Heart rate and heart rate variability during a novel object test and a ...

Physiology & Behavior 76 (2002) 289–296
Heart rate and heart rate variability during a novel object test
and a handling test in young horses
E.K. Visser a,b, *, C.G. van Reenen a , J.T.N. van der Werf a , M.B.H. Schilder c , J.H. Knaap d ,
A. Barneveld b , H.J. Blokhuis a
a Division of Animal Sciences, Institute for Animal Science and Health, P.O. Box 65, NL-8200 AB Lelystad, The Netherlands
b Department of Equine Sciences, Faculty of Veterinary Medicine, Utrecht University, P.O. Box 80.163, NL-3508 TD Utrecht, The Netherlands
c Department of Animals and Society, Faculty of Veterinary Medicine, Utrecht University, Yalelaan 17, NL-3584 CZ Utrecht, The Netherlands
d Research Institute for Animal Husbandry, P.O. Box 2176, 8203 AD Lelystad, The Netherlands
Received 28 November 2001; received in revised form 21 February 2002; accepted 18 March 2002
Abstract
Forty-one Dutch Warmblood immature horses were used in a study to quantify temperamental traits on the basis of heart rate (HR) and
heart rate variability (HRV) measures. Half of the horses received additional training from the age of 5 months onwards; the other half did
not. Horses were tested at 9, 10, 21 and 22 months of age in a novel object and a handling test. During the tests, mean HR and two heart
variability indices, e.g. standard deviation of beat-to-beat intervals (SDRR) and root mean square of successive beat-to-beat differences
(rMSSD), were calculated and expressed as response values to baseline measures. In both tests, horses showed at all ages a significant
increase in mean HR and decrease in HRV measures, which suggests a marked shift of the balance of the autonomic nervous system towards
a sympathetic dominance. In the novel object test, this shift was more pronounced in horses that had not been trained. Furthermore, statistical
analysis showed that the increase in mean HR could not be entirely explained by the physical activity. The additional increase in HR, the
nonmotor HR, was more pronounced in the untrained horses compared to the trained. Hence, it is suggested that this nonmotor HR might be
due to the level of emotionality. HR variables showed consistency between years, as well as within the second year. These tests bring about a
HR response in horses, part of which may indicate a higher level of emotionality; and horses show individual consistency of these HR
variables over ages. Therefore, it is concluded that mean HR and HRV measures used with these tests quantify certain aspects of a horse’s
temperament. D 2002 Elsevier Science Inc. All rights reserved.
Keywords: Heart rate; Heart rate variability; Consistency; Behavioral tests; Temperament; Emotionality; Horses
1. Introduction
Knowledge about temperament of horses may have very
important implications for the breeding, housing and management
of the horse. It may contribute to breeding or
selecting the winning athlete in sports or to select the right
horse to match with a recreational rider. These applications
emphasize the need for quantification of the horse’s temperamental
characteristics. Temperament has been defined
as the individual’s basic stance towards continuing changes
and challenges in its environment [1]. To select the horse
* Corresponding author. Division of Animal Sciences, Institute for
Animal Science and Health, P.O. Box 65, NL-8200 AB Lelystad, The
Netherlands. Tel.: +31-320-238-200; fax: +31-320-238-094.
E-mail address: e.k.visser@id.wag-ur.nl (E.K. Visser).
with the right temperament for a specific goal, one has to
review several aspects underlying the overall temperament.
We presume that most of these temperamental aspects might
be exhibited in situations where the individual shows
emotional responses towards novel stimuli and in handling
situations where the animal shows reactivity to the handler.
Over the last decade, a growing number of studies in
many different species have focused on individual differences
in animals in responses to challenges in behavioral
tests (e.g., Refs. [2–7]). These individual differences have
been related to different coping strategies or to differences in
temperamental traits. The most commonly used variables to
quantify temperamental traits are behavioral variables.
However, recently, it was shown that heart rate (HR) and
heart rate variability (HRV) in rats also are useful to
differentiate between individuals [8]. Hence, they seem
0031-9384/02/$ – see front matter D 2002 Elsevier Science Inc. All rights reserved.
PII: S0031-9384(02)00698-4

290
E.K. Visser et al. / Physiology & Behavior 76 (2002) 289–296
relevant additional variables to quantify individual differences
in temperamental aspects.
HR represents the net effect of the parasympathetic nerves
that slow it down and the sympathetic nerves that accelerate
it. In resting conditions, both parts of the autonomic nervous
system are thought to be tonically active. Different stressors
can induce a shift of the autonomic balance towards either a
sympathetic or a parasympathetic dominance [9–12]. A
decreased HRV reflects a shift of the autonomic balance
towards a more sympathetic dominance [8,13,14]. Several
studies on rodents have shown that stressful conditions,
depending on their nature, can result in a decrease in HRV
and hence low levels of parasympathetic nerve activity [8].
For example, social defeat in rats produces lower HRV
measures compared to baseline recordings, whereas other
stressors including restraint, swimming and shock-probe test
do not [15]. In a human study by Friedman and Thayer [16],
it was shown that patients with chronic anxiety (panic
disorder) exhibit lower levels of HRV and higher levels of
mean HR compared to their controls. The parasympathetic
branch of the autonomic nervous system seems to be
associated with adaptive responsivity to the environment
[17]. Individuals with a higher parasympathetic activity
would be more exploratory and adaptive to environmental
demands. In another study, Porges [18] showed that parasympathetic
activity is suppressed during autonomic and
behavioral responses to stress. Thus, responses of the autonomic
nervous system on challenging situations differentiate
between individuals. These differences could therefore
reflect differences in temperament. It was shown that horses,
when confronted with a challenging situation, exhibited
individually different behavioral responses, and this was
suggested to be linked with differences in temperament [7].
We hypothesize that responses of the autonomic nervous
system may also be useful in differentiating between individual
horses in terms of temperament.
In the present study, we aimed to investigate whether
responses of HR and HRV to challenging situations could
differentiate between individual horses’ temperament. It is
hypothesized that a confrontation with a novel object causes
an emotional state, possibly ‘fear’, which is reflected in an
additional increase in mean HR and a decrease in HRV.
Since HR (and HRV) is also affected by physical activity
[8], physical activity was explicitly considered in the statistical
analysis in an attempt to distinguish between variation
in HR due to physical activity and variation in
nonmotor HR [19]. A second test involving human handling
was used to investigate whether individual horses responded
differently to situations of challenge with or without the
support of a human.
2. Materials and methods
All procedures involving animal handling and testing
were approved by the Animal Care and Use Committee of
the Institute for Animal Science and Health (ID-Lelystad) in
Lelystad, the Netherlands.
2.1. Animals, housing, handling and management
A total of 41 Dutch warmblood horses, 26 colts and 15
fillies were used in this study. Experiments were carried out
at the Research Station for Animal Husbandry in Lelystad.
Horses were simultaneously used in a project aiming at the
development of scientific criteria for the selection and
effective and injury-free training of show jumpers. To
evaluate the effects of training at a young age, half of the
horses were trained from the age of 6 months, and the other
half were not. Horses were blocked by pedigree, date of
birth, sex and free-jumping performance at 5 months of age,
and they were randomly allocated to one of the two
treatment groups: training (n = 21: 13 colts, 8 fillies) and
nontraining (n = 20: 13 colts and 7 fillies). Horses were held
in groups in half-open, straw-bedded stables in winter and
on pasture in summer. At stable, horses were fed concentrates
in the morning and in the afternoon. Before noon,
horses received additional straw and grass silage. Water was
available ad libitum. (For more details, see Ref. [7].) Due to
health problems, three horses were excluded from this study
before the start of the second winter period.
2.2. Testing procedures
During two winter periods, a novel object test and a
handling test were conducted. Both tests were carried out
four times, i.e. when the horses were approximately 9, 10,
21 and 22 months old. The experimental design used in this
study is described in detail by Visser et al. [7].
Briefly, the novel object test was carried out in an indoor
arena of 18 21 m. After a 2-min waiting period in a socalled
starting box, the horse was released into the arena. Two
minutes after the horse entered the arena, the novel object (an
open blue and white umbrella) was lowered from the ceiling,
from approximately 6 m height. The umbrella stood straight
up for another 5 min allowing the horse to investigate it.
Then, the horse was caught by the handlers and led back to its
stable. The handling test comprised of a familiar handler
leading the horse to a so-called ‘bridge’. This bridge consisted
of four concrete plates (four plates of 2 1 m), creating
a bridge 2 m wide and 4 m long. The top level of which was
elevated 15 cm above the ground. While approaching the
bridge, the handler did not pull, touch or talk to the horse. If
the horse stopped walking, the handler walked on until the
rope was slightly tensioned. If the horse continued the
approach, the handler moved on. Tough resistance, like
pulling backwards, rearing or walking along the sides, was
followed by another trial. Each horse had a maximum of three
trials per test to cross the bridge. The starting point of every
trial was approximately 12 m in front of the bridge, the horse
facing the bridge. After the horse had crossed the bridge
or reached the maximum number of trials, it was led back to

E.K. Visser et al. / Physiology & Behavior 76 (2002) 289–296 291
Table 1
HR and HRV indices (means ± S.E.M.) for the baseline values and the
corresponding novel object test values at 9, 10, 21 and 22 months of age
(n = 41, 41, 38 and 38, respectively)
9 Months 10 Months 21 Months 22 Months
HR (bpm)
Baseline 53.7 ± 1.4 a 54.0 ± 2.1 a 47.3 ± 1.1 a 47.6 ± 1.1 a
Test 159.5 ± 4.4 b 142.6 ± 4.6 b 129.4 ± 5.6 b 113.9 ± 5.4 b
SDRR (ms)
Baseline 131.1 ± 9.5 a 133.8 ± 7.0 a 286.9 ± 23.0 a 208.8 ± 15.5 a
Test 91.4 ± 7.0 b 106.6 ± 7.1 b 131.2 ± 12.0 b 132.3 ± 9.1 b
rMSSD (ms)
Baseline 46.3 ± 4.8 a 42.0 ± 2.8 a 61.3 ± 4.7 a 56.1 ± 6.1 a
Test 15.8 ± 1.9 b 18.6 ± 1.7 b 22.8 ± 2.9 b 28.2 ± 2.8 b
a,b Means of baseline and corresponding test value with different letter per
HR variable per age differ significantly ( P < .05).
the stable by a handler. This latter procedure was executed at
9 and 10 months of age. During the second year
(21 and 22 months of age), the following adjustment was
made to ensure enough variation between individuals.
Yellow wooden plates were attached on top of the concrete
plates to avoid habituation. Three horses had to be excluded
from this test at the age of 22 months due to dangerous
behavior towards the experimenters during testing.
2.3. Data acquisition and processing
HR was recorded with Polar Vantage (Polar Electro OY,
Kemple, Finland), which consisted of an electrode belt with
built-in transmitter and a wristwatch receiver. Warm water
and methylated spirit were used to optimize the contact
between electrode and skin. The receiver had a memory
function and stored data during the tests from the transmitter.
Afterwards, data were downloaded via a Polar Interface
(Polar Electro OY) to a PC. Before each test, a baseline
recording of 2 min was made in the home environment
(half-open stable). Recordings in the novel object test were
exactly 5 min for each horse, in the handling test, between 2
and 9 min per horse.
The following HR variables were quantified: (1) mean
HR (bpm) (2) the standard deviation of beat-to-beat intervals
(SDRR) and (3) the root mean square of successive
beat-to-beat differences (rMSSD, ms) [13]. Generally speaking,
reductions in the values of HRV indices (SDRR and
rMSSD) reflect a shift of the autonomic balance towards a
more sympathetic dominance, while increased values of
these indices indicate a shift towards a more parasympathetic
dominance. The SDRR estimates the overall HRV and
therefore includes the contribution of both branches of the
autonomic nervous system to HR variations: It measures the
state of balance between sympathetic and parasympathetic
activities of the heart. rMSSD focuses on high-frequency,
short-term variations of HR, which are mainly due to the
activity of the parasympathetic nervous system [13,14,20].
Mean HR and HRV measures were performed after systematic
removal of artefacts on the basis of visual inspection.
Approximately 8% of all HR recordings was not used
because of excessive artefacts.
Behavior during both tests was videotaped with a camcorder
and was analyzed using the Observer software
program (Noldus Information Technology, Wageningen,
The Netherlands). Several behavioral variables were scored
(see Ref. [7]). In the present study, we used the percentage
of physical activity of the total test time as a measure for
physical activity. Since we were able to control the level of
physical activity in the handling test, horses were led to the
bridge at a low speed; we did not use a physical activity
variable in the handling test to correct mean HR for activity.
2.4. Statistical analysis
To obtain homogeneous variances, observations for
SDDR and rMSSD were log-transformed prior to analysis.
All HR variables considered in this present paper were
expressed as response values to baseline, i.e. for each
variable, the difference was calculated between the value
during the test and the value during baseline. In view of
differential durations of HR recordings in the handling test,
we used regression analyses within ages to examine whether
HR variables were affected by the length of the test. The
regression model used measures of HR and HRV (response
value relative to baseline) as dependent variables and length
of the test as explanatory variable. Factors for housing
groups, sex and training were introduced as fixed effects.
None of the HR variables was significantly affected by the
length of the test, which was therefore not considered as a
source of variation in subsequent analyses.
The estimation of main effects of sex, training and age
involved two steps. First, variables (response value relative
to baseline) were analyzed per age with an analysis of
variance model with factors for housing groups (five
levels), sex (two levels) and training (two levels). Tests
Table 2
Response values (means ± S.E.M.) for the HR and HRV indices (difference
between test values and corresponding baseline values) for the trained
(n = 21) and untrained horses (n = 20) in the novel object test at 9, 10, 21
and 22 months of age (n = 41, 41, 38 and 38, respectively)
9 Months 10 Months 21 Months 22 Months
HR (bpm)
Training 95.8 ± 7.4 a 78.9 ± 5.3 a 64.8 ± 6.4 a 44.5 ± 5.3 a
Nontraining 116.3 ± 6.4 b 99.8 ± 5.4 b 100.2 ± 6.6 b 90.5 ± 6.3 b
SDRR (ms)
Training 29.3 ± 19.8 a 8.2 ± 13.5 a 78.1 ± 34.4 a 17.9 ± 25.2 a
Nontraining 59.5 ± 17.1 b 54.3 ± 13.9 b 248.9 ± 35.4 b 162.1 ± 29.8 b
rMSSD (ms)
Training 33.4 ± 8.3 a 16.2 ± 4.4 a 34.2 ± 6.7 a 5.5 ± 9.1 a
Nontraining 26.3 ± 7.2 a 32.3 ± 4.5 b 43.3 ± 7.2 b 49.3 ± 10.5 b
a,b Means of trained and untrained horses with different letter per HR
variable per age differ significantly ( P < .05).

292
E.K. Visser et al. / Physiology & Behavior 76 (2002) 289–296
of Models A and B, expressed as a percentage of the
residual variance of Model A. The residuals of Model B
were saved as a new variable that effectively represents HR
corrected for physical activity.
Consistency of variables was assessed by calculating
Spearman correlations between residuals e = y m from
the aforementioned analysis for each pair of ages. These
residuals can be regarded as observations corrected for
possible effects of housing group, sex and training. Hence,
correlations can be interpreted as pooled correlations within
combinations of housing group, sex and training. All
calculations were performed with Genstat [23]. Statistical
significance was set at P < .05.
Fig. 1. Response values (means ± S.E.M.) for the HR and HRV indices
(difference between test values and corresponding baseline values) in the
novel object test at 9, 10, 21 and 22 months of age (n = 41, 41, 38 and 38,
respectively). The HRV indices SDRR and rMSSD have been logtransformed.
Means of a single HR variable with different letters differ
significantly between ages ( P < .05).
for main effects and interactions were based on the common
F test. Residuals e = y m, where y denotes an observation
and m the corresponding estimated mean were saved and
used as variables corrected for fixed effects. Second, differences
between all ages were also analyzed as repeated
measurements with a mixed analysis of variance model
with a general covariance matrix for the observations at the
different ages. Dispersion parameters were estimated by
restricted maximum likelihood (REML) [21]. Tests for main
effects for housing groups and main effects, and interactions
between sex, training and time, were performed with the
Wald test [22].
To assess the amount of variance in mean HR as
explained by physical activity, at each age, two models for
mean HR were compared: the model with factors sex and
training (Model A) and the model with an additional
covariate for the physical activity, including interaction
terms (Model B). The percentage of variance explained by
physical activity was defined as the reduction of the residual
variance, i.e. the difference between the residual variances
3. Results
3.1. Effects of the novel object test on HR variables
Mean HR and the HRV indices SDRR and rMSSD all
differed significantly between baseline and test values at all
ages (see Table 1). All HR variables showed a more pronounced
response for the untrained horses, apart from
rMSSD at 9 months of age (see Table 2). There was no
significant effect of sex on any of the HR variables at any age.
Fig. 1 shows the response values, i.e. differences between
baseline and test values, for the different HR variables per
age. The HR response decreased over ages, the HRV indices
did not exhibit an apparent decrease over ages.
3.2. Effect of physical activity on HR in the novel object test
The percentage of HR, expressed as a response value,
explained by physical activity decreased within years
(9 months: 41.1; 10 months: 23.3: 21 months: 42.3; and
22 months: 29.9). The trained and untrained horses in the
novel object test tended to differ in their physical activity at
9 months [ F(1,38) = 3.3, P < .1], differed significantly at
10 months of age [ F(1,38) = 7.2, P < .05] but did not differ
at 21 and 22 months (Table 3). Meanwhile, the untrained
horses displayed a significantly higher mean HR than the
trained horses at all ages. The predicted nonmotor HR
Table 3
Means ( ± S.E.M) of physical activity, response values for mean HR and nonmotor HR for the trained (n = 21) and untrained horses (n = 20) in the novel object
test at 9, 10, 21 and 22 months of age (n = 41, 41, 38 and 38, respectively)
Model Dependent variable 9 Months 10 Months 21 Months 22 Months
A Physical activity Training 69.6 ± 3.3 a 63.8 ± 3.3 a 55.7 ± 4.4 a 55.8 ± 4.1 a
A Physical activity Nontraining 78.4 ± 3.4 b 75.6 ± 3.5 c 62.7 ± 2.7 a 60.8 ± 3.1 a
A HR Training 95.8 ± 7.4 a 78.9 ± 5.3 a 64.8 ± 6.4 a 44.5 ± 5.3 a
A HR Nontraining 116.3 ± 6.4 c 99.8 ± 5.4 c 100.2 ± 6.6 c 90.5 ± 6.3 c
B Nonmotor HR Training 97.6 ± 5.8 a 82.9 ± 4.9 a 70.5 ± 5.3 a 43.9 ± 4.9 a
B Nonmotor HR Nontraining 107.7 ± 6.1 b 97.5 ± 6.0 c 95.7 ± 8.1 c 86.7 ± 6.0 c
Means of physical activity and HR were predicted from Model A (model with factors housing group, sex and training), means of nonmotor HR were predicted
from Model B (same as Model A with an additional covariate for the physical activity, see Section 2.4).
a,b,c Means of trained and untrained horses with different letter per variable per age differ significantly ( a,c P < .05) or tend to differ significantly ( a,b P < .1).

E.K. Visser et al. / Physiology & Behavior 76 (2002) 289–296 293
Table 4
HR and HRV indices (means ± S.E.M.) for the baseline values and the
corresponding handling test values at 9, 10, 21 and 22 months of age
(n = 41, 41, 38 and 35, respectively)
9 Months 10 Months 21 Months 22 Months
HR (bpm)
Baseline 54.1 ± 1.7 a 51.0 ± 1.0 a 48.5 ± 1.7 a 47.5 ± 0.9 a
Test 112.5 ± 2.8 b 105.5 ± 2.9 b 96.2 ± 3.0 b 90.5 ± 3.2 b
SDRR (ms)
Baseline 143.2 ± 11.1 a 131.5 ± 8.9 a 239.9 ± 16.8 a 193.1 ± 9.4 a
Test 90.9 ± 6.4 b 102.3 ± 4.7 b 100.5 ± 6.0 b 89.2 ± 9.0 b
RMSSD (ms)
Baseline 49.2 ± 8.9 a 48.9 ± 5.1 a 57.8 ± 4.7 a 50.6 ± 2.7 a
Test 24.6 ± 2.1 b 24.2 ± 1.8 b 24.4 ± 2.2 b 26.6 ± 3.0 b
a,b Means of baseline and corresponding test value with different letter per
HR variable per age differ significantly ( P < .05).
(Model B) was significantly higher for the untrained horses
than for the trained horses at 10 [ F(1,36) = 4.1, P < .05], 21
[ F(1,30) = 15.1, P < .01] and 22 [ F(1,31) = 39.5, P < .001]
months of age and tended to be higher at 9 months of age
[ F(1,32) = 3.4, P < .1].
3.3. Effects of the handling test on HR variables
Table 4 shows that the baseline values of the mean HR
(HR) and the HRV indices (SDRR and rMSSD) all differed
between baseline and test values for all ages. Although the
untrained horses had more pronounced HR and HRV
responses, these were not, except one, significantly different
from the trained horses (see Table 5). When horses got
older, the mean HR response decreased significantly
( P < .05). The response of SDRR differed between years
(9 and 10 months vs. 21 and 22 months of age) and the
response of rMSSD was not significantly different between
ages (see Fig. 2). There were no interaction effects among
the sexes, training and ages.
Table 5
Response values (means ± S.E.M.) for the HR and HRV indices (difference
between test values and corresponding baseline values) for the trained
(n = 21) and untrained horses (n = 20) in the handling test at 9, 10, 21 and 22
months of age (n = 41, 41, 38 and 35, respectively)
9 Months 10 Months 21 Months 22 Months
HR (bpm)
Training 57.3 ± 4.6 a 49.7 ± 3.9 a 43.0 ± 4.4 a 39.1 ± 4.0 a
Nontraining 60.4 ± 4.9 a 59.0 ± 4.1 a 54.5 ± 4.6 a 46.6 ± 5.0 a
SDRR (ms)
Training 28.8 ± 18.8 a 27.9 ± 13.0 a 121.9 ± 24.0 a 91.7 ± 16.4 a
Nontraining 75.2 ± 19.9 b 29.9 ± 13.8 a 169.5 ± 25.4 a 122.2 ± 20.6 a
RMSSD (ms)
Training 26.1 ± 12.6 a 24.9 ± 7.6 a 29.0 ± 8.3 a 22.4 ± 4.8 a
Nontraining 18.8 ± 13.3 a 25.5 ± 8.0 a 37.2 ± 8.8 a 28.2 ± 6.0 a
a,b Means of trained and untrained horses with different letter per HR
variable per age differ significantly ( P < .05).
Fig. 2. Response values for the HR and HRV indices (difference between
test value and corresponding baseline value) in the handling test at 9, 10, 21
and 22 months of age (n = 41, 41, 38 and 35, respectively). The HRV
indices SDRR and rMSSD have been log-transformed. Means of a single
HR variable with different letters differ significantly between ages
( P < .05).
3.4. Consistency over ages in the novel object test and
handling test
Table 6 shows the Spearman rank correlations between
ages for the responses of the HR variables in the novel
object test. All responses of HR variables and the residual of
HR (after correction for physical activity) were consistent
between years (9 and 21 months of age) and within the
second year. In addition, in the handling test, horses showed
significantly consistent responses between years (9 and 21
Table 6
Spearman rank correlations between responses of HR variables (difference
between test values and corresponding baseline values) in the novel object
test at 9, 10, 21 and 22 months of age (n = 41, 41, 38 and 38, respectively)
HR, SDRR, rMSSD and residual
HR at different ages (months)
Variable Age (months) 9 10 21
HR (bpm) 10 .51
21 .59 .50
22 .36 .27 .43
SDRR (ms) 10 .19
21 .51 .24
22 .33 .24 .46
rMSSD (ms) 10 .33
21 .49 .30
22 .27 .11 .40
Residual HR 10 .06
21 .65 .26
22 .41 .17 .50
Residual HR is the residual that is saved from Model B (model with factors
housing group, sex and training with an additional covariate for the physical
activity) and is used as a new variable, which relates to the predicted
nonmotor HR. Correlations in bold are statistically significant ( P < .05),
those in bold and italic show a tendency (.05 < P < .1).

294
E.K. Visser et al. / Physiology & Behavior 76 (2002) 289–296
Table 7
Spearman rank correlations between responses of HR variables (difference
between test values and corresponding baseline values) in the handling test
at 9, 10, 21 and 22 months of age (n = 41, 41, 38 and 35, respectively)
HR, SD and rMSSD at different
ages (months)
Variable Age (months) 9 10 21
HR (bpm) 10 .61
21 .21 .38
22 .14 .39 .53
SDRR (ms) 10 .22
21 .11 .46
22 .16 .45 .70
rMSSD (ms) 10 .34
21 .49 .27
22 .00 .31 .40
Correlations in bold are statistically significant ( P < .05), those in bold and
italic show a tendency (.05 < P < .1).
months or 10 and 22 months of age) and within the second
year (see Table 7).
4. Discussion
The results show that the challenge of the novel object
test induced a physiological state characterized by an
increase in mean HR and a decrease in HRV. In addition,
for the handling test, in which a challenge was faced with
the support of a familiar handler, a similar pattern was
found. HR variables showed significant positive correlations
within tests both between years and within the
second year.
4.1. The response of the autonomic nervous system to
challenge tests
At all ages, in both tests, a significant increase in mean
HR was found together with a significant decrease in the
HRV indices SDRR and rMSSD. This indicates a shift of the
autonomic balance towards a sympathetic dominance [8].
Because both the SDRR and the rMSSD decreased significantly,
it is suggested that the shift towards the sympathetic
dominance results from the lack of a sufficient parasympathetic
counteraction to sympathetic activation [12]. A
similar response was found for rats facing a social defeat
challenge [15]. Furthermore, studies showed that fast alteration
in mean HR due to sudden excitement or fear, in horses
up to 110 bpm, are almost entirely attributed to decreases in
parasympathetic nerve activity, while slow HR alterations
can be mediated by both sympathetic and parasympathetic
nerve activity [18,24–26]. This also seems the case in this
study. A high degree of parasympathetic control allows for
enhanced responsivity of the HR to environmental demands
[18] and a low degree of parasympathetic tone is found with
panic [16,17]. In the present study, individual horses differed
in their HR responses to both challenge tests and
therefore may also have differed in their adaptiveness to
environmental demands. Horses with a low degree of
parasympathetic nerve activity are thought to be more
emotionally stressed compared to individuals with a higher
degree of parasympathetic nerve activity. Hence, it is
suggested that HR responses may relate to aspects of
temperament, especially emotionality. Aspects of emotionality,
like fear and frustration, emerging during isolation or
other stressors, may be relevant for horse owners and
trainers to be able to predict horses’ responses to environmental
demands. Subsequently, a horse trainer can adjust
training methods accordingly and a horse owner can select a
specific horse for a specific goal.
4.2. Effect of physical activity on mean HR
Behavioral efforts of coping with the challenge influence
the magnitude of the cardiovascular responses [27]. For
example, the physical activity displayed by the individual
might influence its HR [8]. The percentage of variation in
mean HR in the novel object test that was accounted for by
physical activity was 41%, 23%, 42% and 30% at 9, 10, 21
and 22 months of age, respectively. This implies that the
remainder, or the nonmotor HR [19], was explained by
variation in a factor other than physical activity, e.g.
emotional reactivity.
4.3. Effect of training
Several horse studies showed that a mean HR was
significantly higher for individuals that were characterized
as more emotional [28,29]. In the novel object test of the
present study, the response of the untrained horses was more
pronounced compared to the response of the trained horses.
Thus, there was a larger increase in HR and a larger
decrease in the HRV indices. Hence, the shift in the balance
of the autonomic nervous system towards a sympathetic
dominance, presumably related to a poor vagal antagonism
to sympathetic activation, was stronger for the untrained
horses compared to the trained horses. The more pronounced
increase in HR in the untrained horses could not
be explained by a higher level of physical activity: While
the level of physical activity hardly differed, the overall HR
and the nonmotor HR were significantly higher for the
untrained horses than for the trained horses, suggesting a
higher level of emotional activation. This difference in
responsivity between trained and untrained horses may be
explained by the fact that the trained horses were exposed
more often to changing environmental demands during their
training sessions. The difference in emotionality between
the trained and untrained horses was evident in the handling
test, as suggested by similar values of HR variables in the
two groups. We speculate that the social support of the
handler has buffered possible differences between these
groups in the handling test [30,31].

E.K. Visser et al. / Physiology & Behavior 76 (2002) 289–296 295
4.4. Consistency of variables within tests
Quantification of potential temperamental traits implies
developing tests in which animals show consistent responses
[2,7,32–36]. Whether the use of variables indicating the
response of the autonomic nervous system would fulfil this
requirement depends on whether these HR variables show
consistency over time within individuals. The principal
findings of the present study were that mean HR and the
HRV indices SDRR and rMSSD showed significant positive
correlations within individuals across ages. Although correlations
were most robust within the second year, all HR
variables showed a significant correlation between the first
and the second year. These findings suggest that mean HR
and HRV are reliable measures and can be used in the
quantification of temperamental traits of horses. It was shown
that behavioral responses of these horses to these tests were
consistent both within the first and within the second year,
but not between years [7]. It was suggested that the lack of
consistencies in behavioral variables between years could be
explained by maturational effects including puberty [7,29].
Since in the present study, we did find significant correlations
between years for the HR variables, we suggest that the effect
of maturation on the variables measured is different for the
HR and for the behavioral responses of the horses.
In a parallel study with adult horses, it was found that HR
variables measured in a novel object test and in a handling
test showed significant correlations with temperamental traits
assessed by riders [37]. In this latter study, it was shown for
example that horses with a low HRV (rMSSD) in the
handling test were characterized by the riders as highly
responsive to changes in the environment. The reliability of
the HR variables in the present study and the correlation of
the HR variables with temperament assessed by riders in the
study of Visser et al. [37] suggest that HR variables are highly
suitable in objectively quantifying aspects of temperament.
5. Conclusion
Horses showed pronounced responses in HR and HRV
measures when confronted with changes in the environment.
The additional increase in mean HR when exposed to a
novel object was related to emotionality rather than physical
activity. This nonmotor HR showed to be higher for the
untrained horses compared to the trained horses. All HR
variables used in this study were consistent over years and
within the second year. Our results strongly suggest that HR
variables are useful in differentiating between individuals
and quantifying aspects of temperament.
Acknowledgments
We thank the personnel of the experimental farm ‘de
Waiboerhoeve’ for taking care of the horses and their
practical assistance. Furthermore, we would like to thank the
following students for their participation in the project:
Godelieve Kranendonk, Elisa Kroese, Tine Geurts and
Dyane van Huisstede. Moreover, we thank Bas Engel and
Willem Buist for the help with the statistical analysis.
References
[1] Mason WA. Animal learning: experience, life modes and cognitive
style. Verh Dtsch Zool Ges 1984;77:45– 56.
[2] Manteca X, Deag JM. Individual differences in temperament of domestic
animals: a review of methodology. Anim Welf 1993;2(3):
247–68.
[3] Hemsworth PH, Brand A, Willems P. The behavioural response of
sows to the presence of human beings and its relation to productivity.
Livest Prod Sci 1981;8(1):67–74.
[4] Hausberger M, Scolan NL, Bruderer C. Le temperament du cheval:
facteurs en jeu et implications practiques. J Rech Equine 1998;24:
159–70.
[5] Boissy A, Bouissou MF. Assessment of individual differences in behavioural
reactions of heifers exposed to various fear-eliciting situations.
Appl Anim Behav Sci 1995;46(1 – 2):17–31.
[6] Boissy A. Fear and fearfulness in animals. Q Rev Biol 1995;70(2):
165–91.
[7] Visser EK, Van Reenen CG, Hopster H, Schilder MBH, Knaap JH,
Barneveld A, Blokhuis HJ. Quantifying aspects of young horses’
temperament: consistency of behavioural variables. Appl Anim Behav
Sci 2001;74(4):241–58.
[8] Sgoifo A, de Boer SF, Westenbroek C, Maes FW, Beldhuis H, Suzuki
T, Koolhaas JM. Incidence of arrhythmias and heart rate variability in
wild-type rats exposed to social stress. Am J Physiol 1997;273(4 Pt 2):
H1754–60.
[9] Hagan JJ, Bohus B. The effects of endorphins on cardiac responses
during an emotional stress. Physiol Behav 1983;31(5):607–14.
[10] Korte SM, Koolhaas JM, Schuurman T, Traber J, Bohus B. Anxiolytics
and stress-induced behavioural and cardiac responses: a study of
diazepam and ipsapirone (TVX Q 7821). Eur J Pharmacol 1990;
179:393–401.
[11] Obrist PA, Gaebelein CJ, Teller ES, Langer AW, Grignolo A, Light
KC, McCubbin JA. The relationship among heart rate, caratid dP/dt,
and blood pressure in humans as a function of the type of stress.
Psychophysiology 1978;15(2):102– 15.
[12] Sgoifo A, Koolhaas JM, Musso E, De Boer SF. Different sympathovagal
modulation of heart rate during social and nonsocial stress
episodes in wild-type rats. Physiol Behav 1999;67(5):733– 8.
[13] Stein PK, Bosner MS, Kleiger RE, Conger BM. Heart rate variability: a
measure of cardiac autonomic tone. Am Heart J 1994;127(5):1376–81.
[14] Task Force of the European Society of Cardiology and the North
American Society of Pacing and Electrophysiology. Heart rate variability:
standards of measurement, physiological interpretation and
clinical use. Circulation 1996;93(5):1043– 65.
[15] Sgoifo A, Pozzato C, Costoli T, Manghi M, Stilli D, Ferrari PF,
Ceresini G, Musso E. Cardiac autonomic responses to intermittent
social conflict in rats. Physiol Behav 2001;73(3):343– 9.
[16] Friedman BH, Thayer JF. Autonomic balance revisited: panic anxiety
and heart rate variability. J Psychosom Res 1998;44(1):133–51.
[17] Friedman BH, Thayer JF. Anxiety and autonomic flexibility: a cardiovascular
approach. Biol Psychol 1998;47(3):243– 63.
[18] Porges SW. Vagal tone. In: Garber J, editor. The development of
emotion regulation and dysregulation. Cambridge: Cambridge Univ.
Press, 1991. pp. 111–28.
[19] Baldock NM, Sibly RM. Effects of handling and transportation on the
heart rate and behaviour of sheep. Appl Anim Behav Sci 1990;28:
15–39.

296
E.K. Visser et al. / Physiology & Behavior 76 (2002) 289–296
[20] Sgoifo A, De Boer SF, Buwalda B, Korte Bouws G, Tuma J, Bohus B,
Zaagsma J, Koolhaas JM. Vulnerability to arrhythmias during social
stress in rats with different sympathovagal balance. Am J Physiol
1998;275(2 Pt 2):H460–6.
[21] Patterson HD, Thompson R. Recovery of inter-block information
when block sizes are unequal. Biometrika 1971;58:545–54.
[22] Rao CR. Linear statistical inference and its application. New York:
Wiley, 1973.
[23] Genstat. Genstat 5 reference manual. Oxford: Clarendon Press, 1993.
[24] Hamlin RL, Klepinger WL, Gilpin KW, Smith CR. Autonomic control
of heart rate in the horse. Am J Physiol 1972;222(4):976–8.
[25] Hodgson DR, Rose RJ. The athletic horse: principles and practice of
equine sports medicine. Philadelphia: Saunders, 1994.
[26] Akselrod S. Components of heart rate variability. In: Malik M, Camm
AJ, editors. Heart rate variability. Armonk: Futura Publishing, 1995.
pp. 147–63.
[27] Herd JA. Cardiovascular response to stress. Physiol Rev 1991;71(1):
305–30.
[28] McCann JS, Heird JC, Bell RW, Lutherer LO. Normal and more
highly reactive horses: I. Heart rate, respiration rate and behavioral
observations. Appl Anim Behav Sci 1988;19(3– 4):201–14.
[29] Jezierski T, Jaworski Z, Gorecka A. Effects of handling on behaviour
and heart rate in konik horses: comparison of stable and forest reared
youngstock. Appl Anim Behav Sci 1999;62(1):1–11.
[30] Cohen S, Wills TA. Stress, social support, and the buffering hypothesis.
Psychol Bull 1985;98(2):310– 57.
[31] Christenfeld N, Gerin W. Social support and cardiovascular reactivity.
Biomed Pharmacother 2000;54(5):251– 7.
[32] Hessing MJC, Hagelso AM, Beek JAMv, Wiepkema PR, Schouten
WGP, Krukow R, Van Beek JAM. Individual behavioural characteristics
in pigs. Appl Anim Behav Sci 1993;37(4):285–95.
[33] Spoolder HAM, Burbidge JA, Lawrence AB, Simmins PH, Edwards
SA. Individual behavioural differences in pigs: intra- and inter-test
consistency. Appl Anim Behav Sci 1996;49(2):185–98.
[34] Van Reenen CG, O’Connell NE, van der Werf JTN, Blokhuis HJ.
Responses of dairy calves to exposure to a novel object: individual
consistency and relationships between behavioural and physiological
variables. Biology of animal stress. CA: Davis, 1998 (abstract published
on internet).
[35] Van Reenen CG, Heutinck LFM, van der Werf JTN, Blokhuis HJ.
Long-term consistency in behavioural and physiological responses of
heifers to a novel environment test. 29th International Congress of the
International Society for Applied Ethology. Exeter, UK: Universities
Federation for Animal Welfare, 1995. pp. 81–2.
[36] Hopster H, Van der Werf JTN, Blokhuis HJ. Short and long term
consistency of behavioural and physiological stress responses in dairy
cows during a novel environment test. In: Hopster H, editor. Coping
strategies in dairy cows. Wageningen, The Netherlands: Wageningen
University, 1998. pp. 67–84 (PhD thesis).
[37] Visser EK, Van Reenen CG, Rundgren M., Zetterqvist M., Morgan K.,
Blokhuis HJ. Responses of horses in behavioural tests correlate with
temperament assessed by riders. Equine Vet J (in press).