ArticlePDF Available

Pair-bonding and companion recognition in domestic donkeys, Equus asinus

Authors:

Abstract

Pair and social bonding has been documented in various taxa, where pair formations are often described as being driven by kinship or sexual motivation. However, pair-bonding between unrelated individuals where sexual motivation is not a factor is not well documented. Many social relationships and pair-bonds between members of a dyad are facilitated by each individual's ability to recognise their partner using cues which are characteristic of that particular individual. The aims of this study were i) to investigate the existence of pair-bonding in domestic donkeys and ii) to determine whether members of a dyad could recognise their companion during a Y-maze recognition test. Subjects were 55 unrelated donkeys (38 gelded males, 15 females) in seven groups of mixed or same sex, comprising 4–14 individuals. Spatial proximity (nearest-neighbour) was observed three times a day over a 22-day period. Using a simulation approach based on observed data to generate randomised nearest-neighbour matrices, the statistical significance of social relationships was estimated. Of these, 42 (79.2%) were involved in significantly (p < 0.05) non-random nearest-neighbour relationships, most of which were reciprocal pair relationships. Based on the spatial data, 24 of the donkeys which had shown significant reciprocal nearest-neighbour preferences for one individual (companion) were then used in a Y-maze recognition test in which they were presented with a choice of their companion and either a familiar donkey from the same group or an unfamiliar donkey from a different group. Donkeys’ spatial location in the Y-maze demonstrated a preference for their companion versus familiar (one sample Wilcoxon signed rank test, W = 239, p = 0.002) or unfamiliar donkeys (W = 222, p = 0.041). These results verify anecdotal evidence from donkey handlers that donkeys often form pair-bonds, and show that reciprocal social preference and recognition are the basis of these. Pair-bond formation and companionship among donkeys have potential implications for their management, husbandry and welfare.
Applied
Animal
Behaviour
Science
143 (2013) 67–
74
Contents
lists
available
at
SciVerse
ScienceDirect
Applied
Animal
Behaviour
Science
jou
rnal
h
om
epa
ge:
www.elsevier.com/locate/applanim
Pair-bonding
and
companion
recognition
in
domestic
donkeys,
Equus
asinus
Leigh
M.A.
Murraya,b,,
Katharine
Byrnec,
Richard
B.
D’Eathb
aRoyal
(Dick)
School
of
Veterinary
Studies,
The
University
of
Edinburgh,
Easter
Bush
Veterinary
Centre,
EH25
9RG,
UK
bAnimal
Behaviour
and
Welfare,
Animal
and
Veterinary
Sciences
Research
Group,
SRUC,
West
Mains
Road,
Edinburgh,
EH26
0PH,
UK
cCollege
of
Medicine
and
Veterinary
Medicine,
University
of
Edinburgh,
UK
a
r
t
i
c
l
e
i
n
f
o
Article
history:
Accepted
27
November
2012
Available online 29 December 2012
Keywords:
Donkeys
Pair-bonding
Companion
recognition
Welfare
Y-Maze
a
b
s
t
r
a
c
t
Pair
and
social
bonding
has
been
documented
in
various
taxa,
where
pair
formations
are
often
described
as
being
driven
by
kinship
or
sexual
motivation.
However,
pair-bonding
between
unrelated
individuals
where
sexual
motivation
is
not
a
factor
is
not
well
doc-
umented.
Many
social
relationships
and
pair-bonds
between
members
of
a
dyad
are
facilitated
by
each
individual’s
ability
to
recognise
their
partner
using
cues
which
are
char-
acteristic
of
that
particular
individual.
The
aims
of
this
study
were
i)
to
investigate
the
existence
of
pair-bonding
in
domestic
donkeys
and
ii)
to
determine
whether
members
of
a
dyad
could
recognise
their
companion
during
a
Y-maze
recognition
test.
Subjects
were
55
unrelated
donkeys
(38
gelded
males,
15
females)
in
seven
groups
of
mixed
or
same
sex,
comprising
4–14
individuals.
Spatial
proximity
(nearest-neighbour)
was
observed
three
times
a
day
over
a
22-day
period.
Using
a
simulation
approach
based
on
observed
data
to
generate
randomised
nearest-neighbour
matrices,
the
statistical
significance
of
social
relationships
was
estimated.
Of
these,
42
(79.2%)
were
involved
in
significantly
(p
<
0.05)
non-random
nearest-neighbour
relationships,
most
of
which
were
reciprocal
pair
relation-
ships.
Based
on
the
spatial
data,
24
of
the
donkeys
which
had
shown
significant
reciprocal
nearest-neighbour
preferences
for
one
individual
(companion)
were
then
used
in
a
Y-maze
recognition
test
in
which
they
were
presented
with
a
choice
of
their
companion
and
either
a
familiar
donkey
from
the
same
group
or
an
unfamiliar
donkey
from
a
different
group.
Donkeys’
spatial
location
in
the
Y-maze
demonstrated
a
preference
for
their
companion
versus
familiar
(one
sample
Wilcoxon
signed
rank
test,
W
=
239,
p
=
0.002)
or
unfamiliar
donkeys
(W
=
222,
p
=
0.041).
These
results
verify
anecdotal
evidence
from
donkey
handlers
that
donkeys
often
form
pair-bonds,
and
show
that
reciprocal
social
preference
and
recog-
nition
are
the
basis
of
these.
Pair-bond
formation
and
companionship
among
donkeys
have
potential
implications
for
their
management,
husbandry
and
welfare.
© 2012 Elsevier B.V. All rights reserved.
1.
Introduction
Social
pair-bonding,
whereby
long-lasting
preferen-
tial
partner
bonds
and
affiliative
behaviour
are
observed
Corresponding
author
at:
2
Longnewton
Place,
St.
Boswells,
Melrose,
TD6
9ES,
UK.
Tel.:
+44
0
7929
263
370.
E-mail
addresses:
leigh
murray75@hotmail.co.uk
(L.M.A.
Murray),
kate.byrne@btinternet.com
(K.
Byrne),
rick.death@sac.ac.uk
(R.B.
D’Eath).
between
individuals,
has
primarily
been
documented
in
primates
(Berghänel
et
al.,
2011;
Mitani,
2009;
Moscovice
et
al.,
2010),
but
has
also
been
studied
in
some
ungu-
lates
(Cameron
et
al.,
2009;
Wasilewski,
2003);
horses
(Heitor
et
al.,
2006;
Proops
et
al.,
2012;
VanDierendonck
and
Spruijt,
2012;)
birds
(Emery
et
al.,
2007;
Massen,
2010;
Svec
et
al.,
2009)
and
rodents
(Wang
and
Aragona,
2004;
Williams
et
al.,
1992;
Young
et
al.,
2011).
However
the
majority
of
research
in
this
area
focuses
on
kinship
and
sexual
motivation
as
the
drivers
of
pair
formation.
0168-1591/$
see
front
matter ©
2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.applanim.2012.11.005
68 L.M.A.
Murray
et
al.
/
Applied
Animal
Behaviour
Science
143 (2013) 67–
74
Research
investigating
non-kin
bonding
among
animals
however,
is
less
well
documented.
The
maintenance
of
‘friendships’
and
close
bonds
among
dyads,
facilitated
by
each
individual’s
ability
to
identify
the
distinct
character-
istics
of
its
preferential
partner
via
individual
recognition
cues,
may
provide
psychosocial
benefits
to
each
of
the
individuals
involved.
The
evolutionary
function
of
unre-
lated
and
non-sexually
motivated
pair-bonds
however,
still
remains
unclear
(Clutton-Brock,
2009;
Silk,
2002;
Wasilewski,
2003).
There
has
been
little
research
into
pair-bonding
in
domestic
donkeys
Equus
asinus.
Based
on
nearest-
neighbour
observations
in
different
contexts,
Wasilewski
(2003)
reported
that
donkeys
had
small
numbers
of
strong
‘friendships’
which
were
long-lasting,
but
also
those
social
preferences
were
situation-specific.
The
social
dynamics
of
domestic
donkey
groups
are
usually
characterised
by
close
proximity
of
individuals,
implying
high
mutual
tol-
erance
and
affiliative
behaviour
(Whitehead
et
al.,
1991).
Anecdotal
evidence
from
donkey
handlers
suggests
that
it
is
‘common
knowledge’
that
donkeys
form
strong
pair-
bonds
with
other
donkeys,
which
become
their
companion,
or
there
may
be
individuals
for
whom
they
have
a
par-
ticular
preference
(Svendsen,
2008;
Whitehead
et
al.,
1991).
Although
the
strength
of
the
bond
may
vary,
it
has
been
noted
that
the
separation
of
bonded
donkeys
can
cause
extreme
distress
to
either
individual,
resulting
in
‘pining’
and
inappetence
which
can
potentially
lead
to
hyperlipaemia
(Svendsen,
2008;
Whitehead
et
al.,
1991).
Pair-bonds
have
not
been
observed
in
wild
or
feral
popu-
lations
(Klingel,
1998;
Moehlman,
1998;
Rudman,
1998).
This
may
not
be
surprising
given
that
domestic
donkeys
are
descended
from
two
independent
domestication
events
each
from
a
separate
subspecies
(Kimura
et
al.,
2011).
One
ancestor
was
the
Nubian
wild
ass,
Equus
africanus
africanus,
while
the
other
was
a
population
from
an
area
in
mod-
ern
Eritrea
(probably
corresponding
with
the
Eritrean
wild
ass
Equus
africanus
dianae
(Clutton-Brock,
1999)).
As
both
of
these
subspecies
are
now
presumed
extinct
in
the
wild
(Kimura
et
al.,
2011),
there
is
no
way
of
ascertaining
the
nat-
ural
behaviour
of
either.
Furthermore,
donkey
populations
are
composed
of
a
mixture
of
descendants
of
both
ances-
tors
(Beja-Pereira
et
al.,
2004),
which
raises
the
possibility
of
derived
social
behaviour
based
on
a
whole
range
of
inter-
actions
between
two
originally
genetically
distinct
groups.
The
only
point
of
reference
is
the
Somali
wild
ass,
Equus
africanus
somaliensis,
for
which
limited
observations
have
been
made
in
the
natural
habitat
and
which
indicate
that
social
bonds
exist
primarily
between
females
and
their
off-
spring.
Other
social
groups
have
been
observed,
including
all
female
groups
and
some
mixed
male–female
groups,
but
these
are
rarer
and
group
numbers
are
small
(Moehlman
et
al.,
2012).
The
suggested
reason
for
this
structure
is
the
ecology
of
desert
environments
in
which
resources
are
limited
and
defended,
however
the
possibility
that
there
may
have
been
differences
among
subspecies
because
of
ancestry
cannot
be
ruled
out.
This
study
aimed
to
investigate
social
preferen-
ces
and
to
determine
whether
there
are
pair-bonds
and
pair-associations
among
domestic
donkeys.
We
define
pair-bonds
as
reciprocal
preferences
between
both
Table
1
The
number
and
size
of
enclosure,
number
of
males
including
age
range
and
number
of
females
including
age
range
in
each
field.
Paddock
Size
(ha)
Males
Females
Age-Range
1
2.02
8
0
2–10
yrs
2
2.02
8
0
6–13
yrs
3 1.42 4 0 5–8
yrs
4 1.50
5
0
5–12
yrs
5
2.10
4
4
4–7
yrs
6
2.12
6
2
9–26
yrs
7
2.27
3
11
8–30
yrs
members
of
a
dyad,
whereas
pair-associations
are
defined
as
reciprocal
and
non-reciprocal
preferences
for
more
than
one
individual.
The
study
comprised
two
components:
1)
the
observation
of
nearest-neighbour
(NN)
preferences
between
individuals
and
thus
identification
of
bonds
based
on
the
frequency
of
spatial
proximity
within
dyads.
2)
A
subset
of
donkeys
showing
a
clear
preference
for
one
other
donkey
(companion
pairs)
was
identified
and
the
ability
of
each
member
of
a
pair
to
recognise
its
companion
was
then
investigated
using
a
Y-maze
social
discrimination
test,
where
the
companion
was
presented
alongside
a
familiar
or
unfamiliar
donkey.
2.
Material
and
methods
2.1.
Animals
The
subjects
of
this
study
were
55
donkeys
(38
gelded
males,
17
females)
at
the
Scottish
Borders
Donkey
Sanctu-
ary,
registered
charity
no.
SC
034
634.
Many
of
the
donkeys
had
been
family
pets
that
could
no
longer
be
cared
for.
Some
of
these
were
established
pairs
that
were
kept
together
in
the
same
group.
Other
individuals
were
rescued,
arrived
without
a
companion
and
were
placed
in
a
group
of
simi-
lar
individuals.
Details
of
group
structure
and
location
are
shown
in
Table
1.
2.2.
Behavioural
observations
to
identify
nearest-neighbour
preferences
Behavioural
observations
of
donkeys
were
carried
out
in
situ
in
their
home
enclosures
to
establish
each
individ-
ual’s
nearest-neighbour
preferences
(observer
L.M.AM.).
Of
the
67
donkeys
at
the
sanctuary,
55
were
observed
(donkeys
kept
only
in
pairs
were
excluded).
Each
don-
key
was
distinguished
based
on
its
own
distinct
features,
such
as
coat
colour
and
length,
height,
markings
and
body
shape.
Instantaneous
scan
sampling
of
donkeys
in
seven
enclosures
was
conducted
at
two-
hourly
intervals,
three
times
per
day
for
22
days
(on
one
observation
day,
only
two
scan
samples
were
done).
The
total
number
of
observations
per
donkey
was
65.
Each
individual
donkey’s
nearest-neighbour
(identity)
was
recorded,
unless
no
other
donkeys
were
within
two
body
lengths
(approx
3.6
m)
in
which
case
the
donkey
was
scored
as
‘on
own’.
During
the
course
of
observations,
two
donkeys
were
introduced
to
each
of
enclosures
1
and
7.
These
four
donkeys
had
been
living
in
pairs,
so
had
not
been
previ-
ously
observed.
From
the
point
of
introduction,
they
were
L.M.A.
Murray
et
al.
/
Applied
Animal
Behaviour
Science
143 (2013) 67–
74 69
included
in
observations.
The
total
number
of
observations
for
these
individuals
was
lower
than
those
of
the
other
donkeys
in
the
study
and
they
were
excluded
from
further
analyses.
2.3.
Companion
recognition
testing
Twelve
companion
pairs
of
donkeys
(n
=
24)
were
cho-
sen
from
behavioural
observation
data.
These
individuals
were
selected
because
they
showed
a
significantly
non-
random
social
preference,
and
from
inspection
of
data,
showed
a
clear
preference
(at
least
one-third
of
observa-
tions)
for
one
specific
donkey.
Finally,
this
preference
had
to
be
mutual,
in
that
both
members
of
a
companion
pair
had
each
other
as
the
most
frequently
observed
NN
(e.g.
Hero’s
most
frequent
NN
=
Flora;
Flora’s
most
frequent
NN
=
Hero).
For
safety
reasons,
all
donkeys
from
enclosure
1
plus
two
from
enclosure
6
were
excluded
from
recognition
tests
due
to
their
large
size
and/or
difficulty
in
handling.
Both
donkeys
in
a
companion
pair
were
used
as
test
subjects
on
different
occasions.
The
experiment
consisted
of
two
treatments:
1)
C/F:
the
test
subject
was
given
the
choice
of
companion
(C)
versus
a
familiar
donkey
(F)
which
they
associate
with
in
their
home
enclosure
(randomly
chosen);
2)
C/U:
the
test
subject
was
given
the
choice
of
companion
(C)
versus
an
unfamiliar
donkey
(U)
from
a
dif-
ferent
(non-adjacent)
enclosure.
Each
donkey
completed
three
tests
of
each
treatment
type.
2.4.
Apparatus
and
experimental
procedure
The
test
arena
consisted
of
a
7.31
m
×
10.97
m
area
of
fenced
paddock,
containing
a
modified
Y-maze,
with
two
fenced-in
pens
(3.65
m
×
4.27
m)
at
one
end,
each
containing
one
stimulus
donkey
(companion/familiar
or
companion/unfamiliar
donkey
depending
on
the
test).
Stimulus
donkeys
were
visible
to
the
test
subject
and
vice
versa,
and
olfactory
and
auditory
cues
remained
available.
For
social
discrimination
tests,
the
companion
and
one
familiar
(or
unfamiliar)
donkey
were
brought
into
the
test
arena
and
placed
in
the
two
stimulus
pens
(according
to
a
pre-arranged
randomised
schedule).
The
test
subject
was
then
collected
after
approximately
2
min
to
allow
a
small
separation
period
between
the
subject
and
its
companion.
Anecdotal
observations
suggest
that
after
a
period
of
sepa-
ration,
donkeys
often
seek
out
their
companion,
therefore
providing
motivation
(social
re-instatement)
for
the
pref-
erence.
The
test
subject
was
released
into
the
centre
of
the
test
area
at
the
end
farthest
from
the
stimulus
donkey
pens,
(Fig.
1)
and
allowed
to
move
freely
around
the
arena
for
3
min.
For
the
purposes
of
behavioural
observations,
the
choice
area
was
divided
into
six
zones
(Fig.
1);
1–3,
left
and
right
zone
1
left
(Z1L),
zone
1
right
(Z1R)
and
so
forth,
with
Z1L
and
Z1R
being
the
closest
to
the
stim-
ulus
donkeys.
When
a
donkey
remained
stationary
for
more
than
3
s
(this
was
the
minimum
time
observed
in
pilot
studies
in
which
a
test
donkey
chose
its
preferred
donkey),
the
time
and
location
(zone)
were
noted.
Body
orientation
relative
to
the
arena
(using
the
shorthand
N,
S,
E,
W
where
N
=
north
=
oriented
towards
the
stimulus
Fig.
1.
Diagram
of
Y-maze
test
arena.
This
example
illustrates
how
don-
key
positions
(including
duration
and
orientation)
were
recorded
during
the
recognition
tests.
The
test
subject
enters
the
arena
and
5
s
into
the
3
min
test,
remains
in
zone
2L
for
40
s,
in
a
north-facing
position.
The
test
subject
then
moves
into
zone
1L
where
it
remains
for
38
s
in
a
north-facing
position,
before
then
moving
into
zone
2R.
The
test
subject
remains
in
a
south-facing
position
for
30
s,
thereafter
moving
into
zone
1L,
and
fac-
ing
north
for
the
remainder
of
the
test
period.
From
this
example
we
see
that
the
test
subject
has
chosen
its
companion
for
3
out
of
4
movements,
orientated
itself
to
face
north
towards
its
companion
and
has
remained
primarily
in
zones
1
and
2,
which
are
closest
to
its
companion.
Note
that
the
test
arena
is
not
to
scale
and
the
dotted
lines
indicate
markers
on
the
perimeter
of
the
arena
to
indicate
zone
areas
(1,
2
and
3)
of
equal
size.
donkey
and
where
S
=
south
=
oriented
away
from
the
stim-
ulus
donkey)
was
also
recorded
(see
Fig.
1
for
an
example).
Note
that
north
was
not
true
compass
north.
Each
subject
donkey
completed
six
social
discrimina-
tion
tests
in
total.
On
a
given
test
day,
each
donkey
received
one
C/F
and
one
C/U
test.
Each
donkey
had
one
such
test
day
per
week,
with
the
three
test
days
each
sepa-
rated
by
seven
days.
Side
preferences
were
avoided,
by
pseudo-randomising
the
location
of
the
companion
donkey
within
and
between
test
days,
and
avoiding
any
structured
sequences.
The
test
arena
was
mucked-out
at
the
end
of
each
day
to
remove
faecal
deposits
as
potential
olfactory
distractions
prior
to
subsequent
tests.
The
experiment
was
carried
out
in
accordance
with
ISAE’s
ethical
guidelines.
All
animals
used
were
accus-
tomed
to
frequent
handling
and
handlers
who
took
part
in
the
experiment
were
experienced
and
knew
the
test
animals.
Stress
during
movement
from
home
enclosures
to
test
arena
was
minimised
through
patient
handling
and
allowing
rest
periods
throughout
the
testing
session.
70 L.M.A.
Murray
et
al.
/
Applied
Animal
Behaviour
Science
143 (2013) 67–
74
Test
donkeys
were
housed
in
nearby
stables
when
not
in
use,
accompanied
by
another
donkey
or
handler.
Larger
donkeys
were
managed
by
two
handlers
to
minimise
any
boisterous,
over-excited
behaviour
due
to
novel
test
donkeys/environment
thus
ensuring
handler
safety.
2.5.
Statistical
analyses
2.5.1.
Behavioural
observations
To
test
the
significance
of
nearest-neighbour
preferen-
ces
a
randomisation
method
described
by
Sibbald
et
al.
(2005)
was
followed.
This
method
requires
that
there
are
no
missing
data;
a
nearest-neighbour
must
be
allocated
to
each
focal
animal
at
every
scan,
and
there
must
be
the
same
number
of
scans
each
day
(3
for
this
data).
The
dataset
had
‘missing’
data
because
animals
were
sometimes
scored
as
‘on
own’
meaning
that
no
nearest-neighbour
was
avail-
able,
sometimes
animals
were
‘out
of
sight’
and
only
two
scans
took
place
on
one
of
the
observation
days.
To
deal
with
this,
observation
days
were
excluded
entirely
where
more
than
1/3
of
observations
were
missing.
This
resulted
in
a
mean
(±s.d.)
number
of
days
included
of
18.4
(±2.4).
Where
less
than
1/3
of
observations
were
absent,
miss-
ing
observations
were
replaced
by
allocating
a
random
nearest-neighbour.
Although
not
ideal,
this
process
was
conservative,
as
it
reduced
the
chance
of
finding
patterns
of
nearest-neighbour
associations
that
were
non-random.
The
randomisation
was
performed
by
carrying
out
simulations
based
on
the
original
enclosure
data.
Each
enclosure
was
analysed
separately.
Data
were
arranged
into
matrices
for
each
observation
day
with
rows
for
each
focal
animal
and
columns
indicating
the
identity
of
its
nearest-neighbour.
The
simulation
then
randomly
real-
located
donkey
identity
across
the
available
rows
and
columns
of
each
matrix,
generating
new
matrices
with
sim-
ilar
numbers
of
observations
overall
and
in
each
cell,
but
arranged
differently.
A
summary
matrix
of
the
simulated
data
across
the
observed
days
was
then
stored.
This
pro-
cess
was
repeated
1000
times
for
each
enclosure.
Using
daily
summaries
as
the
basis
for
the
simulations
makes
the
assumption
that
observations
within
the
same
day
may
not
be
independent,
but
that
different
days
are
independent.
For
each
enclosure,
the
mean
values
in
each
cell
from
the
1000
randomly
simulated
summary
matrices
were
cal-
culated.
Reciprocal
values
(e.g.
Jethro
has
Hamish
as
his
NN,
Hamish
has
Jethro
as
his
NN)
were
first
added
together.
The
distribution
of
the
1000
most
highly
associated
pairs
(i.e.
largest
mean
value
in
the
matrix
cell)
was
then
generated
and
ranked.
The
summary
matrix
of
the
observed
data
was
then
inspected.
If
the
observed
value
for
the
most
highly
associated
pair
fell
outside
the
central
0.95
of
simulated
values
(i.e.
it
was
larger
than
the
976th
ranked
value),
then
that
pairing
was
considered
to
be
significant
at
p
<
0.05
(or
p
<
0.01
if
the
observed
value
fell
outside
the
central
0.99
of
simulated
values;
i.e.
it
was
larger
than
the
996th
ranked
value).
The
distribution
of
the
second
most
highly
associ-
ated
pair
from
the
simulated
summary
matrices
was
then
generated
and
compared
with
the
second
most
highly
asso-
ciated
pair
from
the
observed
summary
matrix
in
a
similar
way
and
so
on,
until
all
significant
pairings
had
been
iden-
tified.
2.5.2.
Companion
recognition
testing
Each
test
subject’s
location
was
used
as
evidence
for
it
having
made
a
choice.
So
if
the
test
subject
predominantly
spent
time
on
the
side
in
front
of
its
companion,
it
was
regarded
as
showing
a
preference
for
the
companion.
For
both
C/F
and
C/U
trials,
the
total
duration
each
test
subject
spent
on
the
companion
side
(in
Z1,
Z2
or
Z3)
was
compared
to
the
total
time
spent
on
the
other
side
(summing
data
from
the
three
replicates)
and
the
difference
between
the
two
was
calculated.
To
allow
for
the
possibility
that
donkeys
were
not
actu-
ally
making
a
choice
when
they
were
far
away
from
the
stimulus
donkeys,
the
location
preferences
using
only
data
from
zone
1
or
from
zone
1
and
2
combined
was
also
ana-
lysed.
As
above,
the
total
duration
on
the
companion
side
was
subtracted
from
the
duration
spent
on
the
other
side.
To
allow
for
the
possibility
that
donkeys
were
only
really
showing
a
preference
when
facing
towards
the
stimulus
donkeys
(north),
analysis
of
duration
in
each
location
was
performed
on
north
facing
data
only
(in
the
same
format
as
previous
analyses).
Descriptive
statistics
revealed
that
the
data
were
not
normally
distributed;
therefore
non-parametric
one
sam-
ple
Wilcoxon
signed
rank
tests
were
applied
to
the
differences
between
the
preference
for
the
companion
and
the
opposite
side
using
the
variables
described
above.
Finally,
to
determine
whether
donkeys
showed
a
pref-
erence
for
the
companion
right
at
the
start
of
the
test
by
moving
towards
him/her
straight
away,
the
first
loca-
tion
a
donkey
moved
to
and
then
stopped
after
entering
the
test
arena
(first
movement)
was
analysed.
Choices
for
the
companion
were
counted
as
+1,
and
choices
for
the
familiar
or
unfamiliar
were
counted
as
-1,
and
summed
for
each
donkey’s
three
tests
to
give
a
score
between
+3
and
-3.
Non-parametric
one
sample
Wilcoxon
signed
rank
tests
were
then
applied
to
this
data
to
determine
which
side
test
donkeys
first
went
to
companion
or
familiar/unfamiliar.
3.
Results
3.1.
Behavioural
observations:
nearest-neighbours
Nearest-neighbour
observations
are
summarised
and
significant
pairings
are
indicated
in
Table
2.
Of
the
55
observed
donkeys,
42
(76.4%)
were
involved
in
significantly
non-random
social
relationships.
One
donkey
in
enclosure
1
(Angus)
had
no
significant
social
preference
in
the
first
10
days
but
did
in
the
last
12
days
of
observations
(see
below).
The
42
donkeys
involved
in
significant
relationships
were
primarily
reciprocal
pairs
(38
donkeys;
69.1%),
although
4
(7.5%)
non-reciprocal
relationships
were
also
seen,
where
a
donkey
preferred
a
member
of
an
existing
pairing
(e.g.
enclosure
2,
Donovan
and
Cookie
preferred
to
associate
with
Ted
and
Joseph
respectively;
enclosure
5
Tinkerbelle
preferred
Molly,
who
is
paired
with
Wogan;
enclosure
1
Jethro
preferred
Hamish,
and
then
Angus).
In
all
three
cases,
a
third
donkey
was
associating
a
pair
that
showed
a
recip-
rocal
bond
with
one
other,
therefore
these
groups
did
not
appear
to
be
trios
with
equal
relationships
among
the
three
individuals.
L.M.A.
Murray
et
al.
/
Applied
Animal
Behaviour
Science
143 (2013) 67–
74 71
Table
2
Percentages
of
nearest-neighbour
(NN)
associations
for
donkeys
in
Enclosures
1–7.
Significant
(*p
<
0.05;
**p
<
0.01)
pair-associations
are
indicated,
based
on
the
results
of
simulations.
Percentages
are
given
for
dyads
used
in
subsequent
recognition
tests:
(1)
Hamish/Seamus,
Pixie/Milo
and
Kofi/Ollie,
(2)
Ted/Joseph
and
Patch/Francis,
(3)
Blue/Pascal
and
Dusty/Daniel,
(4)
Harry/Jasper,
(5)
Lily/Dolly,
(6)
Ben/Craig,
Cocoa/Charlotte
and
Denis/Twinkle,
(7)
Hero/Flora,
Felix/Freda
and
Perky/Sparkle.
All
donkeys
had
65
observations
of
NN,
with
the
exception
of
Kofi
and
Ollie
(36),
Amelie
(24)
and
Octavia
(15).
Enclosure
Donkey
Freq.
Companion
%
NN
P-Value
1 Hamish
Seamus
63%
<0.01
Seamus Hamish
62%
<0.01
Pixie Milo
46%
<0.01
Milo
Pixie
45%
<0.01
Kofi Ollie 47%
<0.01
Ollie
Kofi
47%
<0.01
Jethro
Hamish
25%
<0.01
Angus
Pixie
29%
<0.01
2
Ted
Joseph
56%
<0.01
Patch
Francis
42%
<0.01
Joseph Ted 60%
<0.01
Francis
Patch
40%
<0.01
Dougal Del
23%
<0.01
Donovan
Ted
31%
<0.05
Del
Dougal
20%
<0.01
Cookie Joseph 27%
<0.05
3 Pascal Blue
53%
<0.01
Dusty
Daniel
47%
<0.01
Daniel Dusty
47%
<0.01
Blue
Pascal
60%
<0.01
4
Shaun
Clive
23%
ns
Jasper
Harry
40%
<0.05
Harry Jasper 35%
<0.05
Clive
Harry
23%
ns
Caz Jasper
17%
ns
5
Dec
Lily
16%
ns
Ant
Storm
19%
<0.05
Dolly
Lily
49%
<0.01
Lily Dolly 44%
<0.01
Molly
Wogan
19%
<0.01
Storm
Ant
22%
<0.05
T-Belle
Molly
26%
<0.01
Wogan
Molly
21%
<0.01
6
Ben
Craig
49%
<0.01
Charlotte
Cocoa
73%
<0.01
Cocoa
Charlotte
75%
<0.01
Craig
Ben
44%
<0.01
Denis
Twinkle
25%
<0.01
Jock
Rambo
19%
<0.05
Rambo
Jock
20%
<0.05
Twinkle
Denis
26%
<0.01
7
Amelie
Octavia
13%
ns
Donna
Silver
11%
ns
Felix
Freda
60%
<0.01
Flora
Hero
83%
<0.01
Freda
Felix
70%
<0.01
Hallan
Felix
14%
ns
Hero
Flora
80%
<0.01
Jenny
Donna
13%
ns
Lottie
Donna
12%
ns
Octavia Hallan
40%
ns
Perky
Sparkle
63%
<0.01
Silver
Donna
11%
ns
Sparkle
Perky
65%
<0.01
Zoe
Donna
9%
ns
All
of
the
donkeys
in
enclosures
2,
3
and
6
and
all
but
one
in
enclosure
5
had
significant
social
relationships.
Of
the
10
donkeys
with
non-significant
social
relationships,
three
were
in
enclosure
4
(n
=
5,
one
significant
reciprocal
pair),
and
six
were
in
enclosure
7,
the
largest
group,
with
the
least
stable
membership.
One
potential
explanation
for
the
instability
of
the
largest
group
is
that
large
group
size
is
incompatible
with
natural
donkey
behaviour
in
the
wild.
Limited
field
observations
of
wild
African
asses
suggest
that
group
size
is
small
(Moehlman
et
al.,
2012).
There-
fore
we
cannot
discount
the
possibility
that
donkeys
in
a
large
group
will
seek
to
limit
social
contacts
by
breaking
into
smaller
subunits
that
matches
their
natural
behaviour
more
closely.
This
issue
can
only
be
resolved
by
further
investigation.
Enclosure
1
(8
donkeys)
makes
an
interesting
case
study
of
the
stability
of
social
relationships,
as
most
relationships
were
unaffected
by
a
change
in
group
membership.
Kofi
and
Ollie,
which
had
been
kept
as
a
pair
in
another
enclo-
sure,
were
introduced
after
day
10.
After
their
introduction,
Kofi
and
Ollie
showed
the
most
significant
(p
<
0.01)
recip-
rocal
paired
relationship
in
the
group.
Pixie
and
Milo
had
a
reciprocal
relationship
before
the
introduction
(p
<
0.01),
with
Angus
preferring
both
Milo
and
Pixie
as
nearest-
neighbours
(12
and
11
out
of
27
observations
respectively,
relationships
not
significant).
After
the
introduction,
all
three
of
the
pairwise
relationships
between
Milo,
Pixie
and
Angus
became
significant
at
p
<
0.01.
The
biggest
change
occurred
for
Jethro.
Prior
to
the
introduction
of
Kofi
and
Ollie,
he
had
a
preference
(p
<
0.01)
for
Hamish
(who
had
a
reciprocal
preference
for
Seamus
p
<
0.01).
After
the
introduction,
Hamish
and
Seamus
continued
with
their
reciprocal
relationship
(p
<
0.01),
while
Jethro
now
showed
a
significant
(p
<
0.01),
but
non-reciprocated
preference
for
Angus
instead
of
Hamish.
3.2.
Companion
recognition
tests
As
hypothesised,
donkeys
spent
longer
on
the
side
of
companion
donkeys,
when
given
the
choice
of
companion
versus
familiar,
and
companion
versus
unfamiliar
donkeys
(Table
3a
and
b).
Also,
when
only
close
approaches
(zone
1)
were
treated
as
a
choice
for
that
stimulus
donkey,
donkeys
spent
signifi-
cantly
longer
on
the
companion
side.
This
was
found
for
the
C/F
tests
and
for
the
C/U
tests
(Table
3).
This
also
occurred
when
approaches
in
both
zone
1
and
2
were
treated
as
a
choice
for
that
stimulus
donkey.
When
only
time
spent
oriented
towards
the
stimulus
donkeys
(‘north’)
was
treated
as
a
choice
for
a
stimulus
don-
key,
donkeys
spent
longer
on
the
companion
side.
This
was
significant
for
C/F
tests,
but
was
a
non-significant
tendency
(p
<
0.1)
for
C/U
tests
(Table
3).
Analysis
of
donkeys’
first
movements
in
the
arena
to
investigate
their
initial
choices
showed
a
significant
pref-
erence
for
companions
over
familiars
(W
=
223.5,
p
=
0.037),
with
47/72
(65.3%)
of
first
movements
being
made
towards
the
companion.
The
initial
preference
for
companions
was
not
observed
when
the
other
donkey
was
unfamiliar
(W
=
199.5,
p
=
0.162),
with
43/72
(59.7%)
of
first
move-
ments
being
made
towards
the
companion.
72 L.M.A.
Murray
et
al.
/
Applied
Animal
Behaviour
Science
143 (2013) 67–
74
Table
3
(a)
Overall
(location
within
the
test
arena),
zone
(1,
1
and
2
combined,
whereby
zone
1
would
be
that
closest
to
the
choice
donkey
–C,
F
or
U),
and
orientation
preferences,
significance
levels,
means
of
summed
durations
from
all
3
replicates
(s)
of
time
spent
on
either
the
side
of
C,
F
or
U,
and
estimated
medians
of
the
difference
(generated
by
the
Wilcoxon
test)
for
24
donkeys
during
companion
(C)
versus
familiar
(F)
recognition
tests.
(a)
Measure
of
choice
Mean
duration
(C)
Mean
duration
(F)
Wilcoxon
statistic
(W)
p-value
Est.
median
Overall
Location
315
198
239
0.002
114
Zone
1
167
98.8
248
0.005
86
Zones
1
and
2 178 93.5
237 0.003
104
Orientation 156.8
111.9
236
0.015
52
(b)
As
(a)
for
24
donkeys
during
companion
(C)
versus
unfamiliar
(U)
recognition
tests.
Measure
of
choice
Mean
duration
(C)
Mean
duration
(U)
Wilcoxon
statistic
(W)
p-value
Est.
median
Overall
Location
309
195
222
0.041
98
Zone
1
189
88.3
227
0.029
98
Zones
1
and
2
128
73.7
229
0.025
100
Orientation
175.4
103.5
209
0.095
59
4.
Discussion
4.1.
Behavioural
observations
Based
on
the
results
from
nearest-neighbour
enclosure
observations,
this
study
verified
anecdotal
evidence
(see
for
example
Burden
et
al.,
2008)
and
demonstrated
that
pair-bonding
in
domestic
donkeys
does
exist.
Over
two-
thirds
of
the
donkeys
analysed
showed
a
clear
preference
for
being
close
to
one
other
individual
(‘companion’),
and
these
preferences
were
reciprocated.
Non-random
social
preferences
for
more
than
one
other
individual
were
observed
in
a
further
four
individuals,
in
which
one
donkey
showed
a
preference
for
another
that
was
in
a
different
reciprocal
pair
relationship.
In
enclo-
sure
1,
following
the
introduction
of
a
bonded
pair,
Jethro
changed
his
(non-reciprocated)
preference
from
Hamish
to
Angus,
while
existing
pairs
were
not
affected.
We
can
speculate
from
this
single
case
study
that
non-reciprocated
relationships
might
be
less
fixed,
but
this
remains
to
be
tested.
Valuable
questions
for
further
investigation
include
how
donkey
pair-bonds
are
formed,
on
what
basis
the
choice
of
companion
is
made,
how
stable
and
long-lasting
these
bonds
are
and
under
what
circumstances
they
can
be
disrupted
or
changed.
We
suggest
that
individual
factors
such
as
age,
temperament
and
history
prior
to
arrival
at
the
sanctuary,
as
well
as
immediate
factors
such
as
group
size,
composition
and
stability
of
membership,
and
perhaps
the
availability
and
distribution
of
resources
are
likely
to
be
important.
Ten
donkeys
showed
no
significant
NN
preference.
Some
of
these
individuals
had
only
recently
arrived
at
the
sanctuary
and
may
not
have
had
time
to
establish
a
pair-bond.
This
was
particularly
true
of
the
six
individ-
uals
in
enclosure
7,
which
had
a
more
fluid
composition
than
other
groups.
Even
though
some
of
these
donkeys
had
been
resident
for
many
years,
the
continual
introduction
of
new
individuals
might
disrupt
existing
relationships.
Alter-
natively,
these
‘loners’
may
be
unable
to
find
a
suitable
companion
or
may
prefer
not
to
have
a
companion.
Studies
describing
pair-associations
in
other
species
have
employed
mutual
grooming
as
a
behavioural
param-
eter
in
addition
to
spatial
proximity
(Cords,
1997;
Wasilewski,
2003).
However,
mutual
grooming
(as
seen
in
primates
(Mitani,
2009)
and
horses
(Feh
and
de
Mazieres,
1993;
Kimura,
1998))
was
not
frequently
observed
among
the
donkeys.
However,
when
companion
pairs
of
donkeys
were
reunited
after
brief
separation,
mounting
often
took
place.
This
has
been
recorded
in
other
ungulate
species,
such
as
fallow
deer
(Holeˇ
cková
et
al.,
2000),
feedlot
steers
(Klemm
et
al.,
1983),
mares
(Gastal
et
al.,
2007),
red
deer
(Vaˇ
nková
and
Bartoˇ
s,
2002;
Hall,
1983a,
1983b),
Thom-
son’s
gazelle
(Walther,
1978).
However,
in
contrast
to
these
species,
in
donkeys
this
behaviour
did
not
appear
to
be
related
to
dominance,
aggression,
oestrus
cycles
or
sexual
motivation.
It
was
observed
in
same-sex
dyads
(both
male
and
female)
with
either
donkey
initiating
mounting.
There-
fore,
just
as
mutual
grooming
may
provide
social
cohesion
among
pairs
in
some
species,
it
is
possible
that
mounting
may
have
a
similar
function
in
donkeys.
Mounting
was
not
systematically
recorded
in
the
present
study,
but
would
be
interesting
to
investigate
in
further
work.
4.2.
Companion
recognition
testing
During
the
Y-maze
recognition
tests,
individuals
chose
to
stand
on
the
same
side
as
their
companion
in
prefer-
ence
to
a
familiar
or
unfamiliar
donkey.
The
only
reward
in
these
tests
was
the
possibility
of
proximity
to
the
preferred
companion.
That
donkeys
chose
to
be
near
their
compan-
ion
demonstrates
the
importance
of
the
bond
between
the
two.
First
movement
results
revealed
highly
significant
pref-
erence
for
companions
in
C/F
tests.
This
highlights
the
ability
of
donkeys
to
recognise
their
companion
quite
quickly
and
from
a
distance.
No
such
preference
was
found
for
first
movement
in
C/U
tests.
Since
the
C/F
first
movement
shows
that
donkeys
can
recognise
their
com-
panion
quickly,
this
lack
of
preference
is
puzzling.
One
possible
explanation
might
be
that
some
donkeys
were
initially
motivated
to
investigate
and
possibly
challenge
the
unfamiliar
donkey
before
returning
to
stand
near
their
companion.
One
explanation
for
the
nearest-neighbour
data
was
that
only
one
member
of
each
pair
showed
a
strong
pref-
erence
to
be
close
to
their
companion
donkey.
The
Y-maze
L.M.A.
Murray
et
al.
/
Applied
Animal
Behaviour
Science
143 (2013) 67–
74 73
tests
excluded
this
possibility:
both
members
of
each
pair
were
tested
as
both
subject
and
stimulus,
and
a
prefer-
ence
for
companions
still
emerged
overall.
This
suggests
that
both
members
of
a
pair-bond
seek
the
proximity
of
their
companion.
Individual
recognition,
defined
as
the
identification
of
numerous
familiar
and
unrelated
individ-
uals
as
‘unique
entities’,
based
on
the
creation
of
mental
representations
of
distinct
recognition
cues
and
individ-
ual
characteristics,
is
abundant
in
diverse
taxa
(Tibbetts
and
Dale,
2007).
Individual
recognition
has
been
docu-
mented
in
various
domestic
species
including
sheep
(Ovis
aries;
da
Costa
et
al.,
2004;
Kendrick
et
al.,
2001),
cattle
(Bos
taurus;
Coulon
et
al.,
2009;
Hagen
and
Broom,
2003),
horses
(Equus
caballus;
Proops
et
al.,
2009)
and
chickens
(Gallus
gallus
domesticus;
D’Eath
and
Stone,
1999)
among
others.
The
function
of
pair-associations
between
individ-
uals
is
generally
driven
by
kinship
(Chiyo
et
al.,
2011;
Frére
et
al.,
2010;
Mitani
et
al.,
2000)
or
sexual
motivation
(Carter
and
Keverne,
2009;
Griggio
and
Hoi,
2011;
Young
and
Lim,
2006).
However,
the
donkeys
in
this
study
are
not
thought
to
form
dominance
hierarchies,
nor
do
they
hold
or
defend
territories.
Females
with
offspring
were
not
observed.
All
individuals
were
unrelated
and
breeding
does
not
occur
as
all
males
were
gelded.
It
is
possible
that
pair-associations
or
‘friendships’
provide
short-term
psychological
benefits
stemming
from
an
evolutionary
fitness
enhancing
func-
tion(s),
however,
the
true
function
of
and
the
motivation
for,
apparently
non-adaptive
pair-bond
formation
under
domestic
conditions
still
remains
unclear
in
donkeys.
The
pair
and
small
group
associations
observed
in
the
majority
of
the
donkeys
appear
to
be
the
mainstay
of
their
social
organisation.
The
motivation
to
be
reunited
with
a
companion
clearly
drove
behaviour
in
the
Y-maze
tests.
Disruption
of
these
social
associations
could
have
welfare
consequences.
Isolation
stress
can
significantly
reduce
an
animal’s
welfare,
and
many
social
animals
are
motivated
to
access
conspecifics
(e.g.
cows:
Bøe
and
Færevik,
2003;
Bosch
et
al.,
2008;
Hennessy,
1997;
rats:
Patterson-Kane
et
al.,
2002;
pigs:
Pedersen
et
al.,
2002;
rodents:
Rault,
2012).
In
sheep
(da
Costa
et
al.,
2004),
face
pictures
of
familiar
conspecifics
presented
to
socially
isolated
sheep
reduced
levels
of
stress
and
fear.
Several
studies
have
shown
that
cattle
form
social
bonds,
for
example
Boissy
and
Le
Neindre
(1997)
demonstrated
that
social
isolation
increased
heart
rate,
vocalisations
and
plasma
cortisol
con-
centrations
in
heifers
and
that
these
were
reduced
when
individuals
were
reunited
with
their
peers.
Dairy
calves
also
exhibit
preferences
for
a
companion
and
association
with
this
familiar
individual
has
a
calming
effect
(Fæverick
et
al.,
2006).
Anecdotal
reports
indicate
that
when
sep-
arated
from
their
companion,
donkeys
become
anxious,
displaying
mental
and
physical
signs
of
distress
such
as
vocalising,
fence-pacing,
dullness
of
character
and
inap-
petence,
leading
to
an
increased
possibility
of
developing
hyperlipaemia
in
response
to
stress
(Burden
et
al.,
2011;
Svendsen,
2008;
Whitehead
et
al.,
1991).
The
importance
of
pair-bonds
or
small
group
associa-
tions
in
donkeys
has
implications
for
their
management.
Long
term,
donkeys
are
likely
to
have
better
welfare
when
kept
in
(socially
compatible)
groups
of
at
least
two,
and
the
longevity
of
donkeys
(average
life
expectancy
27
years;
some
over
40;
Svendsen,
2008)
is
likely
to
mean
that
lifelong
pair-bonds
could
form.
Thus
disruption
of
stable
pair-bonds
has
the
potential
to
have
detrimental
effects
on
both
members
of
the
dyad.
5.
Conclusion
The
results
of
this
study
have
verified
anecdotal
evidence
that
pair-bonds
and
associations
in
domestic
donkeys
are
a
predominant
feature
of
their
social
organisa-
tion.
Where
pair-bonds
do
occur;
both
individuals
within
a
dyad
are
capable
of
individually
recognising
their
compan-
ion,
and
prefer
to
be
close
to
them.
Management
practices
which
avoid
stress
due
to
separation
of
donkeys
from
their
companion
are
likely
to
positively
influence
the
psycholog-
ical
and
physical
welfare
of
domestic
donkeys.
Acknowledgements
This
research
was
self-funded
by
the
corresponding
author,
and
formed
part
of
an
MSc
dissertation
for
the
Applied
Animal
Behaviour
and
Animal
Welfare
MSc
course
at
The
University
of
Edinburgh.
We
thank
the
SBDS
Trustees
for
permission
to
carry
out
the
project.
We
also
thank
Sarah
Maclaren,
the
sanctuary
volunteers
and
students
from
Bor-
ders
College
Animal
Care
course
for
their
assistance
with
donkey
handling.
Angela
Sibbald
(James
Hutton
Institute)
kindly
provided
us
with
Genstat
code
for
Nearest
Neigh-
bour
analysis
originally
written
by
David
Elston
of
BioSS.
SAC
is
supported
by
the
Scottish
Government
Rural
and
Environment
Science
and
Analytical
Services
Division.
References
Beja-Pereira,
A.,
England,
P.R.,
Ferrand,
N.,
Jor.n,
S.,
Bakhlet,
A.O.,
Abdalla,
M.A.,
Mashkour,
M.,
Jordana,
J.,
Taberlet,
P.,
Luikart,
G.,
2004.
African
origins
of
the
domestic
donkey.
Science
304,
1781.
Berghänel,
A.,
Ostner,
J.,
Schröder,
U.,
Schülke,
O.,
2011.
Social
bonds
pre-
dict
future
cooperation
in
male
Barbary
macaques,
Macaca
sylvanus.
Anim.
Behav.
81,
1109–1116.
Bøe,
K.E.,
Færevik,
G.,
2003.
Grouping
and
social
preferences
in
calves,
heifers
and
cows.
Appl.
Anim.
Behav.
Sci.
80,
175–190.
Boissy,
A.,
Le
Neindre,
P.,
1997.
Behavioural,
cardiac
and
cortisol
responses
to
brief
peer
separation
and
reunion
in
cattle.
Physiol.
Behav.
61
(5),
693–699.
Bosch,
O.J.,
Nair,
H.P.,
Ahern,
T.H.,
Neumann,
I.D.,
Young,
L.J.,
2008.
The
CRF
system
mediates
increased
passive
stress-coping
behavior
following
the
loss
of
a
bonded
partner
in
a
monogamous
rodent.
Neuropsy-
chopharmacology
34,
1406–1415.
Burden,
F.A.,
Du
Toit,
N.,
Hazell-Smith,
E.,
Trawford,
A.F.,
2011.
Hyper-
lipemia
in
a
population
of
aged
donkeys:
description,
prevalence,
and
potential
risk
factors.
J.
Vet.
Int.
Med.,
http://dx.doi.org/10.1111/
j.1939-1676.2011.00798.x
(article
first
published
online:
21.09.11).
Cameron,
E.Z.,
Setsaas,
T.H.,
Linklater,
W.L.,
2009.
Social
bonds
between
unrelated
females
increase
reproductive
success
in
feral
horses.
Proc.
Natl.
Acad.
Sci.
U.S.A.
106
(33),
13850–13853.
Carter,
C.S.,
Keverne,
E.B.,
2009.
4
The
Neurobioogy
of
Social
Affiliation
and
Pair
Bonding.
Hormones,
Brain
and
Behaviour,
second
ed.,
part
1,
pp.
37–166.
Chiyo,
P.I.,
Archie,
E.A.,
Hollister-Smith,
J.A.,
Lee,
P.C.,
Poole,
J.H.,
Moss,
C.J.,