avoidance
16.4
20.6
15.4
The percentages of each working equine species displaying the behaviours recorded here. The percentages are compiled from 5481 donkeys, 4504 horses, and 858 mules across nine developing countries. In the general alertness measurement, apathy and depression were combined, and in the observer approach test, moving and turning away were summed as ‘avoidance’.
25
635
Table 3 Behaviour General alertness Apathetic/ depressed
Predictor Observer approach - No response Walk-beside - No response
Observer approach Avoidance
Chin-contact Avoidance Walk-beside - Signs of attention
Observer approach No response
Walk-beside - No response Chin-contact - Accepts
Observer approach Aggression
Chin-contact Avoidance
Observer approach Friendly
Walk-beside - Signs of attention
Walk-beside - No response Tail-tuck
636 637 638 639 640 641 642 643 644 645
Chin-contact - Accepts Chin-contact - Accepts Observer approach Avoidance Walk-beside - Signs of attention Chin-contact Avoidance
Species Donkey
Odds ratio ± SE 1.361 ± 1.110
P-value 0.0030
Horse Donkey
1.401 ± 1.124 3.177 ± 1.124
0.0042 <0.0001
Horse Mule Donkey
2.307 ± 1.126 3.815 ± 1.298 2.408 ± 1.091
<0.0001 <0.0001 <0.0001
Horse Mule Donkey
2.214 ± 1.104 3.022 ± 1.234 1.931 ± 1.140
<0.0001 <0.0001 <0.0001
Mule Donkey
3.130 ± 1.527 2.361 ± 1.129
0.0071 <0.0001
Horse Mule Donkey Horse Mule Donkey
2.344 ± 1.125 4.183 ± 1.411 1.982 ± 1.088 1.859 ± 1.089 2.012 ± 1.240 3.158 ± 1.428
<0.0001 <0.0001 <0.0001 <0.0001 0.0012 0.0012
Horse Donkey
5.104 ± 1.244 2.377 ± 1.251
<0.0001 0.0001
Horse Mule Donkey Horse
2.678 ± 1.182 3.180 ± 1.692 1.377 ± 1.156 1.445 ± 1.138
<0.0001 0.0281† 0.0268† 0.0043
Donkey
2.257 ± 1.087
<0.0001
Donkey
2.416 ± 1.181
<0.0001
Donkey
1.623 ± 1.096
<0.0001
Statistically significant associations between behaviours recorded as part of a working equine welfare assessment, split by equine species. Five behavioural measurements were made: general alertness, observer approach, walk-beside, tail-tuck (in donkeys only), and chin-contact. Because the study was correlational, the associations between the behaviours in the first and second columns could be stated either way around, but the way they are reported in the table reflects the statistical models. The effect sizes are presented as odds ratios (± SE), and the alpha level was set at a P-value of 0.016 to adjust for the false discovery rate; †indicates trends for which P <0.05, but which failed to meet the adjusted significance level.
26
646
Table 4 Behaviour
Predictor
Species
P-value
Donkey
Odds ratio ± SE 2.570 ± 1.093
General alertness Apathetic/ depressed
Lower body condition score
Horse Mule Donkey Horse Mule Donkey Horse Mule Donkey Horse Donkey Horse Donkey Horse Horse Horse Donkey Donkey Donkey
1.859 ± 1.094 2.654 ± 1.271 1.166 ± 1.046 1.133 ± 1.058 1.350 ± 1.146 1.279 ± 1.097 1.379 ± 1.115 1.786 ± 1.259 1.224 ± 1.103 1.283 ± 1.108 2.052 ± 1.071 1.310 ± 1.080 1.300 ± 1.130 2.034 ± 1.433 2.356 ± 1.256 2.492 ± 1.586 1.527 ± 1.163 1.944 ± 1.307 1.504 ± 1.067
<0.0001 0.0001 0.0006 0.0258† 0.0275† 0.0085 0.0031 0.0120 0.0403† 0.0156 <0.0001 0.0005 0.0320† 0.0485† 0.0002 0.0479† 0.0051 0.0131 <0.0001
Abnormal mucous membranes Higher body condition score
Horse Mule Donkey Horse Mule Donkey
1.156 ± 1.064 1.571 ± 1.192 1.225 ± 1.054 1.339 ± 1.064 1.636 ± 1.213 1.328 ± 1.070
0.0194† 0.0103 0.0001 <0.0001 0.0112 <0.0001
Fewer tendon/joint swellings Normal gait Higher body condition score
Donkey Horse Donkey
1.310 ± 1.117 2.004 ± 1.208 2.509 ± 1.297
0.0144 0.0002 0.0004
Younger age Younger age
Mule Donkey
3.963 ± 1.586 1.324 ± 1.083
0.0029 0.0004
Horse Donkey Mule Donkey Donkey Horse Donkey Mule Donkey Mule Horse Horse
1.537 ± 1.077 1.358 ± 1.097 1.632 ± 1.221 1.357 ± 1.140 1.212 ± 1.082 1.252 ± 1.078 1.347 ± 1.103 1.933 ± 1.278 1.150 ± 1.061 1.974 ± 1.269 1.099 ± 1.046 1.363 ± 1.102
<0.0001 0.001 0.0148 0.0204† 0.0147 0.0027 0.0023 0.0073 0.0186† 0.0043 0.0383† <0.0001
More severe lesions Abnormal mucous membranes Faecal soiling Older age
General alertness - Alert Observer approach - No response
Ectoparasites Abnormal gait Eye abnormalities Teeth missing Abnormal hoof shape Sole abnormalities Lower body condition score
Older age Observer approach Avoidance Observer approach Aggression Observer approach Friendly
Higher body condition score Walk-beside - No response
Faecal soiling Older age Lower body condition score
Tail-tuck Chin-contact - Accepts
647 648 649 650 651
Younger age Faecal soiling More severe lesions Younger age
<0.0001
Statistically significant associations between behaviours and physical problems recorded as part of a working equine welfare assessment, split by equine species. Five behavioural measurements were made: general alertness, observer approach, walkbeside, tail-tuck (in donkeys only), and chin-contact. Because the study was 27
652 653 654 655 656 657
correlational, the associations between the variables in the first and second columns could be stated either way around, but the way they are reported in the table reflects the way they were tested in the statistical models. The effect sizes are presented as odds ratios (± SE), and the alpha level was set at a P-value of ≤0.016 to adjust for the false discovery rate; †indicates trends for which P <0.05, but which failed to meet the adjusted significance level.
28