Applied Animal Behaviour Science 140 (2012) 33–43

Contents lists available at SciVerse ScienceDirect

Applied Animal Behaviour Science

journal homepage: www.elsevier.com/locate/applanim

A novel scale of behavioural indicators of stress for use

with domestic horses
Tamsin Young a,∗ , Emma Creighton b , Tessa Smith a , Charlotte Hosie a
a
Department of Biological Sciences, University of Chester, Chester, CH1 4BJ, UK
b
School of Agriculture, Food and Rural Development, Newcastle University, Newcastle, NE1 7RU, UK

a r t i c l e i n f o a b s t r a c t

Article history: Behaviour scores (BS) offer non-invasive, objective and easy to use ways of assessing wel-
Accepted 22 May 2012 fare in animals. Their development has, however, largely focused on behavioural reactions
Available online 22 June 2012 to stressful events (often induced), and little use of physiological measures has been made
to underpin and validate the behavioural measures. This study aimed to develop a phys-
Keywords:
iologically validated scale of behavioural indicators of stress for the purpose of welfare
Horse
assessment in stabled domestic horses. To achieve this, behavioural and physiological data
Behaviour scores
Cortisol
were collected from 32 horses that underwent routine husbandry procedures. Principal
Saliva component analysis (PCA) of the behavioural and physiological data revealed three mean-
Welfare assessment ingful components that were used as the basis of the scale. Analysis of video clips of the
Non-invasive horses’ responses to the husbandry procedures was undertaken by a panel of equestrian
industry professionals using a free choice profiling (FCP) methodology. These results were
added to the scale along with key definitions from relevant literature. Salivary cortisol levels
were significantly correlated with the BS confirming the scale was meaningful and reflected
physiological stress. The scale offers an easy to use ‘tool’ for rapid, reliable non-invasive
welfare assessment in horses, and reduces the need for potentially invasive physiological
measures.
© 2012 Elsevier B.V. All rights reserved.

1. Introduction developed by focusing on expression of behaviours

assumed to indicate stress rather than making use of estab-
Use of behaviour scores (BS) offers objective, imme- lished physiological indicators of stress to underpin them.
diate methods of welfare assessment in animals (Minka In addition, BS are often used rather simply, ignoring all but
et al., 2009). They have been used, with varying degrees of ‘negative’ behaviours (e.g. Minka et al., 2009) and therefore
success, to measure stress levels for the purpose of wel- losing potentially valuable information. Of the small num-
fare assessment in various species (horses: Visser et al., ber of behaviour scoring approaches to welfare assessment
2010; Munsters et al., 2012; cats: McCune, 1994; Kessler available, only the Cat Stress Score (McCune, 1994; Kessler
and Turner, 1997; McCobb et al., 2005; Dybdall et al., and Turner, 1997) provides a scale of stress responses that
2007; goats: Minka et al., 2009; cattle: Maria et al., can be used flexibly to assess welfare in different environ-
2004b; ostriches: Minka and Ayo, 2008; chickens: Maria ments. This scale has not been physiologically validated,
et al., 2004a). These scores have, however, been largely however.
Assessment of a stress response is clearly best car-
ried out using a combination of both behavioural and
∗ Corresponding author at: Glyndŵr University, Northop Campus, physiological measures (Broom, 1991; Mason and Mendl,
Mold, Flintshire, CH7 6AA, UK. Tel.: +44 01978 293962. 1993; Dawkins, 2003). This provides a more compre-
E-mail address: [email protected] (T. Young). hensive measurement of stress, and avoids drawing

0168-1591/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.applanim.2012.05.008
34 T. Young et al. / Applied Animal Behaviour Science 140 (2012) 33–43

misleading conclusions that could be reached by taking kept in similar management and exercise regimes, at four
just a single measure (Broom, 1991; Mason and Mendl, different locations (Table 1). All horses were housed in indi-
1993; Dawkins, 2003). Previous success at correlating vidual stables on either straw or shavings bedding, and
behavioural measures with physiological measures has, received hay or haylage and water with up to two hard
however, been mixed. Minka et al. (2009) established feeds i.e. mix or pellets, at around 07:00 h and 16:00 h. All
that certain behaviours and physiological indices of stress horses were in light to medium work (receiving around 2 h
were related during the handling and loading of goats for of exercise daily) throughout the study. When they were
transportation, but McCobb et al. (2005) were not able to not being exercised all horses received up to 4 h turn out
correlate BS with urinary cortisol measures in cats. Clearly to pasture daily, and remained in their usual daily man-
the behavioural and physiological measures must both be agement routine apart from undergoing routine husbandry
sufficiently sensitive and reliable to successfully correlate procedures.
them for use in a behavioural scale for welfare assessment. A total of 32 horses were used in this study and assigned
The only scale of BS available for use with domestic to experimental groups on the basis of their availability (see
horses was developed to assess, specifically, whether horse Table 1 for details of location, age, gender and husbandry
and rider combinations were appropriate (Munsters et al., procedure). In summary, analysis of behaviour was com-
2012). The scale was adapted from a scoring system used pleted from all 32 horses; nineteen of the horses were used
by Visser et al. (2010) to assess the temperaments of sports for saliva collection for cortisol measurement, with a fur-
horses exposed to novel objects, not to evaluate stress ther 10 horses from location one used as a control group
responses. Some level of experience/expertise was needed for this part of the study; eighteen horses were used for
to accurately assess temperament in this latter study. Phys- measurement of HR, with a further 10 horses from location
iological measures were not used to develop the score, one again used as a separate control group for this aspect
although heart rate variability (HRV) was used when the of the study.
score was tested.
The aim of the current study was, therefore, to develop
2.2. Husbandry procedures used for data collection
an easy to use scale of BS that could be used to rapidly but
reliably assess stress levels in domestic stabled horses. The
Horses were subjected individually to one of four dif-
scale was developed using both behavioural and physiolog-
ferent 10 min husbandry procedures. They remained loose
ical measures obtained from a wide range of horses (N = 32)
in their stables, except during grooming when they were
during naturally occurring daily routine husbandry proce-
loosely held by a familiar handler and had access to hay and
dures. Use of such husbandry procedures, that would have
water. A 10 min period was deemed adequate to induce
taken place irrespective of the study, was considered an
and measure a potential stress response, but not so long
ethical approach to data collection rather than artificially
that habituation to the stressor should occur (Visser et al.,
exposing horses to potentially stressful situations.
2001).
Since assessment of welfare is most effective using
multiple measures, analysis of the behavioural responses
to the routine husbandry procedures used two differ- 2.2.1. Procedure 1 – sound of electric coat clippers
ent approaches. Behavioural responses were quantified in Horses (from location 1 only, for logistical reasons) were
detail using The Observer (Noldus Information Technology exposed to the sound of electric coat clippers (Heiniger
Software Ltd) for 12 of the horses. Both quantitative and Handy Clipper, Switzerland) turned on to maximum clip-
qualitative assessment of the behaviour of all study horses ping velocity. Typical sound emissions from such clippers
was made, by expert panel members, using the experi- were 80.1 db. Clippers were switched on and held by hand
mental free choice profiling (FCP) approach (Wemelsfelder in an adjacent stable 3.66 m away from the study horse.
et al., 2000, 2001).
We again used two measures for the physiological data
2.2.2. Procedure 2 – social isolation
collection, heart rate (HR) and salivary cortisol, to validate
Horses (from locations 1, 2, 3) were caught from the
the BS scale. Both cortisol (Ralston et al., 1988; Toutain et al.,
field and returned to their usual stable. This process took
1995; McBride and Cuddeford, 2001; Covalesky et al., 1992;
no longer than 5 min, and horses showed no resistance to
Stegman and Jones, 1998) and HR (Rietmann et al., 2004;
capture. The horses were stabled in the absence of any
Von Borell et al., 2007; Visser et al., 2008) are established
other horses on the yard for 10 min. At the end of the social
indicators of stress in horses, and detect different aspects
isolation period the horse’s usual neighbouring horse was
of the stress response. Both indices of stress were used
returned to the adjacent stable.
here as their measurement can be achieved by non-invasive
means, thus avoiding any further stress to the study horses.
2.2.3. Procedure 3 – grooming procedures
2. Method A head-collar and lead rope was fitted to the horse, and a
familiar handler held the lead rope approximately half way
2.1. Subjects used for the study along its length to restrain the horse loosely. Mane combing
and mane pulling (a procedure used to thin and shorten the
The study aimed to ensure a wide range of horses was mane by taking small sections of hair back combing them
used to build the BS scale. The horses used in the study and pulling out the remaining long hairs) then took place.
consisted of various breeds of stabled mares and geldings Animals were used from location 1.
 T. Young et al. / Applied Animal Behaviour Science 140 (2012) 33–43 35

Table 1
Details of the subjects used in the study (N = 32).

Horse Location Age Gender Breed/type Husbandry procedures

Coat Social Grooming Firework

clippers isolation sounds
√
1 1 10 Mare Thoroughbred X
√
2 1 7 Gelding Cob X
√
3 1 12 Mare Thoroughbred X
√
4 1 6 Gelding Thoroughbred
√
5 1 10 Gelding Clydesdale
√
6 1 11 Mare Thoroughbred X
√
7 1 8 Mare Thoroughbred
√
8 1 5 Mare Thoroughbred
√
9 1 15 Gelding Thoroughbred X
√
10 1 14 Mare Welsh
√
11 1 20 Mare Welsh
√
12 1 10 Mare Welsh
√
13 1 9 Mare Cob X
√
14 1 8 Mare Arab X
√
15 1 8 Gelding Cleveland Bay
√
16 1 13 Mare Thoroughbred X ID
√
17 1 9 Mare Welsh
√
18 1 20 Gelding Thoroughbred
√
19 1 12 Mare Welsh X
√
20 1 10 Mare Welsh
√
21 1 9 Gelding Cob X
√
22 1 9 Gelding Thoroughbred
√
23 1 10 Mare Thoroughbred X
√
24 2 12 Gelding Thoroughbred X
√
25 2 11 Gelding Thoroughbred X
√
26 3 12 Gelding Irish sports horse
√
27 3 18 Gelding Thoroughbred
√
28 3 12 Gelding Irish sports horse
√
29 4 13 Gelding Cob X
√
30 4 7 Gelding Irish sports horse
√
31 4 9 Gelding Thoroughbred
√
32 4 12 Gelding Cob X

Location one = 23 horses, location two = 2 horses, location three = 3 horses, location four = 4 horses.
Husbandry procedures: clippers = 13 horses, social isolation = 6 horses, grooming = 9 horses, firework sounds = 4 horses.

2.2.4. Procedure 4 – the sound of fireworks played on a 2.3. Behavioural measurement during the husbandry
CD procedures
Police horses (from location 4 only) were used for this
procedure as it involved the sound of fireworks played on The behaviour of the subjects was recorded during all
a compact disc (CD), which was used as part of riot training husbandry procedures using a Sanyo CCD/BW video cam-
with Police horses. The CD player was situated on a table era (Sanyo Electric Co., Ltd, Osaka, Japan) secured at ceiling
outside the horse’s stable 3.66 m away. height in an appropriate position opposite the stable to gain
The husbandry procedures were carried out over a num- an adequate field of view. The video camera was linked to
ber of weeks as the opportunities to collect data arose. a Mitsubishi HS-1024E time-lapse recorder (Osaka, Japan),
Where the same procedures were carried out with a num- set to 3 h real time for recording of images onto 3 h video
ber of horses on the same yard e.g. exposure to the tapes (BASF Vision Chrome Videocassette, BASF plc, Mid-
sound of electric coat clippers, one week was left between dlesex, UK).
tests to minimise the effects of habituation on the horses
that had not yet been sampled from. The control horses 2.3.1. Analysis of behaviour using The Observer
were chosen randomly from location one, as these had The first 5 min of the behavioural reactions exhibited
not been subjected to the husbandry procedures exam- during the husbandry procedures, for 12 of the horses, was
ined in the study for a minimum of eight weeks. This analysed using The Observer 5.0. Behaviours were recorded
opportunistic data collection strategy meant that some using a pre-defined ethogram based on equine stable
procedures/control were tested in a single location, oth- behaviour (Table 2). The ethogram had been compiled from
ers across more than one. Location effects were tested six weeks of ad-hoc observation of race-horses and stabled
for in the analysis to check that this had no confounding riding horses, together with literature research (see Houpt,
impact. 1993; Winskill et al., 1996; McBride and Cuddeford, 2001;
36 T. Young et al. / Applied Animal Behaviour Science 140 (2012) 33–43

Table 2
The ethogram used for The Observer configuration.

Behaviour Behavioural description

Whole body
Rolling Drop to knees, roll onto side then back and perhaps all the way over then return to feet (adapted from Strand et al., 2002).
Lying Lying with sternum in contact with ground surface, legs folded under the body or stretched out on side with legs stretched out
(adapted from Heleski et al., 2002).
Standing at front Standing with weight resting on 3 or 4 legs with one or both ears in front half of stable (adapted from Seaman et al., 2002).
Standing at back Standing with weight resting on 3 or 4 legs with both ears in back half of stable (adapted from Seaman et al., 2002).
Walking Four beat gait of forward movement (Heleski et al., 2002).
Trotting Two beat diagonal gait of forward movement (Heleski et al., 2002).
Buck A springing motion consisting of raising the hindquarters off the ground sometimes accompanied by a longitudinal twisting of
the body (adapted from Mal et al., 1991).
Rear Front legs raised off the ground with forehand higher than hindquarters (adapted from Heleski et al., 2002).
Barging Forceful contact directed at a part of the stable either from a standstill or from any gait.
Scratching Rubbing any part of the body against part of the stable or using hind foot or teeth to scratch part of own body (adapted from
Heleski et al., 2002).
Pawing Striking a vertical or horizontal surface or the air with forelimb (Seaman et al., 2002).
Kicking One or both hind legs thrust backwards or to the side, contact with part of the stable may be achieved (adapted from Seaman
et al., 2002).
Weaving Lateral movement of the head, neck and shoulders from side to side in a rhythmic repetitive manner with alternation of the
weight onto the contralateral foreleg with respect to the position of the head (adapted from McBride and Cuddeford, 2001).
Shake Rapid rhythmic rotation of head, neck and upper body along the axis whilst standing with feet planted (McDonnell, 2003).
Stretch Rigid extension of the limbs and arching of the neck and back (McDonnell, 2003).
Tail
Raised Fleshy part of tail outstretched horizontally or elevated above horizontal.
Neutral Fleshy part of tail relaxed against body or moving slowly from side to side never raised to horizontal level.
Swish Tail is flicked to one side and/or the other of the quarters.
Clamped Fleshy part of the tail is forced close to the dock. Possible tensing of the quarters.
Defecation/Urination Elimination of faeces and urine
Neck
Above withers Eye level elevated above height of withers.
Below withers Eye level parallel or below height of withers.
Over 45 degrees Neck raised over 45 degrees
Ears
Forward Ears pricked up pointing forwards and stationary for 3 s or more.
Scanning Ears moving back and forth at varying speeds.
Back Ears apart or gently back and stationary for 3 s or more.
Flat Ears pressed caudally against the head and neck (McDonnell and Haviland, 1995).
Mouth
Eating/Drinking Prehending, masticating or swallowing food or water (adapted from Winskill et al., 1996).
Exploratory Lick, sniff or touch with muzzle or tongue parts of the stable or floor (adapted from Strand et al., 2002).
Self-Care Non-ingestive behaviours involving the muzzle and teeth including allogrooming and swatting flies on body.
Repetitive oral A repeated, relatively invariant sequence of movements with no obvious function using the teeth or lips in contact with parts
of the stable or own body (adapted from Mason, 1991). Includes cribbing, teethscraping, wood chewing, windsucking, biting,
lipsmacking and self mutilation.
Yawn Deep long inhalation with mouth widely open and jaws either directly opposed or moved from side to side (McDonnell, 2003).
Head
Surveying Head scanning through forty-five degrees or more.
Repetitive head A repeated, relatively invariant sequence of movements with no obvious function (adapted from Mason, 1991) including
movement movements of the head such as headshaking, nodding, bobbing and circling.

Notes to support categories of mouth and head movements:

Repetitive oral
[•]
• Cribbing – horse grips onto a fixed object using incisor teeth, leans back onto hindquarters and contracts the strap muscle of the neck to bring the head
into an arched position. Air is sometimes taken into the oesophagus to produce a grunting sound (McBride and Cuddeford, 2001).
• Teeth scraping – lips are curled back to expose incisors. Lateral and corner incisors are scraped back and forth against side of solid object.
• Biting – bite movement directed at or in contact with part of the stable.
• Lipsmacking – incisors are kept shut whilst lips are opened and closed.
• Self-mutilation – bite movement directed at own body usually the flanks or the chest and limbs (adapted from Houpt, 1993).
• Wood chewing – teeth are used to chew parts of the stable. Ingestion of wood may occur.
• Windsucking – head and neck outstretched, mouth open slightly, air is taken into oesophagus.
Repetitive head movement
[•]
• Head shaking – head is shaken from side to side.
• Nodding – head is moved up and down.
• Bobbing – head performs a ‘pecking action’, where nose is only moved a small distance up and down.
• Circling – whole head is circled to the left or right. Whole or half circles maybe exhibited.
Self care
[•]
• Autogrooming – nibbling, biting, licking or rubbing a part of the body (McDonnell, 2003).
• Swat flies – head is swung to hit insects on body or limbs (adapted from Strand et al., 2002).
 T. Young et al. / Applied Animal Behaviour Science 140 (2012) 33–43 37

Strand et al., 2002; Heleski et al., 2002; Seaman et al., 2002; 2.5. Measurement of heart rate (HR) during the
McDonnell, 2003). husbandry procedures

HR was recorded from the experimental and control

groups of horses at 60-s intervals for 2 min prior to the
2.3.2. Analysis of behaviour carried out by a panel of
start of the husbandry procedure to provide a mean base-
equestrian professionals
line HR. Recording of HR then continued at 60-s intervals
A panel of 13 professionals who had a working back-
for the first 5 min of the husbandry procedure for the exper-
ground with horses, and held a minimum of the British
imental group of horses, and over the same time intervals,
Horse Society (BHS) stage one qualification was convened.
in the absence of a husbandry procedure, for the control
They were briefed on the nature of the study and asked to
group.
view a video of the initial 2 min of each horse’s behavioural
HR was recorded using a Polar HR monitor (S610i) (Polar
reaction to the husbandry procedures, and to provide a BS
Electro, Öy, Kempele, Finland). The HR monitor consisted
between zero and ten according to how stressed they per-
of an electrode belt that picked up the electrical activity
ceived the horse to be. They were told that ten equated to an
of the horse’s heart, with a transmitter attached enabling
extremely stressed horse. The panel also had to describe,
wireless transmission of the HR to a wrist watch receiver.
using their own descriptive terms, the horse’s behaviour
The belt was fitted around the horse’s thorax with both
exhibited during the video. They, finally, were asked to
electrodes sited to the left-hand side of the horse. One elec-
state at which point on their subjective scale that they
trode was placed approximately 10 cm below the withers,
believed the onset of stress occurred in the horses under-
and the other about 10 cm behind the elbow over the heart.
going the husbandry procedures.
Warm water and electrode gel (The Wyke of Shifnal, Shrop-
shire) was used to optimise contact between the horse’s
skin and the electrodes. The wrist watch receiver was taped
2.4. Measurement of salivary cortisol concentrations to a leather strap fastened around the horse’s neck. All
during the husbandry procedures horses were given 10 min to habituate to the equipment
(Rietmann et al., 2004).
Saliva was collected 60 min and 30 min prior to the start
of the husbandry procedures, and then at the end of the 2.6. Statistical analysis
10 min procedure and at 10 min intervals up to 40 min.
Forty minutes was chosen to provide enough time for peak Principal component analysis (PCA) was used to inves-
cortisol to be reached following the onset of the poten- tigate whether there were any relationships between
tial stressor (the husbandry procedure). Plasma cortisol behavioural and physiological changes that took place
peaks in horses 30 min post exercise stress (Foreman and during the husbandry procedures. ‘Data reduction’ was
Ferlazzo, 1996; Marc et al., 2000; Hamlin et al., 2002, and necessary to look for smaller sets of factors or compo-
20 min following restraint stress (Hydbring et al., 1996). nents in the data (Pallant, 2004; Ennos, 2007) from which
Saliva was collected from the control group of horses the scale of BS could be compiled. The percentage change
to ensure that the swabbing procedure did not affect their in cortisol concentration from the median baseline value
stress levels. Collection took place at 60 min, 30 min, and to the peak concentration was calculated for each horse.
0 min before the husbandry procedure would have begun, This percentage, together with the percentage duration
and then at the same time intervals that the experimen- of all behaviours included in the ethogram underwent
tal group of horses had their saliva collected, except for PCA.
the extra swab was taken at 0 min from control horses to PCA of the cortisol and behavioural data exhibited
provide a robust control measure. during the husbandry procedures revealed correlation
Saliva was collected using sterilised flexi-swabs (Medi- coefficients of 0.3 and above (following Pallant, 2004).
cal Wire & Equipment Co (Bath) Ltd) that were introduced An oblimin rotation of three factor solution was used
into the corner of the horses’ mouths first on the horse’s to reduce the number of variables into meaningful com-
left and then on the horse’s right. The horses were allowed ponents (Pallant, 2004). Each behaviour and change in
to manipulate the swabs using their tongues for approxi- cortisol concentration received a score for each compo-
mately 20 s per introduction of the swab. The swabs were nent denoting whether the behaviour was performed or
then placed into sterile 20 ml plastic screw top containers, not, or whether change in cortisol was relevant. A median
labelled and stored on ice until frozen at −20 ◦ C the same BS was calculated for each horse used in the study, as
day to await cortisol extraction. scored by members of the professional panel. The terms
Saliva was extracted from the thawed cotton wool used by the panel to describe each horse’s behaviour
swabs by centrifugation using a Sorvall T.C. centrifuge was pooled for horses with the same BS. Panel descrip-
(Thermo Scientific, Basingstoke, Hampshire, UK) for 2 min tions of behaviour were added to the relevant sets of
at 800 × g. The supernatant was then centrifuged using factors or components revealed by the PCA, and the scale
a Hettick Mikro 20 centrifuge (Tuttilgen, Germany) at of BS for use with stabled domestic horses subsequently
15,000 × g for 2 min. The supernatant was taken off using a devised.
pipette and frozen to await analysis. Salivary cortisol con- All HR and salivary cortisol data measured during the
centrations were quantified using a modified version of an husbandry procedures and controls were log transformed
EIA described by Smith and French (1997). to reduce heterogeneity of variance. The impact of location
38 T. Young et al. / Applied Animal Behaviour Science 140 (2012) 33–43

Table 3
Pattern matrix for PCA of the cortisol and duration of state behaviour recorded during the four different routine husbandry procedures using oblimin
rotation of a three factor solution.

Component

1 – no stress 2 – medium stress 3 – low stress

Change in cortisol concentration −0.648

Standing at the front 0.886
Standing at the back −0.806 −0.374
Walking 0.802
Scratching 0.863
Weaving 0.740
Eating and drinking −0.797 0.304
Exploring 0.860
Self-care 0.446 −0.402
Repetitive oral behaviour 0.364 −0.661
Other mouth behaviour 0.817
Surveying with head 0.602 −0.310 0.342
Repetitive head behaviour 0.838 0.477
Other head behaviour −0.600 −0.395
Tail raised 0.861
Neutral tail posture −0.817
Other tail behaviour 0.816
Neck height above withers −0.931
Neck height below withers 0.307 0.449 −0.346
Neck height over 45 degrees 0.936
Ears pricked 0.592 0.550
Ears scanning 0.885
Ears back −0.879 −0.309
Ears flat 0.740
Other ear behaviour −0.345
Tail swishing 0.635 0.349
Defecation 0.811
Eigenvalues 8.23 7.04 3.01
% variance accounted for 30.51 26.07 11.13

on each physiological measure was investigated to assess 3. Results

potential confounds caused by using subjects at different
locations. To do this, levels of baseline cortisol and HR data 3.1. Behavioural data
were compared across the two locations in the case of HR
data using an independent samples T test, and across the PCA of the percentage change in salivary cortisol
four locations in the case of salivary cortisol data using a from baseline to peak, and percentage duration of state
one-way between subjects ANOVA. behaviour exhibited during the husbandry procedures
Any impact of the Polar heart monitor over time was identified three components in the pattern matrix. They
explored during a control study by analysing HR at time were labelled no stress (factor 1), low stress (factor 3) and
zero and 60, 120, 180, 240 and 300 s as compared to medium stress (factor 2) according to the type of behaviour
the experimental timings, under control conditions. The and change in cortisol identified (Table 3).
impact of collecting saliva swabs on the HPA response Median behaviour scores were calculated for the study
was examined during a control study which assessed horses, and ranged between one and eight. The terms used
cortisol under control conditions at time 0, then 10, by the panel to describe each horse’s behaviour was pooled
20, 30 and 40 min later using a repeated subjects for horses with the same BS, and panel descriptions of
ANOVA. behaviour were added to the three components revealed
Changes in HR data and levels of salivary cortisol were by the PCA.
explored using General Linear Models (GLM) with sex, hus- Descriptions used for horses with a BS of one and two
bandry procedure and time of collection (i.e. before and were added to the component labelled no stress, as the
after the procedure in terms of the HR data, and peak com- mean score representing the onset of stress as judged by
pared to baseline cortisol titres for the salivary cortisol the panel was three. Descriptions used for horses with a BS
data) as fixed factors. Post hoc analyses were conducted of three to seven were added to low and medium stress.
using Tukey test and alpha was set at 0.05. The BS of five was used as the onset of medium stress,
To investigate whether the devised BS scale reflected based on the type of behaviour included in the component
physiological stress, median BS as calculated from the pro- extracted by the PCA. Descriptions used for horses with a BS
fessional panel, and peak salivary cortisol following the of eight to ten formed a new category labelled high stress,
husbandry procedures were investigated using Spearman’s because the PCA analysis did not include horses with a BS
Rank Order Correlation. of this level. Relevant literature was also used to form this
 T. Young et al. / Applied Animal Behaviour Science 140 (2012) 33–43 39

Table 4
A scale of behavioural indicators of stress in domestic stabled horses, as revealed by principal component analysis (PCA) and behavioural assessment
completed by a professional panel.

Stress level Behaviour Behavioural indicators

score

No stress 1 Standing at the front of the stable, looking around or head below wither height, eating. Ears pricked, back
or slowly scanning, tail still or gently swishing. Some repetitive oral behaviour.
Horse described as:
Horse calm, unconcerned, relaxed, quiet, listening, accepting.
2 Behaviour exhibited for previous BS plus:
Walking.
Horse also described as:
Horse alert and watching.
Low stress 3 Behaviour exhibited for previous stress level plus:
Occasional weaving behaviour, box walking and repetitive head movements. Ears occasionally flattened.
Defecation.
Horse described as:
Listening, interested, alert.
4 Behaviour exhibited for previous BS plus:
Pacing. Approaching potential stressors e.g. noise from outside the stable. Repeated tail swishing.
Horse also described as:
Curious, unsettled, barging.
Medium stress 5 Behaviour exhibited for previous stress level plus:
Scratching against stable walls or fittings, pawing at ground with front legs. Nostrils flared. Repeatedly
looking around. Tail raised.
Horse described as:
Restless, showing tension in the body, fidgeting when still.
6 Behaviour exhibited for previous BS plus:
Approaching and retreating away from potential stressors. Stopping eating to focus on potential stressor.
Horse also described as:
Jumpy, easily startled.
7 Behaviour exhibited for previous BS plus:
Keeping away from potential stressors and remaining still to focus on them.
Horse described as for previous BS.
High stress 8 Behaviour exhibited for previous stress level plus:
Repeated performance of stereotypic behaviour e.g. weaving, box walking repetitive head movements.
Stamping of hind feet. Snorting.
Horse described as:
Very unsettled and alert.
9–10 Behaviour as exhibited for previous BS.
Horse also described as:
Agitated, fidgety, anxious, active, aggressive, uncomfortable (McDonnell et al., 1999; Strand et al., 2002).

category. The scale of behavioural indicators of stress for procedures compared to baseline values (Fig. 1) [F(1,
use with stabled domestic horses was subsequently com- 26) = 10.083, P < 0.0001].
piled (Table 4). There was no effect of sex on HR values [F(1, 26) = 0.261,
NS], but there was a significant interaction between sex and
3.2. Physiological data husbandry procedure [F(1, 26) = 4.315, P < 0.05]. Further
analysis using a t test showed that HR values before and
Both measures of HR and salivary cortisol were used to after combined, for the clippers, were significantly higher
underpin the behavioural measures in the development of for mares than the same values for geldings (t = −3.403,
the scale of behavioural scores. df = 18, P < 0.003). In contrast for grooming, before and after
HR values (combined) were similar for both mares and
3.2.1. HR data geldings (t = 1.294, df = 6, NS). Only geldings were exposed
Baseline HR values were comparable across the two to the fireworks.
locations (t = −0.660, df = 16, NS) thus ruling out loca- There was no effect of husbandry procedure alone on HR
tion as a confound in the study. There were, also, values [F(2, 26) = 2.444, NS], or interaction between hus-
no significant changes in HR between baseline (mean bandry procedure and sampling time [F(2, 26) = 0.621, NS],
HR 37.20 ± 8.34 bpm) and the ‘test’ period (mean HR sex and sampling time [F(2, 26) = 0.169, NS], or between the
38.98 ± 15.65 bpm) in the control study suggesting that the three variables [F(1, 26) = 0.820, NS].
presence of the Polar heart rate monitor over a period of
time did not cause the horses any stress (t = −0.381, df = 9, 3.2.2. Salivary cortisol data
NS). Baseline salivary cortisol titres levels were comparable
HR values were explored using a general linear model across subjects in the four locations confirming that loca-
with sex, husbandry procedure and time of collection (i.e. tion was not a confound in our study [F(3, 18) = 1.824, NS].
before versus after the procedure) as the fixed factors. HR Changes in physiological data collected from control
values were significantly raised following the husbandry horses were explored to ensure that the methods of
40 T. Young et al. / Applied Animal Behaviour Science 140 (2012) 33–43

50.00

Mean Heart Rate +/- 1.0SE (bpm)

40.00

30.00

20.00

10.00

Baseline Husbandry
Fig. 1. Mean heart rate (bpm) ± 1.0 SE before (baseline) and during the husbandry procedures (N = 18).

data collection did not affect the horses. There were no 3.3. Correlating behavioural and physiological data
changes in levels of salivary cortisol across the control
study (mean ± s.e.m.) baseline 0.42 ± 0.12 ng/ml; at 10 min To ensure the final BS scale reflected increased levels
0.43 ± 0.2; at 20 min 0.35 ± 0.17; at 30 min 0.39 ± 0.15; at of both behavioural and physiological stress, their rela-
40 min 0.49 ± 0.25) suggesting that the saliva swabbing tionship in response to the husbandry procedures was
was not stressful to our subjects [F(4, 32) = 0.821, NS]. investigated. A significant correlation existed between
There was a significant effect of husbandry procedure median BS and peak salivary cortisol concentration mea-
on levels of salivary cortisol [F(3, 22) = 5.644, P < 0.005]. Post sured during the husbandry procedures (rs = 0.54, P = 0.02,
hoc analysis revealed that levels of the hormone (baseline N = 18) confirming that the final BS scale was a reflection of
and peak concentrations combined) pertaining to expo- both behavioural and physiological stress.
sure to the fireworks were significantly higher than those
related to clipping (P < 0.01) and social isolation (P < 0.01), 4. Discussion
but not grooming (NS). Hormone values relating to groom-
ing were higher than those relating to clipping (P < 0.05) In this study a scale of BS has been developed to measure
and social isolation (P < 0.01). stress levels in stabled domestic horses for the purposes
There was also a significant effect of time (i.e. base- of welfare assessment. To do this, behaviour and phys-
line cortisol versus peak concentration of cortisol) since iology were each measured by two different techniques
hormone levels were significantly elevated following the in a range of horses undergoing standard husbandry pro-
husbandry procedures compared to prior to the procedures cedures. The physiological data were used to underpin
(Fig. 2) [F(1, 22) = 6.077, P < 0.05]. There was, however, behavioural measures. Physiological data revealed that
no interaction between husbandry procedure and sample husbandry procedures did significantly elevate HR. HR was
time showing that despite gross differences in hormone greater for mares than geldings for the clipping proce-
values across the different procedures, the manner and dure but overall patterns of change in HR before and after
magnitude of the change in cortisol levels pre and post procedures was the same for both sexes. This shows that
treatment were comparable across the four husbandry the husbandry procedures used here were perceived as
procedures [F(3, 22) = 0.827, NS]. This was confirmed by stressful by the horses, and provides a biological valida-
two separate one-way ANOVAs which revealed compara- tion of these procedures as stressful events. These data also
ble hormone levels across the husbandry procedures before demonstrate the sensitivity of HR as a reliable physiological
[F(3, 14) = 3.035, NS] and after [F(3, 14) = 2.292, NS] the measure of stress.
treatments. There was no effect of sex on hormone val- Similarly, cortisol measures demonstrated that the
ues [F(1, 22) = 0.645, NS], and no interaction effects, i.e. husbandry procedures had a significant effect on stress
condition and sex [F(2, 22) = 2.184, NS], sex and time [F(1, physiology. Importantly these data also showed that loca-
22) = 0.000, NS], condition, sex and time [F(2, 22) = 0.701, tion (source of horses), sex and other factors did not affect
NS]. cortisol measures. Again, this provides biological validation
 T. Young et al. / Applied Animal Behaviour Science 140 (2012) 33–43 41

Mean Salivary Cortisol (ng/ml)

1.50

1.00

0.50

0.00

Baseline Peak
Fig. 2. Mean salivary cortisol (ng/ml) ± 1.0 SE before the husbandry procedures (baseline) and at peak following the procedure (N = 19).

of these procedures as stressful events and demonstrates sophisticated analysis to assess (Maria et al., 2004a). The
that salivary cortisol is a sensitive, reliable measure of BS indicators in our scale developed here are straightfor-
stress. These two physiological measures can, therefore, be ward and do not require great experience of horses to be
considered reliable to use to underpin the development easily used. In addition, the behavioural descriptions pro-
of the scale of behavioural scores. Schmidt et al. (2010) vide a simple scale to enable a range of levels of stress to
recently also established use of HR and salivary cortisol be measured. Of the small number of behaviour scoring
together as sensitive parameters to detect stress in ‘routine’ approaches to welfare assessment published, only the Cat
transport procedures for horses. Stress Score (McCune, 1994; Kessler and Turner, 1997) pro-
Behavioural data (percentage durations) from each vides a sliding scale of stress responses that can be used
horse during the husbandry procedures was combined flexibly to assess welfare in different environments, as used
with percentage change in cortisol from baseline to peak by McCobb et al.(2005) and Dybdall et al. (2007). This latter
in PCA analysis. The three components revealed by the scale however, has not been physiologically validated, and
PCA were combined with the expert panel descriptions McCobb et al. (2005) were not able in their study to cor-
to derive the scale of behavioural indicators of stress that relate BS with urinary cortisol measures in cats housed in
make up the behavioural stress score this work has devel- traditional or enriched shelter environments. We believe
oped. To ensure the final scale of BS reflected increasing our comprehensive approach to combining behaviour and
levels of behavioural and physiological stress, the rela- physiology has resulted in a non-invasive, user-friendly,
tionship between behavioural and physiological changes physiologically validated, behavioural stress score that can
in response to the husbandry procedures was investigated. be used in a variety of environments to measure stress and
Measures of salivary cortisol concentration following the assess welfare in domestic horses.
routine husbandry procedures, and the median BS cal- The work undertaken in developing this behavioural
culated for the same animal were seen to correlate. In stress score also sheds interesting light on the role of
doing this we have successfully combined two differ- stereotypies in assessing stress in horses. Abnormal or
ent behavioural approaches, underpinning them with two stereotypic behaviour was included in all components
well-established physiological measures. extracted by the PCA. ‘No stress’ horses showed an associa-
Whilst other studies have used BS these are often rela- tion with repetitive oral behaviour such as crib-biting, low
tively simple measures of behaviours. For example Minka stress horses exhibited weaving, and both low and medium
et al. (2009) use a number of ‘negative’ behaviours as a stress horses carried out repetitive head movements such
measure of stress, and Maria et al. (2004a) use a greater as head shaking or nodding. It has been suggested that
complexity of behaviours as an indication of less stress. In performance of stereotypic behaviour may serve as a way
addition, these may involve a degree of subjectivity in their of reducing stress levels, or as a way of horses’ pro-
use (Minka et al., 2009; Munsters et al., 2012) or require viding themselves with some sort of control over their
42 T. Young et al. / Applied Animal Behaviour Science 140 (2012) 33–43

environment (Cooper and Albentosa, 2005). This may Ennos, R., 2007. Statistical and Data Handling Skills in Biology, 2nd ed.
Pearson Education Limited, Harlow.
explain the fact that horses perceived as experiencing no
Foreman, J.H., Ferlazzo, A., 1996. Physiological responses to stress in the
or low stress were exhibiting stereotypies. Weaving, which horse. Pferdeheilkunde 12 (4), 401–404.
is indicative of chronic frustration in horses usually associ- Hamlin, M.J., Shearman, J.P., Hopkins, W.G., 2002. Changes in physiological
ated with attempts to gain social contact with other horses parameters in overtrained standardbred racehorses. Equine Vet. J. 34
(4), 383–388.
(Visser et al., 2008), was evident in medium stressed horses. Heleski, C.R., Shelle, A.C., Nielson, B.D., Zanella, A.J., 2002. Influence on
This together with repetitive head movements suggests an housing on weaning horse behaviour and subsequent welfare. Appl.
increased level of frustration experienced by the horses in Anim. Behav. Sci. 78, 291–302.
Houpt, K.A., 1993. Equine stereotypies. Compendium 15 (9), 1265–1271.
the low and medium stress groups. These findings suggest Hydbring, E., Nyman, S., Dahlborn, K., 1996. Changes in plasma cortisol,
that stress assessment that takes a comprehensive view plasma behaviour endorphin, heart rate, haematocrit and plasma pro-
of all behaviour exhibited, rather than a focus on ‘nega- tein concentration in horses during restraint and use of a naso-gastric
tube. Pferdeheilkunde 12 (4), 423–427.
tive’ behaviours, including stereotypies, is desirable. The Kessler, M.R., Turner, D.C., 1997. Stress and adaptation of cats (Felis sil-
behavioural stress score developed here now makes this vestris) housed singly, in pairs and in groups in boarding catteries.
possible for domestic horses. Anim. Welfare 6, 243–254.
Mal, M.W., Friend, T.H., Lay, D.C., Vogelsang, S.G., Jenkins, O.C., 1991.
Behavioural responses of mares to short-term confinement & social
5. Conclusion isolation. Appl. Anim. Behav. Sci. 31, 13–24.
Marc, M., Parvizi, N., Elendorff, F., Kallweit, E., Elsaesser, F., 2000.
Plasma cortisol & ACTH concentrations in the warmblood horse
The scale of behavioural indicators of stress devel- in response to a standardized treadmill exercise test as physio-
oped in this study provides a quick and easy, reliable logical markers for evaluation of training status. J. Anim. Sci. 78,
1936–1946.
method of measuring the stress levels of domestic sta-
Maria, G.A., Escos, J., Alados, C.L., 2004a. Complexity of behavioural
bled horses to assess their welfare. It was developed using sequences and their relationship to stress conditions in chickens (Gal-
both behavioural and physiological measures, so the final lus gallus domesticus): a non-invasive technique to evaluate animal
behavioural scores that make up the scale also provide reli- welfare. Appl. Anim. Behav. Sci. 86, 93–104.
Maria, G.A., Villarroel, M., Chacon, G., Gebresenbet, G., 2004. Scoring sys-
able indices of physiological change in response to stress. tem for evaluating the stress to cattle of commercial loading and
The relationship between behavioural and physiological unloading. Vet. Rec. 154, 818–821.
changes inferred in the scale was further confirmed by the Mason, G.J., 1991. Stereotypies and suffering. Behav. Process. 25, 103–115.
Mason, G., Mendl, M., 1993. Why is there no simple way of measuring
correlation seen between change in salivary cortisol and animal welfare? Anim. Welfare 2 (4), 301–319.
the same horse’s behavioural score. This, therefore, reduces McBride, S.D., Cuddeford, D., 2001. The Putative welfare – reducing
the need to measure various physiological parameters sep- effects of preventing equine stereotypic behaviour. Anim. Welfare 10,
173–189.
arately to validate the use of the scale and so makes it a McCobb, E., Patronek, G., Marder, A., Dinnage, J., Stone, M., 2005. Assess-
valuable, cost-effective tool that could be used by horse ment of stress levels among cats in four animal shelters. J. Anim. Vet.
owners and behavioural scientists alike. Med. Assoc. 226, 548–555.
McCune, S., 1994. Caged cats: avoiding problems and providing solutions.
Newslett. Companion Anim. Study Group 7.
Acknowledgements McDonnell, S.M., Haviland, J.C.S., 1995. Agonistic ethogram of the equid
bachelor band. Appl. Anim. Behav. Sci. 43, 147–188.
McDonnell, S.M., Freeman, D.A., Cymbalaluk, N.F., Scott, H.C., Hinchcliffe,
We are grateful to the horse owners for allowing us to K., Kyle, B., 1999. Behaviour of stabled horses provided continuous
collect behavioural and physiological data from their ani- or intermittent access to drinking water. Am. J. Vet. Res. 60 (11),
1451–1456.
mals. We also appreciated the support given to us by the
McDonnell, S., 2003. A Practical Field Guide to Horse Behaviour. Eclipse
stable staff at Welsh College of Horticulture, Granary Livery, Press, Hong Kong.
Belle Vue Livery, and Merseyside Mounted Police Unit, and Minka, N.S., Ayo, J.O., Sackey, A.K.B., Adelaiye, A.B., 2009. Assessment and
by undergraduate students from the University of Chester scoring of stresses imposed on goats during handling, loading, road
transportation and unloading, and the effect of pre-treatment with
who helped with data collection during the study. We also ascorbic acid. Livest. Sci. 125, 275–282.
thank members of the professional panel who gave up their Minka, N.S., Ayo, J.O., 2008. Assessment of the stresses imposed on adult
time to contribute to this study. Thanks are also extended ostriches (Senahio camelus) during handling, loading, transportation
and unloading. Vet. Rec. 162, 846–851.
to Dr. Rob Coleman for his help with the cortisol assay, and Munsters, C.C.B.M., Visser, K.E.K., van der Broek, J., van Oldruiten-
to Dr. Joanna Hockenhull for her assistance with this study. borgh Oosterbaan, M.M.S., 2012. The influence of challenging objects
and horse rider matching on heart rate, heart rate variability and
behaviour score in riding horses. Vet. J. 192 (1), 75–80.
References Pallant, J., 2004. SPSS Survival Manual. A Step by Step Guide to Using SPSS.
Open University Press, Maidenhead.
Broom, D.M., 1991. Animal welfare: concepts and measurement. J. Anim. Ralston, J.M., Stenhouse, A.M., Stenhouse, N.S., Buck, G.J., Lucks, S.F.,
Sci. 20, 5–19. Reynoldson, J.A., Bolton, J.R., 1988. Cortisol concentrations in blood
Cooper, J.J., Albentosa, M.J., 2005. Behaviour adaptation in the domes- and urine of horses. Aust. Vet. J. 65 (1), 1–5.
tic horse: potential role of apparently abnormal responses including Rietmann, T.R., Stuart, A.E.A., Bernasconi, P., Stauffacher, M., Auer, J.A.,
stereotypic behaviour. Livest. Prod. Sci. 92, 177–182. Weishaupt, M.A., 2004. Assessment of mental stress in warm-
Covalesky, M.E., Russoniello, C.R., Malinowski, K., 1992. Effects of blood horses: heart rate variability in comparison to heart rate
show-jumping performance stress on plasma cortisol and lactate con- and selected behavioural parameters. Appl. Anim. Behav. Sci. 88,
centrations and heart rate and behaviour in horses. J. Equine Vet. Sci. 121–136.
12 (4), 244–250. Schmidt, A., Mostl, E., Wehnert, C., Aurich, J., Muller, J., Aurich, C., 2010.
Dawkins, M.S., 2003. Behaviour as a tool in an assessment of animal wel- Cortisol release and heart rate variability in horses during road trans-
fare. Zoology 106, 383–387. port. Horm. Behav. 57 (2), 209–215.
Dybdall, K., Strasser, R., Katz, T., 2007. Behaviour differences between Seaman, S.C., Davidson, H.P.B., Waran, N.K., 2002. How reliable is temper-
owner surrender and stray domestic cats after entering an animal ament assessment in the domestic horse (Equus caballus)? Appl. Anim.
shelter. Appl. Anim. Behav. Sci. 104, 85–94. Behav. Sci. 78, 175–191.
 T. Young et al. / Applied Animal Behaviour Science 140 (2012) 33–43 43

Smith, T.E., French, J.A., 1997. Psychosocial stress and urinary cortisol Visser, K., Karlas, K., van Deurzen, I., Workel, van Reenen, K., 2010.
excretion in marmoset monkeys (Callithrix kuhli). Physiol. Behav. 62, Experts’ assessment of temperament in sport horse. J. Vet. Behav. 5
225–232. (4), 214–215.
Stegman, G.F., Jones, R.S., 1998. Perioperative plasma cortisol concentra- Von Borell, E., Langbein, J., Despres, G., Hansen, S., Leterrier, C., Marchant-
tion in the horse. J. S. Afr. Vet. Assoc. 69 (4), 131–142. Forde, J., Marchant-Forde, R., Minero, M., Mohr, E., Prunier, A., Valance,
Strand, S.C., Tiefenbacher, S., Haskell, M., Hosmer, T., McDonnell, S.M., D., Veissier, I., 2007. Heart rate variability as a measure of autonomic
Freeman, D.A., 2002. Behaviour and physiological responses of mares regulation of cardiac activity for assessing stress and welfare in farm
to short-term isolation. Appl. Anim. Behav. Sci. 78, 145–157. animals – a review. Physiol. Behav. 92, 293–316.
Toutain, P.L., Lassourd, V., Popot, M.A., Laroute, V., Alunerie, M., Bonnaire, Wemelsfelder, F., Hunter, E.A., Mendl, M.T., Lawence, A.B., 2000. The spon-
Y., 1995. Urinary cortisol excretion in the resting and exercising horse. taneous qualitative assessment of behavioural expressions in pigs:
Equine Vet. J. Suppl. 18, 457–462. first explorations of a novel methology for integrative animal welfare
Visser, E.K., van Reenen, C.G., Hopster, H., Schilder, M.B.H., Knaap, J.H., measurement. Appl. Anim. Behav. Sci. 67, 193–215.
Barneveld, A., Blokhuis, H.J., 2001. Quantifying aspects of young Wemelsfelder, F., Hunter, T.E.A., Mendl, M.T., Lawence, A.B., 2001. Assess-
horses’ temperament: consistency of behavioural variables. Appl. ing the ‘whole animal’: a free choice profiling approach. Anim. Behav.
Anim. Behav. Sci. 74, 241–258. 62, 209–220.
Visser, E.K., Ellis, A.D., van Reenen, C.G., 2008. The effects of two different Winskill, L.C., Waran, N.K., Young, R.J., 1996. The effect of a foraging device
housing conditions on the welfare of young horses stabled for the first (a modified ‘Edinburgh Foodball’) on the behaviour of the stabled
time. Appl. Anim. Behav. Sci. 114, 521–533. horse. Appl. Anim. Behav. Sci. 48, 25–35.