The Japanese Psychonomic Society
NII-Electronic Library Service
TheJapanesePsychonomic Society
TheJapaneseJburnatofPsychonomicScience
2D13,
VeL32,No. 1,143-144Summary
of
Awarded
Presentation
DisruP
Megumi
SuzuKi
tion
ofvisual
search
by
the
misbinding
illusion
a*,
Jeremy
WoLFEb,
Tbdd
S.
HoRowiTzb,
and
Yasuki
NoGucHi
"Kbbe
university
bHarvardMedical
School
In
cluttered environments, attentive searchingis
necessaryto
find
targets
defined
by
a coajunction of multiplefeatures
(e.g.,
color andshape),
Thus
it
has
classicallybeen
assumedthat
the
allocation of attention causesbinding,
In
contrast,here
we show aphenomenon
wherethe
causalitybetween
attention andbinding
is
reyersed;the
visualprecessing
offeature
binding
affects spatial shifts of attention.Subjects
searchedfor
targets
with acombination
ofparticular
colors and orientations(e,g,,
red, vertiealbars)that
were embeddedin
peripheral
areas
of
a
search array.When
acentral area consisted ofbars
with regular color-orientationpairings,
these
regularpairings
automaticallychanged
perceptual
bindings
in
the
peripheral
area,re$ultingin
target
omissien.These
resultsindicate
that
the
visu-alsystemperforms
a rapidextrapolation
ofbinding
information
spreadingfrorn
centerto
peripheral
areas, whichthrows
offattention.Our
resultsthus
provided
evidence supportingthe
hypothesis
that
binding
affbcts attention.Key
words:preattentive
process,
feature
binding,
misbindingillusion,
visual
searchMounting
evidenceindicates
that
visualfeatures
such asshape and celQr are
processed
in
separate regionsin
the
brain.
Hew
ourbrain
binds
the
features
associated withthe
same
eb-ject
is
animportant
problem.
A
dominant
theory
(Treisman
&
Gelade,
1980)
argues thatthe
integration
offeatures
requires attention,IYpical
examplesare
findings
in
visual searchpara-digms,
When
targets
canbe
distinguished
frorn
distracters
by
a
simple
feature,
these
targets
are readily apparent,When
the
targets
aredefined
by
aceajunction of rnultiplefeatures,
how-ever, an attentive searchis
necessaryto
find
them.
This
indi-catesthat
attentiondeployment
resultsin
the
integration
of
different
features
of an object,In
contrast,there
arealso
psychophysical
data
suggesting
an early mechanismfor
feature
binding
that
does
not requireat-tention
(e.g.,
the
McCollough
effect),It
remains unclearhow
this
earlybinding
mechanismis
relatedte
the
visual searchfindings
above.In
this
study we addressedthis
issue
usingthe
misbinding
illusion
(Wli,
Kanai,
&
Shimojo,
2004),
aphenom-enon representing
the
earlybinding
mechanism,In
this
illu-sion, central andperipheral
portions
of atransparent
motionfield
combine color and rnotionin
opposite ways.In
observersof such stimuli,
perceptual
color-motionbindings
in
the
periphery
change to matchthe
bindings
in
the
center,resu!t-ing
in
erroneous celer-metionpairings
in
the
periphery
The
*
Corresponding
authon
Department
ofPsychologF
Grad-uate
School
ofHumanities,
Kobe
University)
1-1
Rokke-dai-cho,
Nada,
Kobe
657-8501,
Iapan.
E-rnail:g-bt-bs-88@vahoo,co.jp
Cepy
current
studyfirst
shewsthat
this
illusion
also
takes
place
in
color-orientatien
bindings
(Exp.
1).
The
illusion
is
then
appliedto
avisual searchparadigm
(Exp.
2).
Experiment
1
Methods.
Eleven
and nine subjectsparticipated
in
Exper-iments
1a
andlb,
respectivelyIn
eachtrial,
wepresented
sub-jects
with an array of864
stimuli(27
deg
×18
deg;
cell size ofthe array:
O.75
deg
×O,75
deg)
in
whichfour
types
ofbars
(red-horizontal,
red-vertical,
green-horizontal,
andgreen-ver-tical)
wereintermixed.
Each
array consisted ofthree
fields:
a
central
field
(13.5
deg
×18
deg,
432
items)
andtwo
peripheral
fields
(Ieft
and right surreunds,6.75
deg
×18
deg,
216
items
for
each),In
the
left
and right surrounds,the
array wascom-posed
of equal numbers of red(R)
andgreen
(G)
bars
(108
for
each).
Half
ofthe
green
bars
werehorizontal
(H)
and otherhalfwere
vertical(V).
Fer
the
peripheral
redbars,
the
ratio ofH
to
V
changed acrosstrials
from
O
:100
to
100
:O.
The
cen-figuration
ofthe
centralfield
was chosenfrem
four
condi-tions,
asfo11ows,
In
the
"No
Stimulus"
(or
NS)
condition,there
was nobar
in
the
centralfield
(peripheral
bars
only),In
the
"Random"
condition, equal numbers ofthe
four
types ofbars
(R-H,
R-Y
G-H,
andG-V)
wereintermixed
in
the
centralfield.
These
two
conditions served as controls.In
the
"Dou-ble"
conditions, allredbars
were of one orientation and allgreen
bars
were ofthe
other orientation.Thus
the
centralfield
in
the
Double-1
condition consisted ofR-H
andG-V
bars,
while
that
in
theDeuble-2
condition consisted ofR-V
andThe Japanese Psychonomic Society
NII-Electronic Library Service
TheJapanesePsychonomic Society
144
The
Japanese
Journal
ofPsychonomic
Science
Vbl.
32,
No.
1
G-Hbars,
Subjects
estimatedthe
ratio ofR-H
te
R-V
bars
in
the
pe-ripheral
fields.
They
pressed
onekey
to
indicate
that
the
num-ber
ofperipheral
R-H
bars
wasgreater
than
the
number ofR-V
bais,
andpressed
anotherto
indicate
the
reverse.They
were askedto
ignore
stimuliin
the
centralfield
and
to
attend
te
the
surround(left
or right, assigned randomlyfor
eachsub-ject).
The
actualpercentage
ofperipheral
R-H
bars
(%R-H)
was varied randomly acrosstrials
from
O
to
100%,
and we measuredthe
probability
of
subjects reportingR-H
bars
asgreater
in
number(%
`H") as afunction
of%R-H.
Misbind-ing
illusiens
shouldbe
observed as changesin
the
point
ef subjective equa!ity(PSE)
between
R-H
andR-V
Thus,
Dou-ble-1
sheuldbias
the
apparent orientation ofperipheral
redbar$
toward
horizontal,
whileDouble-2
shouldbias
the
appar-entorientationofperipheralredbarstowardvertical.
Every
trial
began
withfixation
for
1
s,fo11ewed
by
a200
mspresentation
of
the
arrayAfter
400
ms, subjects wereprompt-ed
for
akey
response.Experiment
1a
involved
the
two
controlconditions
(NS
andRandom),
Subjects
performed
four
ses-sions of90
trials
ea[h.In
Experiment
lb,
arlfour
cenditions were randomlyintermixed.
Each
participant
completed six sessions of90
trials
each,Results.
In
Experiment
la,
no significantdifference
in
PSE
was ebservedbetween
NS
andRandorn
(p=.64),
whichindicated
that
the
bars
in
the
centralfieLd
themselves
did
notinduce
anyperceptual
biases.
In
Experirnent
lb,
PSEs
in
the
two control conditions
(NS
andRandom)
againdid
notdiffer
{p=.14).
In
contrast, we obseryed a significantdecrease
ofthePSE
in
Double-1
cemparedto
Random
(p=.O1),
andasignifi-cant
increase
in
Double-2
comparedto
Random
(p=.03).
These
results
are similarto
those
observedin
color-motion misbinding(Wu,
Kanai,
&
Shimojo,
2004),
indicating
that
the
misbinding
illusion
can alsotake
place
between
color anderi-entationinformation.
Experiment2
Methods.
In
Experiment
2,
we appliedthe
color-erienta-tion
misbindingillusion
to
a visual searchparadigrn.
Eleven
subjects
participated
in
Experiment
2,
As
in
Experiment
1,
the
array consisted of
four
types
of
bars
and was subdividedinto
three
fields,
Each
field
had
an equal number of red andgreen
bars,
andthe
H
:V
ratio oftheperipheral
green
bars
wasfixed
at
50
/50.
The
targets
ofvisual
search wereeither
R-V
or
R-H
bars
(determined
randemlyfbr
each subject).These
target
bars
wereplaced
in
eitherthe
left
or right surround.In
the
target
field,
a number oftarget
bars
couldbe
2,
12,
22,
38,
54,
or70,
which correspondedto
respectivetarget
per-centages(%
Thrgets)
of2,
11,
20,
35,
50,
or6S%ofthe
108
to-tal
redbars
in
atarget
field.
There
wereno
targets
in
the
cen-tralateral
field,
Subjects
reported whetherthe
left
or right surround containedthe
target
bars,
ignertng
the
stimuliin
the
central
field,
We
comparedtask
accuracyfor
the
two
types
of centralfields
(Double
andNS).
In
both
conditions,target
con-junctions
were absentfrom
the
centralfield.
Each
participant
performed
six session$of
96
trials
each.
Fer
eachtype
of centralfield,
changesin
task
accuracy as afunction
ofthe
%Tlargets
werefitted
by
a sigmoidfunction.
We
comparedthe
75%threshold
ofthe
function
between
Double
(misbinding)
andNS
(no
misbinding) conditions.Results.
We
observedasignificant
increase
in
the
75%
threshold
in
Double
comparedto
NS
(p=.O07).
These
results were censistent withthe
viewthat
the
centraldispla7
in
the
Double
trials
induced
illusory
bindings
in
the
peripheral
tar-gets,
whichdrew
attention awayfrom
subjects and resultedin
the
failure
to
detect
targets.
Discussion
In
contrastto
aprevieus
view
that
an allocation ofattention causes anintegration
offeatures
at an attendedposition,
wetested
the
hypothesis
that
the
visualprecessing
offeature
binding
affectsthe
allocation of attentienduring
visual search.SpecificallM
we usedthe
misbindingillusion
as an example of earlyfeature
binding
(ipdependent
ofattention) and examinedhow
the
illusion
interacted
with attentionin
visual search.Results
revealedthat
the
misbindinginduced
by
the
centralstimuli
changed
perceptual
color-orientationbinding
in
the
peripheral
stimuli(Exp.
1),
reducingthe
saliency
oftargets
in
the
peripheral
field
(Exp.
2),
These
resultsprovide
evidencesupporting