It's not just my fault: Neural correlates of feedback processing in solo and joint action

Abstract

People often coordinate their actions with others' in pursuit of shared goals, yet little research has examined the neural processes by which people monitor whether shared goals have been achieved. The current study compared event-related potentials elicited by feedback indicating joint errors (resulting from two people's coordinated actions) and individual errors (resulting from one's own or another person's observed actions). Joint errors elicited a reduced feedback-related negativity (FRN) and P3a relative to own errors, and an enhanced FRN relative to observed errors. In contrast, P3b amplitudes did not differ between joint and individual errors. These findings indicate that producing errors together with a partner influences neural activity related to outcome evaluation but has less impact on activity related to the motivation to adapt future behavior.
Copyright © 2015. Published by Elsevier B.V.

Full text

Biological
Psychology
111
(2015)
1–7
Contents
lists
available
at
ScienceDirect
Biological
Psychology
jo
ur
nal
home
p
age:
www.elsevier.com/locate/biopsycho
It’s
not
just
my
fault:
Neural
correlates
of
feedback
processing
in
solo
and
joint
action
Janeen
D.
Loehra,c,∗,1,
Dimitrios
Kourtisb,c,1,
Inti
A.
Brazilc,d
aDepartment
of
Psychology,
University
of
Saskatchewan,
9
Campus
Drive,
Saskatoon,
SK,
Canada
S7N
5A5
bDepartment
of
Experimental
Psychology,
Ghent
University,
Henri
Dunantlaan
2,
9000
Ghent,
Belgium
cDonders
Institute
for
Brain,
Cognition
and
Behaviour,
Radboud
University
Nijmegen,
PO
Box
9102,
6500
HC
Nijmegen,
The
Netherlands
dPompestichting,
Weg
door
Jonkerbos
55,
6532
CN
Nijmegen,
The
Netherlands
a
r
t
i
c
l
e
i
n
f
o
Article
history:
Received
13
January
2015
Received
in
revised
form
6
August
2015
Accepted
7
August
2015
Available
online
11
August
2015
Keywords:
Action
monitoring
Joint
action
Interpersonal
coordination
Event-related
potentials
FRN
a
b
s
t
r
a
c
t
People
often
coordinate
their
actions
with
others’
in
pursuit
of
shared
goals,
yet
little
research
has
examined
the
neural
processes
by
which
people
monitor
whether
shared
goals
have
been
achieved.
The
current
study
compared
event-related
potentials
elicited
by
feedback
indicating
joint
errors
(result-
ing
from
two
people’s
coordinated
actions)
and
individual
errors
(resulting
from
one’s
own
or
another
person’s
observed
actions).
Joint
errors
elicited
a
reduced
feedback-related
negativity
(FRN)
and
P3a
rel-
ative
to
own
errors,
and
an
enhanced
FRN
relative
to
observed
errors.
In
contrast,
P3b
amplitudes
did
not
differ
between
joint
and
individual
errors.
These
findings
indicate
that
producing
errors
together
with
a
partner
influences
neural
activity
related
to
outcome
evaluation
but
has
less
impact
on
activity
related
to
the
motivation
to
adapt
future
behaviour.
©
2015
Elsevier
B.V.
All
rights
reserved.
Efficient
and
flexible
behaviour
often
requires
that
people
mon-
itor
the
outcomes
of
their
actions
to
determine
whether
their
goals
have
been
achieved.
Research
investigating
the
cognitive
and
neural
mechanisms
underlying
action
monitoring
(see
Ullsperger,
Danielmeier,
&
Jocham,
2014a,
for
a
review)
has
focused
predom-
inantly
on
how
people
monitor
individual
action
outcomes,
i.e.,
outcomes
of
a
single
person’s
actions.
However,
humans
are
highly
social
beings;
a
significant
portion
of
our
behavioural
repertoire
is
obtained
through
social
interactions
that
require
sensitivity
to
the
outcome
of
both
our
own
and
others’
actions.
Recent
stud-
ies
have
begun
to
unravel
the
impact
of
social
context
on
action
monitoring,
typically
by
examining
how
people
process
observed
individual
outcomes
in
tasks
that
require
them
to
monitor
their
own
and
another
person’s
performance
on
successive
trials
(Koban
&
Pourtois,
2014).
Little
research
has
examined
action
monitoring
in
social
contexts
that
require
people
to
actively
coordinate
their
actions
with
each
other
in
pursuit
of
a
joint
action
outcome
(e.g.,
scoring
a
goal
in
hockey
as
a
result
of
multiple
players
passing
the
∗Corresponding
author
at:
Department
of
Psychology,
University
of
Saskatchewan,
9
Campus
Drive,
Saskatoon,
SK,
Canada
S7N
5A5.
Tel.:
+1
306
966
6082;
fax:
+1
306
966
6630.
E-mail
addresses:
janeen.loehr@usask.ca
(J.D.
Loehr),
Dimitrios.Kourtis@UGent.be
(D.
Kourtis),
i.brazil@donders.ru.nl
(I.A.
Brazil).
1Shared
first
authorship.
puck
back
and
forth;
see
Sebanz,
Bekkering,
&
Knoblich,
2006).
Joint
outcomes
present
a
challenge
for
the
action
monitoring
system,
as
each
person
involved
has
only
partial
control
over
the
outcome
despite
actively
adapting
their
own
actions
to
other
people’s.
Fur-
thermore,
joint
outcomes
often
take
the
form
of
external
feedback
that
provides
information
about
the
joint
performance
as
a
whole
and
is
available
only
at
the
end
of
the
entire
shared
action
sequence.
Nevertheless,
people
are
typically
able
to
identify
errors
that
have
been
committed,
attribute
them
to
a
single
person
or
to
the
group
as
a
whole,
and
adapt
their
future
behaviour
accordingly.
Researchers
have
identified
several
event-related
potential
(ERP)
components
associated
with
evaluating
individual
action
outcomes
and
adapting
ongoing
behaviour
based
on
external
feedback.
The
feedback-related
negativity
(FRN)
is
an
anterior,
negative-going
ERP
that
peaks
∼250
ms
after
feedback
(Gehring
&
Willoughby,
2002;
Miltner,
Braun,
&
Coles,
1997).
FRN
amplitudes
are
larger
following
negative
feedback
indicating
an
unfavourable
outcome
(e.g.,
an
error)
compared
to
positive
feedback
(Walsh
&
Anderson,
2012).
The
FRN
is
thought
to
reflect
an
initial
evaluation
of
the
outcome
as
better
or
worse
than
expected
(Nieuwenhuis,
Holroyd,
Mol,
&
Coles,
2004)
or
as
simply
unexpected
(Ullsperger,
Fischer,
Nigbur,
&
Endrass,
2014b).
The
FRN
is
often
followed
by
a
P3,
a
positive-going
potential
with
two
sub-components:
an
earlier,
anterior
P3a
and
a
later,
posterior
P3b.
The
P3a
is
thought
to
reflect
an
orienting
response
related
to
the
initial
evaluation
of
stimuli
as
task-relevant
(Polich,
2007).
The
P3b
is
thought
to
reflect
internal
http://dx.doi.org/10.1016/j.biopsycho.2015.08.004
0301-0511/©
2015
Elsevier
B.V.
All
rights
reserved.

2
J.D.
Loehr
et
al.
/
Biological
Psychology
111
(2015)
1–7
Fig.
1.
Schematic
illustration
of
the
sequence
production
task
in
the
individual
and
joint
settings.
Following
instructions
and
fixation,
participants
heard
a
series
of
isochronous
pacing
clicks
(illustrated
by
eighth
note
symbols)
and
then
produced
a
sequence
of
tones
(illustrated
by
combined
button
press
and
eighth
note
symbols,
labelled
P1
and
P2
for
Participants
1
and
2,
respectively).
After
producing
the
last
tone,
participants
received
feedback
indicating
whether
the
sequence
they
produced
matched
the
pace
set
by
the
isochronous
clicks.
decision-making
processes
that
facilitate
appropriate
behavioural
responses
to
task-relevant
stimuli
(Nieuwenhuis,
Aston-Jones,
&
Cohen,
2005).
Given
the
ubiquity
of
joint
actions
in
everyday
life
and
the
need
to
establish
healthy
brain
functioning
during
social
interactions
to
better
understand
social
disorders
such
as
psychopathy
(Brazil
et
al.,
2011;
de
Bruijn,
2012),
it
is
critical
to
examine
how
the
neural
mechanisms
underlying
action
monitoring
are
modulated
by
the
social
context
in
which
actions
are
performed.
The
cur-
rent
study
compared
ERPs
elicited
by
jointly
committed
errors
to
ERPs
elicited
by
individual
errors
committed
by
oneself
or
another
person.
We
predicted
that
joint
errors
would
elicit
reduced
ERP
amplitudes
relative
to
one’s
own
errors
but
enhanced
ERP
ampli-
tudes
relative
to
another
person’s
errors,
based
on
previous
work
showing
that
reduced
control
over
action
outcomes
reduces
both
FRN
and
P3
amplitudes
(e.g.,
Li,
Han,
Lei,
Holroyd,
&
Li,
2011;
Li
et
al.,
2010;
Yeung,
Holroyd,
&
Cohen,
2005)
and
that
observing
errors
elicits
reduced
FRN
amplitudes
compared
to
producing
them
(e.g.,
Bellebaum,
Kobza,
Thiele,
&
Daum,
2010;
Yu
&
Zhou,
2006).
1.
Method
1.1.
Participants
Twenty
two
adults
(8
male,
4
left-handed,
mean
age
=
24.23,
SD
=
3.12)
participated
in
the
study.
Participants
were
recruited
in
pairs
without
regard
for
specific
gender
combinations.
Of
the
11
pairs,
4
pairs
consisted
of
two
females,
1
pair
consisted
of
two
males,
and
6
pairs
were
mixed-gender.
All
participants
provided
written
informed
consent
according
to
procedures
reviewed
by
the
medi-
cal
ethics
committee
at
Radboud
University
Nijmegen.
Participants
were
compensated
with
D
30
for
their
participation.
1.2.
Design
and
procedure
In
order
to
compare
ERPs
elicited
by
individual
vs.
jointly
com-
mitted
errors,
we
employed
a
sequence
production
task
that
could
be
performed
either
alone
or
in
coordination
with
a
partner.
Specif-
ically,
participants
were
asked
to
produce
sequences
of
4
or
6
tones
that
matched
the
pace
set
by
an
initial
series
of
isochronous
clicks
(see
Fig.
1).
Participants
produced
the
tone
sequences
in
two
sett-
ings:
individual
and
joint.
In
the
individual
setting,
each
member
of
the
pair
produced
the
tone
sequences
alone
while
the
other
member
of
the
pair
sat
quietly
beside
them.
Each
sequence
in
the
individual
setting
therefore
elicited
ERP
responses
to
own
action
outcomes
for
the
participant
who
produced
the
tone
sequences,
and
to
observed
action
outcomes
for
the
participant
who
observed
the
other
person
produce
the
tone
sequences.
In
the
joint
setting,
the
two
participants
alternated
button
presses
so
as
to
produce
the
tone
sequences
together.
Sequences
produced
in
the
joint
setting
elicited
ERP
responses
to
joint
action
outcomes
for
both
partici-
pants.
During
the
experiment,
participants
sat
next
to
each
other
on
the
same
side
of
a
table.
A
computer
screen
was
centered
between
them,
approximately
80
cm
from
the
edge
of
the
table.
Each
partic-
ipant
had
a
Logitech
Gamepad
F310
game
controller
aligned
with
their
right
hand,
approximately
20
cm
from
the
edge
of
the
table.
The
game
controllers
were
modified
to
include
pressure
sensitive
buttons
(2
cm
diameter)
that
registered
presses
without
providing
auditory
feedback.
Each
button
press
triggered
a
1000
Hz
sinusoidal
tone
(100
ms
duration;
20
ms
rise/fall
time;
sound
pressure
level
70
dB).
The
initial
series
of
isochronous
clicks
was
produced
in
a
snare
drum
timbre.
Tones
and
clicks
were
presented
via
speak-
ers
placed
on
either
side
of
the
computer
screen.
Stimuli
were
presented
using
Presentation
software
(Neurobehavioral
Systems,
Inc.;
Albany,
CA),
which
also
recorded
participants’
button
presses.
Participants
were
fitted
with
EEG
caps
and
then
performed
two
practice
trials
(one
for
the
individual
setting
and
one
for
the
joint
setting)
during
which
the
experimenter
controlled
the
presentation
of
the
events
that
comprised
a
trial
and
explained
the
task.
Partic-
ipants
then
completed
a
training
phase
(two
blocks
of
18
trials)
and
a
test
phase
(12
blocks
of
30
trials)
with
the
timing
described
below.
At
the
beginning
of
each
block,
instructions
presented
on
the
computer
screen
indicated
the
sequence
length
and
whether
participants
were
to
produce
the
tones
alone
or
together.
Blocks
alternated
between
individual
and
joint
settings,
the
order
of
which
was
held
constant
through
both
the
training
and
test
phases
and
was
counterbalanced
across
pairs.
Sequence
length
was
either
4
or
6
tones,
held
constant
for
a
given
block
and
randomly
determined
with
the
constraint
that
half
of
the
blocks
in
each
setting
were
of
length
4
and
the
other
half
of
length
6.
The
person
producing
the
sequence
(individual
setting)
or
the
first
tone
in
the
sequence
(joint
setting)
was
randomly
determined
on
each
trial
with
the
constraint
that
one
participant
produced
or
started
the
sequence
on
half
of
the
trials
and
the
other
participant
produced
or
started
the
sequence
on
the
other
half
of
the
trials.
Each
trial
consisted
of
the
following
sequence
of
events,
shown
in
Fig.
1.
A
cue
indicating
which
person
was
to
produce
the
tone
sequence
(individual
setting)
or
the
first
tone
in
the
sequence
(joint
setting)
appeared
on
a
black
computer
screen
for
2000
ms.
The
cue
consisted
of
a
cartoon
face
with
two
arms,
one
of
which
was
colored
red
to
indicate
that
the
person
on
that
side
of
the
table
should
pro-
duce
or
begin
the
sequence.
A
white
fixation
cross
then
appeared
and
remained
on
the
screen
until
the
last
tone
of
the
sequence
was
produced.
Three
pacing
clicks
were
presented
at
500
ms
intervals
(beginning
500
ms
after
the
onset
of
the
fixation
cross).
Participants
were
instructed
to
produce
the
tone
sequence
while
maintaining
the
pace
set
by
the
clicks.
After
the
last
tone
was
produced,
a
black
screen
was
presented
for
700
ms,
followed
by
feedback
indicating
whether
the
participants
had
successfully
maintained
the
required
pace.
The
feedback
was
presented
for
700
ms,
after
which
a
black

J.D.
Loehr
et
al.
/
Biological
Psychology
111
(2015)
1–7
3
screen
was
presented
for
700
ms
before
the
next
trial
began.
The
feedback
consisted
of
a
green
or
red
Euro
symbol
indicating
cor-
rect
or
incorrect
performance,
respectively.
Participants
were
told
that
they
would
each
gain
or
lose
D
0.05
from
the
initial
fee
of
D
25
per
person
for
every
sequence
they
produced
with
correct
or
incor-
rect
timing,
respectively,
regardless
of
who
produced
or
started
the
sequence
on
a
given
trial.
In
reality,
all
participants
were
paid
D
30
at
the
end
of
the
experiment.
The
accuracy
of
participants’
performance
on
each
trial
was
determined
based
on
the
average
interonset
interval
(IOI)
produced
by
the
participant(s)
on
that
trial.
A
trial
was
deemed
correct
if
the
average
IOI
fell
within
a
window
around
500
ms.
The
window
size
was
set
to
60
ms
at
the
beginning
of
the
experiment
(i.e.,
sequences
were
considered
correct
if
the
average
IOI
fell
within
500
±
30
ms).
At
the
end
of
each
block,
the
window
size
increased
by
10
ms
if
more
than
30%
of
the
sequences
in
the
block
were
produced
with
incorrect
timing.
Otherwise,
the
window
size
decreased
by
10
ms
(until
the
minimum
window
size
of
20
ms
was
reached).
The
win-
dow
size
was
adjusted
separately
for
the
individual
and
joint
setting
blocks.
Only
the
last
12
of
the
18
training
trials
that
comprised
the
first
block
in
each
setting
were
included
in
the
first
window
size
adjustment.
This
procedure
resulted
in
mean
error
rates
of
28.08%
(SD
=
8.17%)
in
the
individual
setting
and
32.42%
(SD
=
7.08%)
in
the
joint
setting
(before
outliers
were
removed
as
described
in
Sec-
tion
1.4).
The
difference
in
error
rates
between
conditions
was
not
significant,
t(10)
=
1.89,
p
=
0.089.
1.3.
EEG
acquisition
EEG
was
recorded
using
32
active
electrodes
(Acticap,
Brain
Products
GmbH,
Germany)
per
participant,
arranged
according
to
an
extended
version
of
the
10–20
system
at
Fz/3/4/7/8,
FCz/1/2/5/6,
Cz/3/4,
CPz/1/2/5/6,
Pz/3/4/7/8,
Oz/1/2,
and
T7/8.
All
electrodes
were
referenced
to
the
left
earlobe.
Vertical
and
horizontal
eye
movements
were
monitored
using
bipolar
electrooculography
(EOG)
electrodes
positioned
above
and
beneath
the
right
eye
and
at
the
outer
canthi
of
both
eyes.
Impedance
was
kept
below
10
k!.
EEG
and
EOG
signals
were
amplified
within
a
bandwidth
of
0.05–100
Hz
and
digitized
with
a
sampling
frequency
of
1000
Hz.
1.4.
Data
processing
For
the
behavioural
data,
inter-tap
intervals
(ITIs)
were
calcu-
lated
between
participants’
taps,
starting
with
the
interval
between
the
last
pacing
tone
and
the
first
tap.
Trials
(279/3960
=
7.05%)
were
excluded
from
analysis
if
they
included
one
or
more
ITIs
that
was
greater
or
less
than
3
standard
deviations
from
the
mean
of
all
intervals
produced
by
a
given
pair.
This
left
an
average
of
20.64
(SD
=
10.37)
error
trials
and
64.32
(SD
=
13.60)
correct
trials
pro-
duced
by
one
participant
or
the
other
in
the
individual
setting,
and
an
average
of
22.86
(SD
=
7.46)
error
trials
and
59.50
(SD
=
7.94)
correct
trials
started
by
one
participant
or
the
other
in
the
joint
setting2.
See
the
Supplementary
material
for
an
analysis
of
IOIs
by
setting
and
accuracy.
EEG
data
processing
was
performed
off-line
using
Brain
Vision
Analyzer
software
(V2.01.3931,
Brain
Products
GmbH,
Germany).
EEG
data
were
re-referenced
to
the
mean
of
both
earlobe
electrodes.
Ocular
artefacts
were
removed
using
Independent
Component
Analyses
(Jung
et
al.,
2000).
The
data
were
filtered
using
a
high-
and
low-pass
filters
of
0.05
Hz
(24
dB/oct)
and
30
Hz
(24
dB/oct)
2Data
were
initially
analyzed
separately
by
which
participant
started
the
sequence
in
the
joint
setting.
Because
there
were
no
differences
based
on
who
started
the
joint
sequence
in
any
of
the
analyses,
all
ps
>
0.25,
we
collapsed
across
starting
person
in
all
analyses
reported
here.
to
remove
slow
drifts
and
excessive
noise,
respectively.
The
cor-
rected
EEG
data
were
segmented
into
epochs
from
200
ms
before
to
700
ms
after
feedback
onset.
Individual
trials
were
removed
if
they
contained
further
artefacts
induced
by
head,
body,
or
arm
move-
ments,
as
indicated
by
a
difference
exceeding
100
!V
between
the
maximum
and
minimum
value
within
a
segment.
One
participant’s
data
were
removed
from
analysis
because
artefact
rejection
left
fewer
than
three
error
trials
in
the
individual
setting
1.5.
Data
analysis
We
compared
ERP
responses
to
own,
joint,
and
observed
errors.
Average
EEG
waveforms
were
calculated
separately
for
each
partic-
ipant
and
error
type.
The
200
ms
before
feedback
onset
was
used
as
the
baseline
period.
Difference
waves
were
computed
on
individual
averages
by
subtracting
the
EEG
waveforms
elicited
by
correct
feed-
back
from
the
waveforms
elicited
by
error
feedback
(see
Fig.
S1
in
the
Supplementary
material
for
non-differenced
waveforms).
The
FRN
was
calculated
on
this
difference
wave
using
a
peak-to-peak
analysis
(Holroyd,
Nieuwenhuis,
Yeung,
&
Cohen,
2003;
Pfabigan,
Alexopoulos,
Bauer,
&
Sailer,
2011;
Picton
et
al.,
2000)
in
which
the
mean
voltage
amplitude
over
a
10
ms
window
around
the
pos-
itive
peak
that
preceded
the
FRN
was
subtracted
from
the
mean
voltage
amplitude
over
a
10
ms
window
around
the
FRN
peak.
The
FRN
peak
was
defined
as
the
most
negative
peak
between
180
and
360
ms
after
feedback
onset,
and
the
preceding
positive
peak
was
defined
as
the
most
positive
peak
between
100
and
220
ms
after
feedback
onset.
Peaks
were
identified
separately
for
each
partic-
ipant
and
error
type
with
the
aid
of
voltage
scalp
topographies.
This
analysis
was
conducted
on
electrodes
FCz
and
Cz,
where
FRN
amplitudes
were
maximal
in
both
the
current
study
(see
Fig.
2)
and
previous
FRN
studies
(e.g.,
Miltner
et
al.,
1997).
FRN
amplitudes
were
compared
across
error
types
by
a
one-way
repeated-measures
ANOVA.
The
FRN
was
followed
by
a
frontocentral
P3a
component
and
a
subsequent
parietocentral
P3b
component.
The
P3a
was
defined
as
the
mean
amplitude
of
the
difference
wave
from
360
to
390
ms
after
feedback
onset
at
electrodes
Fz,
FC1,
and
FC2.
The
P3b
was
defined
as
the
mean
amplitude
from
430
to
560
ms
after
feedback
onset
at
electrodes
Pz,
CP1,
and
CP2.
Electrode
sites
were
chosen
based
on
previous
studies
on
the
P3a
and
P3b
and
the
locations
of
their
maximal
amplitudes
in
the
current
study
(see
Fig.
3).
To
confirm
that
the
two
subcomponents
had
distinct
time
windows
and
scalp
distributions,
we
compared
the
mean
amplitudes
with
a
2
(interval:
early
[360–390
ms],
late
[430–560
ms])
×
2
(location:
frontocentral
[Fz,
FC1,
FC2],
parietocentral
[Pz,
CP1,
CP2])
×
3
(error
type:
own,
joint,
observed)
repeated
measures
ANOVA.
Finally,
we
checked
for
attentional
effects
by
examining
the
pos-
terior
P1
and
N1
components
evoked
by
the
feedback
stimulus,
as
the
amplitudes
of
these
components
depend
on
the
level
of
visual
attention
that
is
allocated
to
a
stimulus
(Luck
&
Kappenman,
2012).
We
measured
the
mean
amplitude
of
the
P1
component
from
110
to
140
ms
after
feedback
onset
at
electrodes
P7
and
P8,
and
the
mean
amplitude
of
the
N1
component
from
150
to
180
ms
after
feedback
onset
at
electrodes
P7
and
P8.
We
compared
the
mean
amplitudes
using
a
3
(condition:
own,
joint,
observed)
×
2
(out-
come:
correct,
error)
ANOVAs
for
each
component.
All
ANOVAs
were
Greenhouse–Geisser
corrected
where
appropriate,
and
post-
hoc
comparisons
were
conducted
with
paired-samples
t-tests.
2.
Results
Fig.
2
shows
that
own
errors
elicited
larger
FRN
amplitudes
than
joint
errors,
which
elicited
larger
FRN
amplitudes
than
observed
errors.
The
ANOVA
on
FRN
amplitudes
confirmed
a
significant

4
J.D.
Loehr
et
al.
/
Biological
Psychology
111
(2015)
1–7
Fig.
2.
(a)
Grand-average
difference
waves
for
each
error
type
at
electrode
FCz,
time-locked
to
feedback
onset,
and
the
scalp
voltage
distribution
of
the
grand-average
difference
wave
within
the
time
window
of
analysis
of
the
FRN.
(b)
Mean
peak-to-peak
amplitudes
of
the
FRN
for
each
error
type.
(For
interpretation
of
the
references
to
colour
in
this
figure
legend,
the
reader
is
referred
to
the
web
version
of
this
article.)
Fig.
3.
(a)
Grand-average
difference
waves
for
each
error
type
at
electrode
Fz,
time-locked
to
feedback
onset,
and
the
current
source
density
of
the
grand-average
difference
wave
within
the
time
window
of
analysis
of
the
P3a.
(b)
Grand-average
difference
waves
for
each
error
type
at
electrode
Pz,
time-locked
to
feedback
onset,
and
the
current
source
density
of
the
grand-average
difference
wave
within
the
time
window
of
analysis
of
the
P3b.
(c)
Mean
amplitudes
for
each
error
type
by
location
and
time
interval.

J.D.
Loehr
et
al.
/
Biological
Psychology
111
(2015)
1–7
5
effect
of
error
type,
F(2,
40)
=
22.22,
p
<
0.001.
Post-hoc
tests
con-
firmed
that
FRN
amplitudes
were
significantly
larger
for
own
errors
compared
to
joint
errors,
t(20)
=
2.77,
p
=
0.012,
and
compared
to
observed
errors,
t(20)
=
5.34,
p
<
0.001.
FRN
amplitudes
were
also
significantly
larger
for
joint
errors
compared
to
observed
errors,
t(20)
=
5.13,
p
<
0.001.
There
were
no
differences
in
the
latencies
of
the
FRN
peak
or
the
preceding
positive
peak
by
error
type
(see
Supplementary
material
for
details).
Fig.
3
shows
the
mean
P3
amplitudes
elicited
by
each
error
type
at
each
combination
of
location
and
time
interval.
The
ANOVA
on
mean
amplitudes
showed
a
significant
location
by
interval
interac-
tion,
F(1,
20)
=
40.69,
p
<
0.001,
such
that
mean
amplitudes
were
larger
at
frontocentral
compared
to
parietocentral
sites
for
the
early
interval,
t(20)
=
3.22,
p
=
0.004,
whereas
mean
amplitudes
were
larger
at
parietocentral
compared
to
frontocentral
sites
for
the
late
interval,
t(20)
=
3.97,
p
=
0.001.
This
confirms
that
there
were
indeed
two
distinct
subcomponents:
an
early
component
that
peaked
frontocentrally
(the
P3a)
and
a
later
component
that
peaked
parietocentrally
(the
P3b).
The
ANOVA
also
showed
a
main
effect
of
error
type,
F(2,
40)
=
4.73,
p
=
0.014,
which
was
qualified
by
an
error
type
by
location
interaction,
F(2,
40)
=
4.76,
p
<
0.014,
and
an
error
type
by
location
by
interval
interaction,
F(2,
40)
=
6.52,
p
=
0.004.
All
other
main
effects
and
interactions
were
non-significant,
Fs
<
1.03,
ps
>
0.35.
To
further
examine
the
3-way
interaction,
we
conducted
2
(location)
×
3
(error
type)
ANOVAs
separately
for
the
P3a
(early
interval)
and
the
P3b
(late
interval).
The
ANOVA
on
P3a
amplitudes
confirmed
a
main
effect
of
loca-
tion,
F(1,
20)
=
10.34,
p
=
0.004,
such
that
mean
amplitudes
were
larger
at
frontocentral
than
parietocentral
sites.
The
ANOVA
also
revealed
a
main
effect
of
error
type,
F(2,
40)
=
5.95,
p
=
0.005,
and
an
error
type
by
location
interaction,
F(2,
40)
=
9.12,
p
=
0.001.
The
left
panel
of
Fig.
3c
shows
that
at
frontocentral
sites,
where
the
P3a
was
maximal,
own
errors
elicited
larger
amplitudes
compared
to
joint
errors,
t(20)
=
2.67,
p
=
0.015,
and
compared
to
observed
errors,
t(20)
=
3.99,
p
=
0.001,
whereas
amplitudes
elicited
by
joint
and
observed
errors
did
not
differ,
t(20)
=
1.17,
p
=
0.26.
At
parieto-
central
sites,
differences
between
error
types
were
smaller,
such
that
own
errors
elicited
larger
amplitudes
than
observed
errors,
t(20)
=
2.16,
p
=
0.043,
but
the
difference
between
own
and
joint
errors
was
not
significant,
t(20)
=
1.85,
p
=
0.079,
nor
was
the
dif-
ference
between
joint
and
observed
errors,
t(20)
=
0.44,
p
=
0.66.
The
ANOVA
on
P3b
amplitudes
revealed
only
a
main
effect
of
location,
F(1,
20)
=
15.73,
p
=
0.001,
confirming
that
mean
ampli-
tudes
were
larger
at
parietocentral
compared
to
frontocentral
sites.
There
was
no
main
effect
of
error
type,
F(2,
40)
=
2.03,
p
=
0.16,
or
error
type
by
location
interaction,
F(2,
40)
=
0.65,
p
=
0.53.
Finally,
the
ANOVAs
on
mean
P1
and
N1
amplitudes
showed
no
main
effects
or
interactions,
all
ps
>
0.05.
Thus,
participants
ini-
tially
attended
equally
the
feedback
stimulus
irrespective
of
the
person(s)
who
had
produced
the
sequence
and
the
correctness
of
their
effort.
3.
Discussion
The
current
study
compared
ERP
responses
to
feedback
indi-
cating
that
joint
errors
(resulting
from
two
people’s
coordinated
actions)
or
individual
errors
(resulting
from
a
single
person’s
own
or
another
person’s
observed
actions)
had
been
committed.
Differ-
ences
between
joint
and
individual
errors
were
largest
for
the
FRN,
the
amplitude
of
which
was
largest
in
response
to
a
person’s
own
errors,
reduced
for
jointly
committed
errors,
and
further
reduced
for
another
person’s
observed
errors.
The
P3a
also
differentiated
between
joint
and
individual
errors,
such
that
P3a
amplitudes
were
larger
for
own
errors
compared
to
both
jointly-committed
and
observed
errors.
In
contrast,
P3b
amplitudes
did
not
differentiate
between
joint
and
individual
errors.
These
findings
indicate
that
producing
an
error
together
with
a
partner
influences
neural
activ-
ity
related
to
outcome
evaluation
(the
FRN
and
P3a)
but
has
less
impact
on
activity
related
to
adapting
future
behaviour
(the
P3b).
The
reduced
FRN
elicited
by
joint
compared
to
own
errors
corroborates
previous
research
showing
that
FRN
amplitudes
are
reduced
for
action
outcomes
that
are
partly
contingent
on
another
person’s
passively
observed
action
(Li
et
al.,
2010)
or
on
a
com-
puter’s
random
selection
(Martin
&
Potts,
2011;
Yeung
et
al.,
2005)
compared
to
action
outcomes
that
are
fully
contingent
on
a
per-
son’s
own
action.
The
current
findings
are
the
first
to
show
that
the
FRN
is
reduced
for
action
outcomes
that
are
partly
contingent
on
another
person’s
actions
even
when
both
people
must
actively
and
continuously
adapt
to
each
other’s
actions
to
achieve
a
desired
outcome.
The
finding
that
FRN
amplitudes
are
reduced
when
peo-
ple
share
control
over
an
action
outcome
is
also
consistent
with
previous
research
showing
reduced
FRN
amplitudes
when
peo-
ple
believe
that
task
outcomes
are
not
controllable
compared
to
controllable
(Li
et
al.,
2011)3.
The
current
findings
also
add
to
pre-
vious
reports
that
the
FRN
is
reduced
for
observed
compared
to
own
action
outcomes
(e.g.,
Bellebaum
et
al.,
2010;
Fukushima
&
Hiraki,
2009;
Koban,
Pourtois,
Bediou,
&
Vuilleumier,
2012;
Leng
&
Zhou,
2010;
Yu
&
Zhou,
2006).
In
most
of
these
studies,
participants
performed
tasks
independently
and
each
was
rewarded
based
on
their
own
task
performance;
only
in
Koban
et
al.
(2012)
did
partici-
pants’
independent
outcomes
result
in
a
combined
reward
for
both
participants.
The
current
findings
extend
the
pattern
of
self-other
differences
in
outcome
processing
to
a
context
in
which
partici-
pants’
outcomes
not
only
resulted
in
combined
rewards
but
were
also
the
product
of
active
interpersonal
coordination.
One
possible
explanation
for
the
modulation
of
FRN
amplitudes
in
the
current
study
is
that
shared
control
reduced
the
affective
value
of
joint
errors
relative
to
own
errors,
and
likewise
the
absence
of
control
reduced
the
affective
value
of
observed
errors
relative
to
own
and
joint
errors.
Some
researchers
have
argued
that
action
out-
comes
that
are
less
contingent
on
a
person’s
actions
have
reduced
affective
value
compared
to
outcomes
that
are
fully
contingent
on
the
person’s
actions
(Yeung
et
al.,
2005).
Moreover,
the
affective
value
of
action
outcomes
modulates
the
FRN
in
both
social
and
non-social
contexts
(see
Koban
&
Pourtois,
2014,
for
a
review).
A
second
possibility
is
that
shared
control
altered
the
subjective
probability
of
correct
vs.
incorrect
outcomes.
Outcomes
that
are
less
objectively
or
subjectively
probable
elicit
larger
FRNs
(Walsh
&
Anderson,
2012).
In
the
current
study,
the
objective
probability
of
correct
vs.
incorrect
outcomes
was
equated
between
individ-
ual
and
joint
settings
through
the
use
of
an
adaptive
window
to
determine
accuracy.
However,
people
often
experience
a
greater
subjective
probability
of
positive
outcomes
than
is
warranted
by
the
objective
probability
(e.g.,
Hajcak,
Moser,
Holroyd,
&
Simons,
2007;
Miller
&
Ross,
1975).
People
are
also
more
optimistic
about
their
own
performance
compared
to
other
people’s
(Krueger,
1998).
Thus,
participants
in
the
current
study
may
have
expected
more
correct
outcomes
for
self-produced
sequences
compared
to
joint
and
observed
sequences,
rendering
errors
more
unexpected
and
therefore
eliciting
larger
FRNs
in
the
former
conditions
(see
Li
et
al.,
2010,
for
a
similar
argument).
Further
work
is
needed
to
deter-
mine
whether
and
to
what
extent
the
affective
value
and
subjective
3It
is
possible
that
sharing
control
over
the
action
outcome
may
have
induced
participants
to
feel
shared
responsibility
for
the
outcome.
Indeed,
control
and
responsibility
often
go
hand
in
hand;
for
example,
Li
et
al.’s
(2011)
manipulation
of
people’s
beliefs
about
control
affected
their
feelings
of
responsibility,
which
cor-
related
with
FRN
amplitudes.
We
use
the
term
shared
control
to
capture
the
fact
that
participants
produced
the
action
sequences
together,
while
remaining
agnostic
with
respect
to
their
feelings
of
responsibility.

6
J.D.
Loehr
et
al.
/
Biological
Psychology
111
(2015)
1–7
probability
of
action
outcomes
are
modulated
by
acting
in
coordi-
nation
with
a
partner.
The
P3a
has
received
relatively
less
attention
than
the
FRN
in
the
action-monitoring
literature.
Most
researchers
treat
the
P3
that
follows
the
FRN
as
a
unitary
component,
although
recent
work
indicates
that
the
P3a
and
P3b
subcomponents
are
differen-
tially
sensitive
to
the
valence
and
magnitude
of
reward
feedback
(West,
Bailey,
Anderson,
&
Kieffaber,
2014).
In
the
current
study,
P3a
amplitudes
were
larger
for
own
errors
compared
to
joint
or
observed
errors.
Given
previous
research
indicating
that
the
P3a
reflects
an
attentional
response
related
to
the
initial
evaluation
of
a
stimulus
as
task-relevant
(Polich,
2007),
this
finding
suggests
that
people
may
allocate
more
attentional
resources
to,
and
evaluate
as
more
task-relevant,
outcomes
over
which
they
have
full
control
compared
to
outcomes
over
which
they
share
or
have
no
control.
The
larger
P3a
amplitudes
elicited
by
own
compared
to
joint
errors
in
the
current
study
is
consistent
with
West
et
al.’s
(2014)
find-
ing
that
both
FRN
and
P3a
amplitudes
were
larger
when
Blackjack
players
experienced
negative
outcomes
that
could
be
attributed
primarily
to
their
own
actions
compared
to
outcomes
that
could
be
attributed
in
part
to
the
dealer’s
actions.
The
lack
of
differ-
ence
between
P3a
amplitudes
elicited
by
joint
and
observed
errors
suggests
that
the
P3a
may
be
particularly
sensitive
to
differences
between
a
person’s
own
action
outcomes
and
outcomes
that
can
be
partly
or
fully
attributed
to
other
people.
Previous
research
examining
how
control
over
action
outcomes
influences
the
P3
has
focused
primarily
on
later
activity
at
more
posterior
electrodes
(i.e.,
Cz,
CPz,
and
Pz;
Li
et
al.,
2010;
Yeung
et
al.,
2005;
Martin
&
Potts,
2011).
This
work
has
shown
that
P3
ampli-
tudes
are
larger
for
outcomes
that
are
fully
contingent
on
a
person’s
own
actions
compared
to
outcomes
that
are
partly
or
not
at
all
contingent
on
the
person’s
actions
(e.g.,
Li
et
al.,
2010;
Yeung
et
al.,
2005).
The
current
study
revealed
no
significant
differences
in
later,
more
posterior
(i.e.,
P3b)
amplitudes
across
conditions,
despite
the
fact
that
own
errors
were
fully
contingent
on
the
participant’s
actions
whereas
joint
and
observed
errors
were
not.
Although
this
may
seem
contradictory
to
previous
findings,
it
might
instead
high-
light
differences
in
the
cognitive
and
neural
processes
elicited
by
performance
monitoring
compared
to
gambling
tasks.
The
current
paradigm
allowed
participants
to
adapt
their
behaviour
after
errors
in
order
to
influence
subsequent
outcomes,
whereas
the
gambling
tasks
used
in
previous
studies
did
not.
In
the
current
paradigm,
both
own
and
joint
errors
signalled
a
need
to
adapt
future
behaviour,
regardless
of
the
degree
of
control
participants
had
over
the
error.
In
addition,
observed
sequences
were
randomly
intermixed
with
self-produced
sequences,
which
may
have
encouraged
participants
to
learn
from
their
partners’
mistakes
as
well
as
their
own.
The
lack
of
differences
between
conditions
is
in
line
with
the
notion
that
P3b
amplitudes
reflect
participants’
motivation
to
adapt
future
behaviour
(Nieuwenhuis
et
al.,
2004),
although
it
may
be
worth
not-
ing
that
there
were
numerical
differences
in
P3b
amplitudes
that
scaled
with
control
over
action
outcomes.
It
is
therefore
possible
that
differences
between
joint
and
individual
errors
could
be
evi-
dent
in
P3b
amplitudes
if
an
alternative
experimental
design
was
implemented.
4.
Conclusion
The
neural
processes
involved
in
monitoring
multiple
people’s
coordinated
actions
have
just
begun
to
be
investigated.
Although
previous
work
has
examined
how
people
monitor
each
person’s
individual
outcomes
within
a
joint
action
(e.g.,
the
individual
tones
that
comprise
a
musical
duet;
Loehr,
Kourtis,
Vesper,
Sebanz,
&
Knoblich,
2013;
see
also
Picton,
Saunders,
&
Jentzsch,
2012),
the
current
study
is
the
first
to
investigate
the
effects
of
interpersonal
coordination
on
monitoring
and
evaluating
action
outcomes.
We
show
that
producing
errors
together
with
another
person
reduces
neural
activity
related
to
outcome
evaluation
but
has
less
impact
on
neural
activity
associated
with
the
motivation
to
adapt
future
behaviour.
These
findings
further
our
understanding
of
how
peo-
ple
monitor
and
evaluate
joint
action
outcomes
to
ensure
that
shared
goals
are
achieved,
and
may
have
important
implications
for
investigations
of
pathological
conditions
that
entail
social
neu-
rocognitive
deficits
such
as
psychopathy
and
autism.
Funding
This
work
was
supported
by
the
European
Union’s
7th
Frame-
work
Programme
[Marie
Curie
International
Incoming
Fellowship,
Project
254419,
to
J.L.].
Appendix
A.
Supplementary
data
Supplementary
data
associated
with
this
article
can
be
found,
in
the
online
version,
at
http://dx.doi.org/10.1016/j.biopsycho.2015.
08.004.
References
Bellebaum,
C.,
Kobza,
S.,
Thiele,
S.,
&
Daum,
I.
(2010).
It
was
not
MY
fault:
Event-
related
brain
potentials
in
active
and
observational
learning
from
feedback.
Cerebral
Cortex,
20,
2874–2883.
Brazil,
I.,
Mars,
R.
B.,
Bulten,
B.
H.,
Buitelaar,
J.
K.,
Verkes,
R.
J.,
&
de
Bruijn,
E.
R.
A.
(2011).
Neurophysiological
correlates
of
the
detection
of
own
and
others’
errors
in
individuals
with
psychopathy.
Biological
Psychiatry,
69,
693–699.
de
Bruijn,
E.
R.
A.
(2012).
It
takes
two:
A
social
neurocognitive
perspective
on
psy-
chiatry.
Acta
Psychiatrica
Belgica,
112,
56–59.
Fukushima,
H.,
&
Hiraki,
K.
(2009).
Whose
loss
is
it?
Human
electrophysiological
correlates
of
non-self
reward
processing.
Social
Neuroscience,
4,
261–275.
Gehring,
W.
J.,
&
Willoughby,
A.
R.
(2002).
The
medial
frontal
cortex
and
the
rapid
processing
of
monetary
gains
and
losses.
Science,
295,
2279–2282.
Hajcak,
G.,
Moser,
J.
S.,
Holroyd,
C.
B.,
&
Simons,
R.
F.
(2007).
It’s
worse
than
you
thought:
The
feedback
negativity
and
violations
of
reward
prediction
in
gam-
bling
tasks.
Psychophysiology,
44,
905–912.
Holroyd,
C.
B.,
Nieuwenhuis,
S.,
Yeung,
N.,
&
Cohen,
J.
D.
(2003).
Errors
in
reward
prediction
are
reflected
in
the
event-related
brain
potential.
NeuroReport,
14,
2481–2484.
Jung,
T.
P.,
Makeig,
S.,
Westerfield,
M.,
Townsend,
J.,
Courchesne,
E.,
&
Sejnowski,
T.
J.
(2000).
Removal
of
eye
activity
artifacts
from
visual
event-related
potentials
in
normal
and
clinical
subjects.
Clinical
Neurophysiology,
111,
1745–1758.
Koban,
L.,
&
Pourtois,
G.
(2014).
Brain
systems
underlying
the
affective
and
social
monitoring
of
actions:
An
integrative
review.
Neuroscience
and
Biobehavioral
Reviews,
46,
71–84.
http://dx.doi.org/10.1016/j.neubiorev.2014.02.014
Koban,
L.,
Pourtois,
G.,
Bediou,
B.,
&
Vuilleumier,
P.
(2012).
Effects
of
social
context
and
predictive
relevance
on
action
outcome
monitoring.
Cognitive,
Affective
&
Behavioral
Neuroscience,
12,
460–478.
Krueger,
J.
(1998).
Enhancement
bias
in
the
description
of
self
and
others.
Personality
and
Social
Psychology
Bulletin,
24,
505–515.
Leng,
Y.,
&
Zhou,
X.
(2010).
Modulation
of
the
brain
activity
in
outcome
evaluation
by
interpersonal
relationship:
An
ERP
study.
Neuropsychologia,
48,
448–455.
Li,
P.,
Han,
C.,
Lei,
Y.,
Holroyd,
C.
B.,
&
Li,
H.
(2011).
Responsibility
modulates
neu-
ral
mechanisms
of
outcome
processing:
An
ERP
study.
Psychophysiology,
48,
1129–1133.
Li,
P.,
Jia,
S.,
Feng,
T.,
Liu,
Q.,
Suo,
T.,
&
Li,
H.
(2010).
The
influence
of
the
diffusion
of
responsibility
effect
on
outcome
evaluations:
Electrophysiological
evidence
from
an
ERP
study.
NeuroImage,
52,
1727–1733.
Loehr,
J.
D.,
Kourtis,
D.,
Vesper,
C.,
Sebanz,
N.,
&
Knoblich,
G.
(2013).
Monitoring
indi-
vidual
and
joint
action
outcomes
in
duet
music
performance.
Journal
of
Cognitive
Neuroscience,
25,
1049–1061.
Luck,
S.
J.,
&
Kappenman,
E.
S.
(2012).
ERP
components
and
selective
attention.
In
S.
J.
Luck,
&
E.
S.
Kappenman
(Eds.),
The
Oxford
handbook
of
event-related
potential
components
(pp.
295–327).
New
York,
NY:
Oxford
University
Press.
Martin,
L.
E.,
&
Potts,
G.
F.
(2011).
Medial
frontal
event-related
potentials
and
reward
prediction:
Do
responses
matter?
Brain
and
Cognition,
77,
128–134.
Miller,
D.
T.,
&
Ross,
M.
(1975).
Self-serving
biases
in
the
attribution
of
causality:
Fact
or
fiction?
Psychological
Bulletin,
82,
213–225.
Miltner,
W.
H.
R.,
Braun,
C.
H.,
&
Coles,
M.
G.
H.
(1997).
Event-related
brain
potentials
following
incorrect
feedback
in
a
time-estimation
task:
Evidence
for
a
“generic”
neural
system
for
error
detection.
Journal
of
Cognitive
Neuroscience,
9,
788–798.
Nieuwenhuis,
S.,
Aston-Jones,
G.,
&
Cohen,
J.
D.
(2005).
Decision-making,
the
P3,
and
the
locus
coeruleus–norepinephrine
system.
Psychological
Bulletin,
131,
510–532.