Upload
rost-alius
View
219
Download
0
Embed Size (px)
Citation preview
7/25/2019 Galin y Thorn (1993) Unmasking Pain
1/17
lournal
of
Social
an d
Clinical
Psychology,
Vol.
12,
No.
2,
1993,
pp.
182-197
UNMASKING
PAIN: DETECTION OF
DECEPTION
IN
FACIAL
EXPRESSIONS
KAREN E. GALIN
VA
Medical
Center,
Pittsburgh,
PA
BEVERLY
E.
THORN
University
of
Alabama
Trained
judges
66
college
students)
distinguished
facial
expressions
of
subjects
in
genuine
pain
hand
in
ic e
water ,
masked
pain, posed
pain,
or
no
pain. Judges
were
given
facial movement
training
based
on
Facia l Action
Coding
System)
plus
limited
feedback
training,
feedback
only training,
or
no
training.
All
judges,
regardless
of
training,
were more ur te
in
detecting
genuine
pain
in
subjects
demonstrating
low
pain
tolerance
than
in
subjects
with
high
tolerance. Relative to
no
training,
feedback
training
enhanced
accuracy
in
identifying
posed
and
genuine pain,
whereas facial action
training
plus
feedback enhanced
accuracy
in
identifying
posed
pain.
Results
suggest
judges
n be
provided
with
information
about
facial movements to
distinguish
between
genuine
and d isto rted
pain
displays.
Pain ssessment
represents
challenge
to clinicians.
The fundamental
difficulty
involves the
private
n ture of
pain,
which n
only
be
revealed
by
what
th e
suffering
person
says
or
does
Fordyce,
1983 .
Several researchers
have
attempted
to
identify
the
behaviors
reliably
associated
with
pain.
Keefe
and Block
1982)
investigated
number of
behav iors assoc ia ted with
low-back
pain,
including guarded
move
ment,
bracing, rubbing, sighing,
and
grimacing.
Other researchers
e.g.,
Craig,
Hyde,
Patrick,
1991;
Craig
Patrick,
1985;
LeResche
Dworkin, 1988;
Patrick,
Craig,
Prkachin, 1986;
Prkachin
Mercer,
1989
have
studied
specific
facial
expressions
characteristic
of
pain.
Investigations using
the
Facial Action
Coding
System
FACS;
Ekman
Friesen,
1969)
indicate that
circumscribed
subset of
facial actions is
7/25/2019 Galin y Thorn (1993) Unmasking Pain
2/17
PAIN
AND
FACIAL
EXPRESSIONS
183
associated
with
pain,
although
there
is
variability
in their
expression
cross
individuals
(Craig
et
al.,
1991 .
LeResche and Dworkin
(1988)
note
variability
ross
patients
in
the
frequency,
duration,
and inten
sity
of
facial
movements
in
response
to
painful
clinical
examinat ion.
Hyde (1986)
suggests
the
existence
of constellation of
AUs
(Action
Units),
any
one of
which
might
be
displayed
by
n
individual
in
pain
p.
98
rather
than
unitary
or
prototypic
pain expression.
Variability
in
the
expression
of
facia l ac tions
may
be
influenced
by
display
rules
and/or
secondary
gain.
Display
rules
(i.e.,
norms
for
managing
emotional
expressions)
may
serve to
intensify,
deintensify,
mask,
or
neutralize facial
expressions
of
pain
(Ekman
Friesen,
1978).
Similarly,
individuals who
receive
reinforcement
fo r their
pain
behav
iors
may
display
different
facial
expressions
than
individuals
for
whom
pain
has
only negative
consequences.
Attention
(Turk,
Meichen-
baum,
Genest,
1983)
and
insurance
compensation
(Fordyce,
1983)
n serve
s
potent
reinforcers
that
may shape pain expressions.
Little
research h as b een done
to
help
clinicians
identify
distortions of
th e
pain
experience,
such s
willfully
minimizing
pain
displays
(masking)
or
exaggeration
of such behav io rs
(posing).
Craig
et al.
1991
have identified the
facial
movements associated
with
simulated
and
masked,
s
well
s
genuine,
pain.
However,
it is unknown
whether
clinicians'
knowledge
of
deceptive
pain expressions
results in more
ccur te ssessments
This
question
w s
addressed
in
the
present
study.
STUDY
METHOD
Thirty
male
and
30
female
volunteers
underwent
Cold-Pressor
pain
(Wolf
Hardy,
1941),
whereby
th e
subject's
hand
is
immersed in w ter
maintained
t 0-4
degrees
Celsius.
Subjects
were asked to endure th e
discomfort s
long
s
they
could . Af te r
five
minutes,
th e
procedure
w s
discontinued.
Subjects
rated their
pain intensity using
Numer ic
Rating
Scale
(Scott
Huskisson,
1976 upon
immers ion
in
th e
w ter
and
upon
withdrawal .
The
scale w s vertical
line
with the numbers
0-100
spaced evenly along
its
length,
with
adjective
ancho rs no
pain,
just
noticeable
pain,
moderate
pain,
and
excruciating pain,
corresponding
to
th e
numbers
0, 10, 50,
and 100. The mount
of t ime
7/25/2019 Galin y Thorn (1993) Unmasking Pain
3/17
184
GALIN
A ND
THORN
focus
on facial
expressions.
Subjects
were
alone
in
the
room
but
were
informed
that
they
were
being
videotaped.
Participants'
facial
expressions
were
videotaped
in
four
conditions:
a
baseline
(the
subject placed
the h an d in
room
temperature
water
for
two
minutes);
b
genuine pain
(standard
cold-pressor
procedure);
c
masked
pain
(standard
cold-pressor
procedure
but
S
instructed
to
express
s little
discomfort
as
possible);
and
(d)
posed
pain
(baseline
procedure,
but S
asked to
pose
n
expression
of
pain).
The
m a sk ed and
posed
conditions
were counterbalanced for order
of
presentation
cross
subjects.
A
completely
counterbalanced
design
w s sacrificed to
ensure
that
the
genuine
pain
condition
represented
spontaneous
behavior
(Lanzetta,
Cartwright-Smith,
Kleck,
1976).
Videotapes
of
15
male
and
15
female
participants
were
selected
randomly
for
Tape
1. The
remaining segments
were
used
for
Tape
2.
The
videotapes
of th e
conditions were
edited
so that the
tapes
c o nta in ed fou r 20 second
segments
of each
subject,
one
segment
from
each of th e
baseline
posed pain,
masked
pain,
and
genuine
pain
conditions.
For the
genuine
an d m ask ed
conditions
th e
segment began
25
seconds
befo re the
subject
withdrew
their
hand
(either
because
they
quit
or because the 5 m inu te
ceiling
w s
reached).
T he 20 second
baseline
and
posed
segments
began
after one minute of
filming.
The
order of
the
segments
on th e stimulus
tape
w s
randomized
with th e
limiting
condition
that
no
more
than
two
segments
of the s me
subject
would be
presented
consecutively.
The
coders
were two
experienced
FACS
analysts
w ho were
blind
to
the
pain
conditions
appearing
in
each
segment.
Prior
to
coding
th e
videotapes,
they
demonstrated
their
expertise
in the
FACS
by
passing
th e final test
developed
by
Ekman and Friesen
1978 .
O ne coder
identified
th e
A U s
in
every segment
on
Tape
1
using
FACS. The other
coder scored
25
of th e data to
provide
a
reliability
me sure
Using
n
occurrence
agreement
formula
intercoder
reliability
w s
.83. Occur
rence
agreement
w s
defined as
agreements
of
occurrence divided
by
agreements
plus disagreements
of occurrence
multiplied
by
100.
RESULTS
Action Units
appearing
at least 17 t imes in
th e
videotape
segments
were
used
in the
analyses.
In
addition
A U s with similar
muscle movements
7/25/2019 Galin y Thorn (1993) Unmasking Pain
4/17
PAIN
A N D
FACIAL
EXPRESSIONS
185
differed
according
to
conditions;
brow
lower
F l,19
7.73,
p
.05;
cheeks
raise lids
tighten,
F
1,19
13.35,
p
.01;
lip
orner
pull,
F l,19
5.21,
p
.05;
dimpler-lips
stretch
F l,19
8.49,
p
.01;
lips
part,
F l,19
5.55,
p
.05;
lips
tighten,
F l,19
9.30,
p
.01;
jaw
drop,
F l,19
4.58,
p
.05;
eyes
closed-blink,
F 3,58
10.93,
p
.0001.
Significantly
more
brow
lower
cheek
raise lids
tighten,
lips
tighten,
and
eyes
close blink
occurred
in
th e
posed
condition. The
baseline,
genuine,
and
masked
cond itions d id not
differ
significantly
fro m each
other in
terms of
the
frequencies
of these
movements The
posed
and
genuine
conditions
yielded
significantly
more
lip
orner
pull
than
the
baseline condition.
The
ma sk ed c on d itio n
did
not
significantly
differ
from the
posed,
genuine,
or
baseline
conditions
in
the
frequency
of
lip
orner
pull.
Significantly
more
dimpler-lips
stretch
lips
part,
and
jaw
drop
also
occurred
in
th e
posed
c on ditio n th an
in
the
m asked and
baseline
condit ions.
The
genuine
cond ition d id
not
significantly
differ
from the
posed,
masked
or baseline conditions in terms of these
variables.
M e a n s
used
in
th e
pairwise
comparisons
are listed
in
Table
1.
Since
heterogeneity
of variance
w s
highest
in the
posed
condition
and
because
identification of differences
among
genuine
pain,
masked
pain,
and baseline w s
necessary
in
order
to
prepare
the
training
package
used
in
Study
2,
n additional
repeated
me sures
MANOVA
w s
performed
withou t
posed pain.
Cheeks
raise lids
tighten
F l,29
5.40,
p
.05 ,
lip
orner
pull F l,29
5.61,
p
.05 , dimpler-lips
stretch
F l,29
4.96,
p
.05 ,
and
lips
part F l,29
4.78,
p
.05
distinguished
among
th e conditions.
Lips
part
and
dimpler-lips
s tre tch occurred
significantly
more
often
in the
genuine
condition.
The
masked
and
baseline
conditions
did not differ
significantly
in
th e
frequency
of these
movements
In
addition
significantly
more cheeks raise lids
tighten
and
lip
orner
pull
occurred
in
th e
genuine
condition than in th e
baseline condition.
The m asked cond ition
yielded
a tre nd to w a rd more
lip
orner
pull
than
the baseline condition
p
10
Baseline
yielded
a
tren d to w ard
more
lips
tighten
than the m asked cond ition
p
.10 .
Means used in th e
pairwise
comparisons
re listed
in
Table
1.
DISCUSSION
A s
predicted,
a subset of action units
distinguished
among
th e
conditions
i.e.,
brow
lower
cheeks
raise
or
lids
tighten,
lips
tighten,
7/25/2019 Galin y Thorn (1993) Unmasking Pain
5/17
186
GALIN
AND
THORN
TABLE 1
Means fo r Pattern of
Action
Units
Function
of Condition
TUKEY
TUKEY
CONDITION
MEAN
SD
GROUPINGb
GROUPING^
Action Unit
4 Brow Lower
Posed
34
.70
A
Genuine
10 .36
B
A
Masked
03
18
B
A
No
pain
02
13
B
A
Action
Unit 6 7 Cheeks Raise Lids
Tighten
Posed
158
2 57
A
Genuine
.51
1 15
B
A
Masked
27
74
B
A B
No
pain
03
18
B
B
Action
Unit 12:
Lip
Corner
Pulld
Posed .37
66
A
Genuine
29 .53
A
A
M asked
20
45
A B
A B
No
pain
05
22
B
B
Action Unit 14 20:
Dimpler-Lips
Stretch
Posed
53
70
A
Genuine
34
63
A B
A
Masked 10
36
B B
No
pain
14 39
B
B
Action Unit
17
Chin
Raise
Posed
23 .54
A
Genuine
.07
25
A
A
Masked
.05
22
A
A
No
pain
05 22
A
A
Action Unit 23
Lips
ighten
Posed
36
66
A
No
pain
12 .38
B
A
Genuine .05
22
B
A
Masked
00 .00
B
A
Action
Unit 251
Lips
Part
Posed 54 89
A
Genuine
41 77
A B
A
Masked
17
46 B
B
No
pain
.15
45 B B
Action Uni t 26:
Jaw
Drop
Posed 46 .75
A
Genuine
24
63
A B
A
Masked
15
52
B
A
No
pain
12 42
B A
Ac tion Un it 43^15:
Eyes
Close Blink
Posed
5 30
3 09 A
Genuine
3 68 2 47
B A
Masked
3 81
3 01
B
A
No
pain
3 29
2 27
B A
Action
Unit
1 2 Inner
and
Outer
Brow
Raise
Posed
44
.89
A
7/25/2019 Galin y Thorn (1993) Unmasking Pain
6/17
PAIN
AND
FACIAL
EXPRESSIONS
187
TABLE
1.
Means for
Pattern o f Ac tio n Units s
Funct ion of Condition
continued
TUKEY
TUKEY
CONDITION
MEAN SD
GROUPINGb GROUP ING
Action
Unit
19:
Tongue
Out
Posed
.20 .55 A
Genuine .19
.57
A
A
No
pain
.07
.25
A
A
Masked
.03 18
A
A
Action Unit 24
Lip
Press
Posed
.27
.49 A
Masked
.20 .52 A
A
Genuine 10
.30
A A
No
pain
10
.36
A A
Action Uni t 41 :
Lids
Droop
Posed
.46
.45
A
Genuine 24
.81
A
A
Masked
.15
.56
A
A
No
pain
.12
.34
A
A
*SD
Standard
deviation.
bPost-hoc
Tukcy pairwise
comparisons
fo r
posed, genuine,
no
pain,
and masked conditions.
Mean s w ith th e
s me letter
re
not
significantly
different.
cPost-hoc
Tukcy pairwise comparisons
fo r
genuine,
no
pain,
a nd ma sk ed con ditio ns
only
Posed
condition
is
removed
from
analysis .
Means w ith the
s me letters re not
significantly
different
Comparisons
of
me ns
fo r
mask ed and no
pain
conditions
yield
trend toward
significance y
.10 .
stretch
and
lips part
than masked
pain
and baseline and
also
more
lip
corner
pull
and cheeks raise
or
lids
tighten
than baseline
in
analysis
with
genuine
pain,
masked
pain,
and baseline
only .
Although
there
is
overlap
in
th e facial
movements
associated with
pain
between this
and o th er
investigations,
variability
between
studies
is also
evident. For
example,
in
Craig
and
Patrick s
1985
experiment
and
in
th e
present study,
cheeks
raise-lids
tighten,
lip
corner
pull,
and
lips
part
occurred
more
frequently during
cold-pressor
exposure
than
during
baseline. On
the
other
hand,
upper
lip
raise,
jaw
drop,
and
eyes
close-blink
occurred
significantly
more often in
genuine
pain
than
baseline
in
only
th e
Craig
and
Patrick
1985
study.
It
is
possible
that th e
videotape
segments
selected for
coding
in
the
two
studies
resulted
in
th e differences
in
action units. The
present
study
used
segment
just prior
to
quit-point.
This
segment
w s
selected
because
our
observations
during
pilot
work
suggested
that
subjects
7/25/2019 Galin y Thorn (1993) Unmasking Pain
7/17
188
GALIN
AND
THORN
Patrick
1985)
because
we
wanted to
eliminate
other
expressions
concomitant with
pain,
such
s
startle
response
or
orienting
behavior
It is
also
possible
that th e
participants
w reness of the
videotaping
procedure
inhibited
their
expression
of
pain during
th e
genuine
condi
tion in
th e
present
experiment.
Kleck
et
al
(1976)
demonstrated that
individuals
con tro l th eir
facial
expressions
more
when
they
re
ob
served than w hen
they
re alone
Yet,
participants
in several other
investigations (e.g., Craig
et
al.,
1991;
Craig
Patrick
1985;
Patrick et
al.,
1986)
were also
in fo rmed tha t
they
were
being
videotaped
and
still
exhibited
variety
of
facial
movements
in
response
to
noxious
stimula
tion The
variat ions found
ross
experiments
suggest
that
facial ex
pressions
of
pain
re influenced
by
number of variables
On y
two
studies
have
compared
genuine
pain,
masked
pain,
posed
pain,
and
baseline
(the
present
study
and
Craig
et
al.,
1991
and
they
suggest
sim ilar conclusions about
the
facial
movements
associated with
simulation
In
both
studies
participants
were
capable
of
masking
expressions
of
pain,
and
they displayed
more
varied
and
intense
facial
activity
when
posing
pain
than
in
th e other conditions
Brow
lower
and
cheeks
raise
occurred
more
frequently
in
posed pain
than
genuine pain
in
both
studies
Thus
it
appears
that facial
expressions
of m ask ed and
posed pain
n be
described
using
FACS It
remains
unclear however
whether
observers
n
be trained
to
distinguish among genuine,
masked
and
posed
pain
displays.
STUDY 2
The
purpose
of
Study
2
w s to determine
whether
training,
using
either
repeated presentations
of
th e
taped
segments
of
th e different
pain
expressions
(masked,
posed, genuine)
or
using
those
constellations
of
facial
action units
found
to
distinguish
among
conditions
is more
useful
than
no
training
condition
in
improving
accuracy
in
detection
of
simulated
expressions
of
pain.
It
w s
expected
that
judges accuracy
in
distiguishing genuine,
masked
and
posed pain
on
pretraining
task
(Trial
1
would not differ
significantly
ross
groups
Groups
w h o receiv ed either facial move
ment
training
based on FACS
plus
lim ited feedback
training
or
feedback
only training
were
expected
to
improve
in
Trial
2,
with th e
greatest
increase
in
accuracy occurring
in th e facial
movement
training
7/25/2019 Galin y Thorn (1993) Unmasking Pain
8/17
PAIN
A ND
FACIAL
EXPRESSIONS
189
pain
than
other
pain
st tes It
w s
expected
that
judges
would be mor
ccur te
in
detecting genuine
and
masked
pain
in
subjects
w ho
demonstrate
low
pain
tolerance
because these
participants
w r
expected
to
be th e
most
expressive.
Subjects
in
th e
masked
condition
w ho had
low
pain intensity ratings
wh e n
they
withdrew
their
hands
from
th e w ter w r
expected
to be successful in
masking
their
discomfort,
and
thus
th e
judges
w r
expected
to believe that
they
w r
not
in
pain.
M E T H O D
Sixty-six
female
psychology
students
tested in
groups
of four
to
eight,
served
s
judges
in
Study
2.
They
v iew e d the
s m
videotape
used
in
Study
1.
Accuracy
in
labeling
the four conditions
genuine
pain,
masked
pain,
posed pain,
and
no
pain
on
th e
videotape
segments
served s th e
dependent
variable
Accuracy
w s determined
by
frequency
count
of the
number
of
videotape
segments correctly
identified
during
no
pain,
genuine
pain,
masked
pain,
and
posed pain.
Accuracy
w s calculated
separately
for th e 15 male and 15
female
faces
on the
videotape.
Thus
perfect
accuracy
would
yield
score of 15
correct
for each
condit ion and
gender,
and
c ha nc e le ve ls
of
accuracy
would
be
t 3 75
segments
labeled
correctly.
The
judges
w r
assigned
randomly
to
training group
control,
facial movement
training
plus
l imited feedback
training,
or feedback
only
training
and
trainer The
primary experimenter
Trainer 1
and
n
advanced
undergraduate
psychology
student
Trainer
2
served
s
th e
trainers Trainer
2
w s blind
to
the
hypotheses
of th e
study.
Each
trainer
presented
information
to
one half of th e
judges
in each
of th e
three
judging
conditions
First
all
judges
viewed
Tape
1
see
Study
1
without
audio
accompa
niment
Following
each of 120
segments,
judges
had five seconds to
indicate
whether
no
pain,
posed
pain,
masked
pain,
or
genuine pain
w s
portrayed.
Next
the
groups
received
either th e
control or on of th e
experimen
ta l interventions
Following
their
viewing
of
Tape
1,
control
group
judges
watched
videotapes
that
discussed
th e n ture
and
tre tment of
pain.
Judges
in th e
feedback
training group
viewed
Tape
1
again
120
total
segments
and
w r told
which
condition
no
pain,
posed,
7/25/2019 Galin y Thorn (1993) Unmasking Pain
9/17
190
GALIN
AND
THORN
expressions
characteristic of
genuine,
masked
and
posed
pain
as
derived
from
Study
1.
Specifically,
the
training
provided
to the facial
m o v e m e n t
training
group
had
three
components.
First,
th e
trainers
presented
detailed
descriptions
of the
changes
in
th e
face
associated
with
masked
posed,
and
genuine pain. Photographs
from the
FACS
manua l
were
used
to
il lus tr ate these
facial movements
Next
the
experimenter
taught
each
judge
to m ake
th e
movements
associated
with
masked
posed,
and
genuine
pain
on her own
face.
Finally,
th e
judges
viewed
segments
from
Tape
1,
during
which
they
w r
given
feedback
regarding
which
condition
w s
displayed.
In
order
that the
training
t ime
provided
to
all
groups
would be
equivalent,
th e to ta l number
of
segments
viewed
by
th e
facial movement
training
group
w s
limited
to
20.
Thus
subjects
in
th e
facial movement
training
group
w r
given
only
limited
feedback
compared
to Ss
in
th e
feedback
training
group.
Following
a
10 minute
recess
all
judges
viewed
Tape
2,
which
contained th e
videotape
segments
of th e
15
male
and
15
female
participants
from
Study
1
that
w r
not
included
on
Tape
1. After each
20 second
segment
on
Tape
2,
judges
had five seconds
to
indicate
whether no
pain,
posed,
masked
or
genuine
pain
w s
displayed.
RESULTS
ANALYSIS OF
VARIANCE
A 4 x
2
x
3
analysis
of
variance
ANOVA)
w s
used
to ss ss
th e effects
of condition
no
pain,
posed,
masked
and
genuine),
trial
1
and
2 ,
and
training group
control,
feedback
facial
movement)
on
judges
accu
racy
in
labeling
the
videotape
segments.
While
other variables
in th e
study gender
of face on
videotape
and
trainer)
did
interact in small
ways
w ith the
condition trial,
and
training
group
variables
relative
to
th e effects of
these
variables
th e
differences
w r
only
slight
and
th e
conclusions
did
not
change.
Therefore,
only
the
3-way
interaction
is
reported.
The
condit ion
x
trial
x
training
interaction
w s
significant,
F 6,387)
5.45,
p
genuine).
Feedback
trained Ss demonstrated a
pattern
of
accuracy
whereby
posed
>
genuine
masked
no
pain
in
Trial 2. Facial
Action trained
Ss
showed
a
pattern
of
accuracy whereby
no
pain
posed
>
masked
genuine
in
Trial 2.
As
hypothesized,
a
significant
main
effect
for condition
w s
found
F 3,387)
111.64,
p
.0001.
Judges
were
significantly
more accurate
in
labeling
the
posed
and no
pain segments
than
the m a sk ed an d
genuine
segments. They
were also
significantly
more accurate
in
labeling
th e
masked
segments
than
the
genuine
segments.
MULTIPLE
REGRESSION
It w s
expected
that
judges
would be
more accurate
in
detecting
genuine
an d m ask ed
pain
in
subjects
w ho
report
high
pain
intensity
at
quit-point
and demonstra te
lo w
pain
tolerance.
To test
this
hypothesis,
a
simultaneous
regression procedure
w s
used
to test the in te rac tion
between
final
pain
ratings
and
pain
tolerance.
The number of
judges
w ho
correctly
labe led the
segment
w s the
dependent
variable.
For th e
genuine
segments
in
Trial 1
Tape
1,
viewed
prior
to
training),
the
interaction
w s not
significant,
F 3,26
.05.
The test of
7/25/2019 Galin y Thorn (1993) Unmasking Pain
11/17
192
GALIN
AND
THORN
9
7
6
c
5
i
4
3
A
C
.sea
M35 rJ
Genune
Condition
6
5
r
2
4
i^
A
3 r
c
Baseline
Posed
rJlas^ ed
Genuine
Condition
Group
A
=
Control
group
Croup
B
=
Feedback
training gToup
Group
C
=
Facial
movement
training group
FIGURE 1. Mean
accuracy
as
a function
of
group,
condition,
and
trial.
F 3,26
=
.484. The test of the overa ll
mode l
was
significant,
with
an
R2
of
.300,
p
J
>rrJ