The Japanese Psychonomic Society
NII-Electronic Library Service
The JapanesePsychonomic Society
TVteYitPaneselburnal uf Rs.ychonomicScience
2003,Vol.22,No,1,27-28
Prizewinner's
Summary1-B-19
Backdrop
motion
illusion
from
Kiyoshi
images
FuJiMoTo*・
** andUniversity
of
Toleyoof
a
walking
human
figure
Takao
SATo***
*・***
We have found a novel vi$ual
illusjon
where an ambiguous motion became unidirectional withthe
superimposition of an image of a hurnanfigure
walking on atrcadmilL
When wepresented
a counterpha$e grating as an ambiguous backdrop, thegrating
appeared to driftin theoppositedirectiontothe
bipedal
locomotion. This phenomenon indicatesthat the visual system wouldevaluate
low-level
motion signals relative toahigh-lcvel
representation of an object's movernentdefined
by
its
biological
motion. In thisstudy we presented images of human figurescither walkingforward
or walking backward inorder toexamine the effects of adirectional
correspon-dence between human rnovement and
form
that was importantfor
thc recognition ofbiological
motion. The results revealed that a
forward
walker produccd the illusionmore reliably than abackward
walker, and confirmed thatrecognition oflocomQtion
isa prirnarydeterminant fortheillusion,
Key words: visual Musion, motion
perception,
biologicalmotion
Current
models of motion perccption attempt to explain how the visual system estimates the move-ment of an object from retinal fiow(Lu
&
Sperling,
2002). However, thevisual system isa]so capable of
representing the movernent of an object even ifan object retains a
fixed
position on thc retinas whentheeyes are stationary. For example, the translation
of a wa]ker isrecognized solely from the articulated
rnovernents of the
limbs.
Our investigationsdemonstrate that such a
high-level
representation of human movementsis
so pow-erful thatit
can modulate low-levelmotion proccss-ing so that an apparent motion in a backgroundpattern isinduced, Inorder toexplere this
phenomc-non we
have
designed
movie clips inwhich the im-age of a walklnghuman
figure
is
superimposed on a counterphase grating(Figure
1).
The
grating isthesum of two gratings moving at an equal vclocity in
opposite
directions.
This grating isusually per-*Intelligent
Modeling
Laboratory, University of
Tokyo,
1--1-1
Yayoi,
Bunkyo-ku, Tokyo
8657
** Present address:
Life
ElectronIcs
Laboratory,tiona] Instituteof Advanced
Industrial
Science
and
Technology,
1-8-3Midori-ga-oka,
Ikeda,
Osaka
563-8577***
Department
of Psychology, Faculty of Letters,
University
ofTokyo,
7-3-1
Hongo, Bunkye-ku,
Tokyo
113-O033
ceived as a
directionless
flicker,which indicatesa non-biasedintegration
of the twodircctional
signals extracted jn the earliest motionprocessing.
How-ever, when an
image
of a walking humanfigure
is
superimposed on thecounterphase grating,the
grat-ing
appears tobe unidirectional, as though a striped wall was moving relative to the stationary walker(for
demonstrations of this illusion,visit http:f!www1.u-netsurLne,jp/
fff/gait/),
Inthepresent study wc have examined theeffects
of a directional correspondence
between
human
movement and
form
by
using an image of a humanwalking ineither oftwo ways: forward or backward.
The backward walker faced
in
theopposite directionto
locomotion,
Iftheform
were crittcal, thc percep-tionof thedirection
ef thegrating movcment would be oppo$ite'to
thedirection
thefigurc
was facing,nttrnely, in the same direction as the
backward
Figure1.
Movie
clips of the backdrop rnotion
illusion.
If
the person iswalking toward the Ieft,thegrating appears to
drift
toward therjght,The Japanese Psychonomic Society
NII-Electronic Library Service
The JapanesePsychonomic Society
28
TheJapanese
Journal
of PsychonomicScience
Vol.22, No. 1 walking.
Method
Apparatus. Visual stimuli were
displayed
on acolor
CRT
monitor by using anApple
PowerMacin-tosh, A gamma correction was applied to
produce
luminance
]inearity.
Each
observer viewed the dis-playbinocularly
from a distanceof90
cm while theirhead
was supportedby
aheadrest,
The
experiment was conductedin
adark
room.StimulL
Theimagesofhumanfigureswalkingon
a treadmillwere designed with
Curious
Labs
Poser
4
software. The figures
faced
either leftor right, and they subtended 2.9×1.2
degrees
of visual angle inheight
and width, respectively. The walkerscom-pleted
one step-cycle ina periodof 1,2seconds,The image of each
human
figure
wassuperim-posed on asinusoidal counterphase grating that
sub-tended
5
degrees
of visual anglein
beth
height
andwiqth.
The
spatial and temporalfrequencies
were 4 cycles per degree and8
Hz,
respectively. Thelumi-nance contrast was 60% and the mean
luminance
was 30cdlm2. The luminance contrast was
multi-pliedby a 2-D Gaussian envelepe with a standard
deviation
of1
degree
of visual angle. The stirnuluswas presented
in
a uniform gray fieldwiththe
meanluminance
of thegrating,The stimulusduration
was O.48seconds.
Procedure.
Eleven
observers were asked tore-port theirperceptual impre$sion$ ofthe gratingwith
aforcedchoice of three alternatives:
flickering,
drift-ing
left,
ordrifting
right,The
twodirectional
re-sponses were classified as an '`oppositeresponse" or a
"same
response" according
to
theirrelationship to the walkingdirection,
Each
participantobserved, and reported on,16
trialsin
which the figurewalkedforward and 16 trialsin which
the
figure
walkedbackward.
Results
For the forward walker, the opposite response was
obtained for
67,1%
of the trials,the sarne responsefor
5,1%
of the trials,and the flickerresponse for27.8%
ofthe trials.For thebackward walker,38.6%
of the responses were
the
opposite response,but
36.4% were
the
same response and25.0%
were theflickerresponse.
Analysis
of thedata with two-tairedt-tests
indicated
that the opposite responsefor
theforward
walker occurred significantly more oftenthan the oppesite response or thesame response for
the
backward
walker, ts(1O)>3.25,Ps<.O1,
Discussion
The
forward
walkerinduced
an apparent motien of the gratingin
adirection
opposite tothe walkingwith the
high
probability.Onthe
otherhand,
images
of the
backward
walker induced apparent motionin
the opposite and same
directions
with the equalprobabilities,and the percentages were significantly
lower
than that for the opposite responsefor
theforward
walker,These
resultsindicate
thatacorre-spondence of
direction
between
human motion andform iscritical ior thc illusion.This isconsistent with arecent model of biologicalmotion recognition, assuming that the recognition
is
achieved by aco-operation between motion and form processing
{Giese
&
Poggio,
2003). Learning also contributes to therecognition, and thismay accountfor
thearnbigu-ous results from the
backward
walker, Backwardwalking
is
rarely encountered ineveryday life.Our
findings
could impact onthe
current models ofmotion perceptienas
follows.
First,
low-level
motion signals could be modulated relative toa high-level representation of object motion which isdefined
by
biologicalmotion. Second. such modu]ation effects
extend toa relatively largevisual
field
over the area inwhichthe
object motionis
represented. The per-cept reflects a scene thatisfrequentlyexperiencedin
everyday life.Thus, ourfindings
suggest the exis-tcnce of a scene-based modulation of motion signalsby a high-levelperceptua] inference.
Referenees
Giese,M,A.
&
Poggio,
T. 2003 Neural mechanisms
for
the
recognition ofbiological
movements.ture
Reviews
Aiburoscience,
4,179-192.Lu,
Z L.&
Sperling,G.2001 Three-systems theory ofhuman
visual motion perception: review andupdate.