Depth perception from point-light biological motiondisplays

Marc H E de Lussanet $Institute for Psychology University of Munster

Munster Germany

Markus Lappe $Institute for Psychology University of Munster

Munster Germany

Humans have a clear impression of facing in depth for point-light biological motion However this has not been measuredsystematically nor is it known on which cues humans rely for their judgment In the present study subjects judged the facingorientation-in-depth of point-light displays The displays represented natural walking and modified versions in which the timesequence was reversed action was perturbed the limbs and joints were nonrigid the temporal sequence was scrambledor the joint positions were scrambled We found that the subjects were best at judging the facing direction of normal andreversed walking with an accuracy of 68 and 108 precision The results show that pendular motion of the limb segments andthe implicit knowledge of the human body play an important role for the precision of the judgment Three further factors wererelevant for the judgment of facing direction (a) the discrimination of the front and back side (b) the facing bias and (c) theimpression of depth from the display probably due to the kinetic depth effect The latter influences the accuracy whichdiffered strongly between subjects The results suggest that the facing bias to perceive the figure as facing toward theobserver rather than away is not related to the recognition of a human figure but rather to the presence of oscillatingmovements of the dots in the display

Keywords human walking pendular motion depth cues kinetic depth effect oblique effect repulsion effect facing bias

Citation de Lussanet M H E amp Lappe M (2012) Depth perception from point-light biological motion displays Journal ofVision 12(11)14 1ndash12 httpwwwjournalofvisionorgcontent121114 doi101167121114

Introduction

Point-light biological motion is possibly the mostabstracted and impoverished visual display of actionsand humans are remarkably good at its recognition(Johansson 1973) Typical studies have used profileviews of walking but human observers have a strongimpression of depth if an oblique or frontal view isdisplayed (Bulthoff Bulthoff amp Sinha 1998) In normallife walking may indeed be one of the most commonwhole-body activities that we may see but a pure sideview is rather uncommon Instead it is relevant toperceive in which direction a person is heading or whichorientation the person has to the observer

Perception of 3D orientation has been systematicallystudied for objects and faces For example rigid 3Dobjects and faces can be best recognized from canonicalviews (Blanz Tarr amp Bulthoff 1999) and humans tendto choose preferred views when learning 3D objects andfaces (Harries Perrett amp Lavender 1991 Peissig ampTarr 2006 Perrett amp Harries 1988) It has also beenshown that local and global strategies are combined ina flexible manner when judging the facing in depth of3D objects (Foster amp Gilson 2002 Troje amp Bulthoff1996 Watson Johnston Hill amp Troje 2005) Different

views of objects and faces are represented by differentpopulations of cells in the ventral temporal cortex inhumans (Grill-Spector et al 1999) In monkeys it hasbeen found that these populations are typicallyorganized in orientation maps (Wang Tanifuji ampTanaka 1998)

For point-light biological motion it has been foundthat some neurons in the monkey inferotemporal cortexare selective for the 3D orientation (facing in depth)(Vangeneugden et al 2011) Also in humans there isevidence for such orientation maps (Michels Kleiser deLussanet Seitz amp Lappe 2009) Psychophysical exper-iments showed that humans can detect changes in theorientation in depth even if they occur during a saccade(Verfaillie amp de Graef 2000) Humans can also easilydiscriminate between forwards and reversed walkingfrom almost every viewpoint even from single framelifetime point lights (Kuhlmann de Lussanet amp Lappe2009) An exception to this is the frontal view in whichthe performance was quite bad and depended entirelyon the presence of the foot-point lights in the display

Two kinds of depth information are available forseeing depth from a point light display which we shalladdress as implicit and explicit depth cues Withexplicit depth cues we mean cues that can be presentin any kind of visual stimulus such as occlusion

Journal of Vision (2012) 12(11)14 1ndash12 1httpwwwjournalofvisionorgcontent121114

doi 10 1167 12 11 14 ISSN 1534-7362 2012 ARVOReceived May 5 2012 published October 22 2012

binocular disparity and perspective deformationsExplicit depth cues do contribute to the perception ofdepth (Vanrie Dekeyser amp Verfaillie 2004) Withimplicit depth cues we mean cues for which implicitknowledge of the human body and its movements isneeded Current theories of biological motion percep-tion are based on such implicit knowledge (Giese 2004Lange amp Lappe 2006 Lee amp Wong 2004) Aninteresting feature of implicit depth cues in point-lightdisplays is that they are ambiguous for symmetricactions This means that a frontal view of walking is thesame as a back view (see also the more detailedexplanation in the Methods section Figure 2C) In thepresent study we will only address implicit depth cues

Humans have a clear impression of whether a point-light display faces toward or away from the observereven if explicit depth cues are absent (Brooks et al2008 Jackson Cummins amp Brady 2008 SchoutenTroje Brooks van der Zwan amp Verfaillie 2010Schouten Troje amp Verfaillie 2011 Schouten TrojeVroomen amp Verfaillie 2011 Sweeny Haroz ampWhitney 2012 Vanrie et al 2004 Vanrie amp Verfaillie2006) Explicit depth cues can disambiguate the point-light display and thus determine which of the twoambiguous facing directions an observer sees (Jacksonamp Blake 2010 Vanrie et al 2004)

However it is unknown how well humans canactually judge the facing in depth of point-lightbiological motion and neither is it known on the basisof what sort of implicit information they do so Theanswers to these questions are relevant because they mayhelp understand the underlying mechanisms that thebrain uses to recognize biological motion Such infor-mation may include dynamic and static cues A numberof dynamic cues have been proposed for recognizing aside view of point-light walking which could in principlebe of help to recognize the facing in depth

First we have the knowledge that human move-ments are typically pendular in the sense that the limbsegments are rigid so that the movement patterns of themajor joints typically revolve with respect to theadjacent joints (Chang amp Troje 2009 Hoffman ampFlinchbaugh 1982 Jokisch amp Troje 2003 Webb ampAggarwal 1982) Second we have implicit knowledgeof the trajectories of individual body parts such aswrists and ankles (Troje amp Westhoff 2006) Third wehave implicit knowledge of the entire movementpattern (Giese 2004) Fourth it has been proposedthat humans use their implicit knowledge of the bodyform at each stage of a movement which is a motion-independent cue (Beintema amp Lappe 2002 LangeGeorg amp Lappe 2006) Finally local dynamic cues(Neri 2009) or local static cues such as the ratio of thedistance between the hips or shoulders over the lengthof limb segments could be informative

In the present study we aimed to measure how wellhuman observers can tell the facing in depth in the

absence of explicit depth cues and to extract the relativeimportance of the different implicit cues mentionedabove Five manipulations were applied to a recordedpattern of human walking These manipulated system-atically the naturalness of the motion the rigidstructure of the body segments the body structurethe movement of the dots and the underlyingcoherency of the points These stimuli faced in arandom horizontal direction The task of the subjectswas to report the facing direction as precisely aspossible In the Discussion we will address the questionof which mechanisms might underlie the depthperception from implicit depth cues on the basis ofthe kinds of errors that the observers make

Methods

Subjects

The nine participants were undergraduate studentsfrom the psychology department They received creditpoints for experimental participation in the regularundergraduate program All participants had normalor corrected to normal vision and no known history ofneurological disorders

Stimuli and setup

The stimuli were computed from a walking cycle of a20-year old female with a regular relatively symmetricgait (note that most peoplersquos gait is surprisinglyasymmetric) A MotionStar (Ascension) motion cap-ture system recorded the 3D motions of markersattached to the body The horizontal translation wassubtracted Then a single regular gait cycle (beginningat right heel ground contact) was selected The smallmismatches of the start and end of each jointrsquostrajectory were removed by adding linear gains Therecordings were then smoothed to remove high-frequency recording noise

The bounding box (the 3D volume that justcontained the trajectories of all points) was computedas well as the length of the upper and lower limbsegments and the hip and shoulder segments The hip-and shoulder-joint-angles were defined between theupper limb segment and the trunk The knee elbowhip and shoulder joint angles as well as the orientationof the trunk hip and shoulder segments over time werecomputed See also the Supplementary Quicktimemovies

The normal condition was an orthogonally projectedview of the point-light walking stimulus without

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 2

occlusion (Figure 1A) The following scrambledversions were all computed from the normal stimulus

Reversed was the normal cycle played backwards(Figure 1B)

For marionette the structure of the body was leftintact The limb segments all had the same lengths andthe joints bended through the same range as duringnormal walking (Figure 1C) The manipulation shiftedthe timing of the bending of the joints The stimuluswas therefore biomechanically possible at all timeswhereas the movement kinematics were very unnaturalAlso the joints of the left and right side of the bodyfollowed different trajectories The stimulus thus gavethe impression of a marionette

For rubber limbs the spatial trajectories of the jointswere left intact but the timing of these trajectories wasshuffled (Figure 1D) Consequently the limb segmentschanged their lengths during the walking cycle and theknees and elbows could adopt angles in the non-physiological range The resulting stimulus was biome-chanically impossible because the limb segments werenot rigid and because the joints bent in non-physio-logical directions This gave the stimulus a nonhumanrubber-like impression

In frame-scrambled on each time frame a randomphase of the normal stimulus was displayed so therewere no trajectories and no apparent motion of thepoints (Figure 1E)

For the position-scrambled stimulus each pointfollowed the normal trajectory and phase but with arandom offset in 3D (Figure 1F) These offsets weresuch that the bounding box was the same size as thenormal stimulus This stimulus does not look humanand uninformed observers would never guess that thestimulus is related to human movement in any way

The relationships between the joints of one limb are notrecognizable

Table 1 summarizes how the stimuli are graduallymore different from the normal stimulus Listed are theimplicit stimulus-related cues that are available forestimating the facing in depth

Procedure

The participant was acquainted with the six differentstimulus types It was then explained that he or she wasto judge the exact direction in depth in which thestimulus was facing After two full cycles (276 s) thestimulus was replaced by a circular mouse-controlledcursor (Figure 2A) The cursor could be rotatedtowards the desired direction by dragging the mouseWhen the cursor direction corresponded with theperceived facing in depth the participant pressed thespace bar and the next stimulus appeared

Each of the six stimulus types (Figure 1) waspresented 60 times Each stimulus was simulated witha random facing in depth (a) The trials were blockedby stimulus type The order of the stimulus types wasbalanced between the participants

Analysis

Three factors are essential to analyze the results thefacing bias the accuracy and the precision

The facing bias

First the facing-towards-the-observer bias wascomputed as the percentage of responses between 08

Figure 1 Schemas of the different point-light stimuli in side-view (08) In grey the underlying body structure Blue and green depict the

trajectories of the visible point-lights The real stimuli consisted only of the point-lights (A) Normal walking (B) Reversed walking (cf

arrow heads in the trajectories) (C) In marionette the body structure and joint angles are natural but each joint of a pair follows a different

unusual trajectory due to a phase offset (D) In rubber limbs the joints follow the normal trajectories but at the wrong time The body

segments are nonrigid and the joints bend in unnatural directions (E) In frame-scrambled the human figure is intact There is no apparent

motion and the point-lights do not follow trajectories because a random phase is presented at every frame (F) In position-scrambled the

joints follow their intact trajectories at the correct time but the wrong location The body structure and relative movements are destroyed

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 3

and 1808 To avoid a possible influence from the stimuliof the previous block only the last 50 trials of eachblock were used for this measure

A profile view neither faces into the screen nor out ofit so it has no bias Therefore only trials in which boththe facing direction of the stimulus and the responsedeviated more than 108 from a side view were included(84 for frame-scrambled 90 for position-scram-bled and 94 for the other conditions)

The accuracy or mean error

The mean error was computed as a measure of theaccuracy For this a positive error was defined in the

direction away from the plane of the display screen(grey areas in Figure 2D) Thus for a side view (ie 08

or 1808) any errors were positive by definition and inthe range 08ndash908 For a front-back view (ie 908 or2708) any errors were negative by definition and in therange 08 to 908 (white range in Figure 2D) Forexample if a stimulus at 458 was responded with 488this would be an error ofthorn38 If the same stimulus wasperceived as facing away (ie 3158) an error of thorn38

would be obtained by responding 3128 Note that thisdefinition has the advantage that it is symmetric withrespect to the midline so that errors in the twohemifields are directly comparable and can be aver-aged

Figure 2 Definition of facing angle a (A) Screen shot of the response mask The white arrow (currently pointing at 888 out of the plane of

screen) was controlled by the mouse (B) Schematic top view of the experimental setting showing the correspondence of the setting in the

response mask and the perceived depth orientation (C) Any given configuration of points corresponds with two depth orientations ndasha and

a which are mirrored in the plane of the display screen (D) The relationship between the angle of facing in depth perceived angles and

errors Correct responses are located on the diagonal grey lines Since a represents the same configuration as ndasha two responses are

correct for any facing angle (except the side views) Errors towards the midplane (908 and 2708) are positive (grayed regions see text)

Stimulus Human form Human movement Pendular motion Point trajectories Local motions

Normal Correct Correct Correct Correct Correct

Reversed Correct Dynamically incorrect Correct Correct Dynamically incorrect

Marionette Correct Dynamically incorrect Correct Dynamically incorrect Dynamically incorrect

Rubber Physically

impossible

Physically

impossible

Physically

impossible

Correct Correct

Frame-scrambled Correct ndash ndash ndash ndash

Position-scrambled ndash ndash Physically

impossible

Correct Correct

Table 1 Presence of implicit cues that might be used for judging the facing in depth in the stimuli used in the experiment Dynamically

incorrect The stimulus is physically consistent with the human body but not normally performed this way Physically incorrect The cue is

Physically impossible eg the joints bend in unnatural directions and limb segments are non-rigid A ndash symbol The cue is absent

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 4

Also note that the measure is insensitive to aconfusion of the front and backside of a stimulus Thiscan occur for example with the position-scrambledstimulus (Figure 1F) which does not have a uniquelydefined front or backside If we apply the examplewhile the observed confuses front and back he wouldnot respond at 488 but at 1808 thorn 488 frac14 2288 whichwould still be an error of 38 (with respect to 2258 theleft-oblique away direction)

The precision or standard deviation

As a measure of the precision of the settings thestandard deviation of the error was computed As areference for the reader A homogeneous randomdistribution of responses would result in a standarddeviation of 3708 whereas the expectancy for givingthe same response on each trial (eg the plane of thedisplay) is 2628

Statistics

The data were plotted with the package R (Version213 with the Rapp GUI for Mac) A locally weightedscatterplot smoothing (LOWESS) moving-average witha window of 108 was fitted to the data of all subjects foreach stimulus type

For statistical comparisons between conditionsrepeated measures analyses of variance (ANOVAs) werecomputed (with stimulus type as a within-factor) Forthe planned pairwise comparisons Fisherrsquos partial least

squares difference test (PLSD) was used In a number ofcases the results were also tested against chance level orthe expectancy level for random responses In these casessigned t tests were used with a Bonferroni-Holmcorrection for multiple comparisons (Holm 1979)

Results

Facing-the-observer bias

The percentage of facing-the-observer responses ispresented in Figure 3 An ANOVA with scramble typeas within-factor revealed a significant effect of scrambletype F(5 48) frac14 45 p frac14 0003 Fisherrsquos PLSD testsconfirmed that frame-scrambled differed significantlyfrom all other conditions (p 002 Figure 3)

T tests showed that the frame-scrambled conditionwas the only one that did not differ significantly fromthe 50 chance level (pfrac14 03) All the other conditionswere significantly above 50 (p 002 Bonferroni-Holm correction) Thus in all conditions but the frame-scrambled condition the observers had a systematic biasto see the stimuli as facing out of the plane of the screen

Error distributions

Figure 4 plots the responses for each of the stimuli asa function of the simulated facing in depth Correct

Figure 3 Facing-the-observer-bias The 100 is a maximal bias at 50 there is no bias The asterisk indicates a significant difference

from all other conditions

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 5

responses are located on the ascending diagonalwhereas the opposite facing direction is depicted bythe descending diagonal The latter kind of errors canoccur if the front and backside are confused Sucherrors were rare and most common in the reversed andmarionette conditions

A different kind of error is the tendency to respondalong the cardinal axes (08 908 1808 and 2708) Theseresponses are clustered around the horizontal 08 908and 1808 lines These errors were most common for therubber and frame-scrambled conditions with a prob-ability of a response within 38 of a cardinal axis ofabout five times the expectancy rate of a random

distribution For the normal reversed and marionette

stimuli this probability was about 25 times the

expected rate

In most conditions the moving average curves show

a slight S-shape In the range of 08ndash908 the curves are

slightly below the diagonal whereas for 908ndash1808 the

curves are somewhat higher This indicates that the

observers tended to underestimate how much the facing

in depth deviated from the plane of the display Due to

the S-shape the errors tend to be negative (cf Figure

2D) In other words the distributions indicate that the

precision tends to be negative

Figure 4 Distribution of the responses in each condition as a function of the simulated facing in depth (a) Each subject is represented by a

different symbol Note that responses between 1808ndash3608 are converted to 1808ndash08 (cf Figure 2D and the inset in the upper left panel) 08

facing right Red curves LOWESS moving-average fit with a window of 108 Ascending diagonal correct responses descending diagonal

wrong facing direction

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 6

Accuracy and precision (mean and variableerrors)

Figure 5 presents the two error measures Note thatthe two panels have the same scale The error bars aremuch larger for the mean errors (the accuracy Panel A)than for the variable errors (the precision Panel B)This reflects that the S-shape in the distributions ofFigure 4 (see above) was strongly expressed for somesubjects but not for all

The S-shape in the response distributions resulted ina negative mean error of68 for the normal stimuli (theaccuracy Figure 5A) The ANOVA on the mean errorsshowed a significant main effect of scrambling typeF(5 48) frac14 106 p 00001 Fisherrsquos PLSD post-hoctests revealed that the differences between position-scrambled and all other conditions as well as betweenframe-scrambled and the rubber and marionetteconditions were statistically significant (p 001)The normal condition differed only significantly fromthe position-scrambled condition T tests showed thatthe mean errors differed significantly from zero in thenormal reversed frame-scrambled and position-scram-bled conditions (p 01 corrected) and did not in therubber condition (p 04)

The ANOVA on the variable errors (the precisionFigure 5B) also showed a significant main effect ofscrambling type F(5 48) frac14 440 p 00001 FisherrsquosPLSD post-hoc tests showed that the only conditionsthat did not differ significantly were normal andreversed and rubber and frame-scrambled (p 05)All other differences were highly significant (p 001)All conditions differed significantly from zero (p 001 corrected t tests) As explained in the Methodsthe optimal guessing strategy ie giving the sameresponse on every trial would result in a precision of2628 The precision was significantly lower than 2628

in all conditions (p 005 corrected) except theposition-scrambled stimuli (pfrac14 006)

The mean and variable errors of subjects were notcorrelated for any of the six conditions (r 026 p 05)

Control experiment Frame duration isirrelevant for frame scrambling

The frame-scrambled condition of main experimentwas presented with a 10-ms frame duration which veryshort and may have deteriorated the performance ofparticipants The control experiment included only theframe-scrambled conditions but with a longer frameduration of 80 or 160 ms With longer frame durationsthe subsequent configurations might have been recog-nized better and thus improved performance

A new group of 11 undergraduate students per-formed in the control experiment Each still configura-tion of the frame-scrambled stimulus remained on thescreen for either eight or 16 monitor refreshes (ie 80or 160 ms)

The repeated-measures ANOVAs on the meanerrors on the standard deviation of the error and onthe percentage of toward-responses did not reveal anysignificant main effects of frame duration F(1 20) 3p 01 These measures were all very similar to thosefound for the scrambled frame-order condition in themain experiment (mean error frac14 898 standarddeviation error frac14 1708 towards responses frac14 47)Three new ANOVAs with experiment as a between-factor confirmed this There was no significant maineffect of experiment for any of the three measures F(128) 10 p 03

The results of the control experiment confirm thatthe frame duration was not a limiting factor for the

Figure 5 (A) Mean error (accuracy) and (B) the standard deviation (precision) of the error NS the only nonsignificant differences (see

text) Error bars standard error

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 7

recognition of the facing direction in the mainexperiment The results also reproduce the lack of afacing bias in the frame-scrambled condition

Discussion

The results of the present study show that non-trained human observers can accurately judge thefacing in depth from two-dimensional point-lightbiological motion displays As we will discuss belowboth the familiar structure of the body and the relativependular motions of the joints contributed to this

The results revealed four different factors to thejudgment of the facing in depth of a point-light displayThese are (a) the discrimination of the front and backside (b) the facing bias (c) the impression of depthfrom the display and (d) the precision of the judgedfacing direction

Confusion of front and back

The participants were generally very good atdiscriminating the front and back sidse (cf Figure 4in all conditions but the position-scrambled there werefar more responses around the ascending dashed linesthan around the descending dashed lines) This is in linewith earlier findings in which the task was only todiscriminate between facing left and right (Beintema ampLappe 2002 for normal walking Lange amp Lappe2007 for frame-scrambled) The position-scrambledcondition showed a substantial number of confusionsof the front side and back side consistent with thefinding of Troje and Westhoff (2006) The present studyextends these findings to facing in depth orientations

Facing bias

We found a strong facing-toward-the-observer biasin most conditions This finding corroborates earlierfindings (Brooks et al 2008 Manera et al 2012Vanrie et al 2004 Vanrie amp Verfaillie 2006) Thesestudies all investigated the depth ambiguity inherent tobiological motion without explicit depth cues In somecases such as walking this ambiguity leads to a strongfacing bias (Vanrie et al 2004 Vanrie amp Verfaillie2006) More generally this kind of bias is known asdepth inversion (Yellott amp Kaiwi 1979) A well knownexample from research on face perception is the hollowface illusion (Hill amp Johnston 2007)

Many biological motion studies on the facing biashave sought the reason for this bias in social or actionrelated explanations Vanrie and Verfaillie (2006) found

large differences in the bias for different actions Theyfound that the presence of a facing bias was neitherrelated to semantic effects nor to whether the actionwas familiar or not A gender bias was reported Brookset al (2008) but not confirmed (Schouten et al 2010)Manera et al (2012) found that the facing bias forwalking disappears when the observer walks on atreadmill but it was unclear whether this was due to theactor and observer performing the same action or tothe observerrsquos head movements

Our results point more to a low-level explanation Astrong bias was present for the position-scrambledstimulus that did not present biological motion Incontrast the facing bias was absent in the frame-scrambled stimulus although subjects did recognize thehuman form in these stimuli We see two implications ofthese data

First the facing bias was absent if the points did notmove The only condition without apparent motionframe-scrambled was the only one without a facingbias This finding was replicated in the main and thecontrol experiments This absence cannot be explainedby a bad recognition of the stimulus because it isknown that human observers interpret the frame-scrambled stimulus as human walking (Lange amp Lappe2007) Also in the main experiment the subjectsrsquoperformance was as good as for the rubberrsquo condition(Figure 5B)

Second the bias did not depend on whether a humanfigure was present The facing-the-observer bias waspresent in the position-scrambled stimulus although nohuman walking can be recognized from such ascrambled stimulus Moreover the largest bias wasfound for the rubber stimulus This stimulus isunnaturally deformed as the joint angles bend inimpossible directions and the limbs are nonrigid

We therefore argue that the facing-the-observer biasis not related to biological motion mechanisms Insteadone could speculate about different mechanisms Forexample the bias might be related to the bias forlooming patterns as opposed to compressing patterns(Ball amp Sekuler 1980 Fahle amp Wehrhahn 1991Georgeson amp Harris 1978) Such a biased system willrespond more with lsquolsquoloomingrsquorsquo than with lsquolsquocompressingrsquorsquoto a stimulus of back and forth oscillating dots Thusintegrating over the biased looming and compressingsignals such a stimulus will appear to be approachingthe observer regardless whether it is human or not(Albright 1989 Lappe amp Rauschecker 1995 Rau-schecker von Grunau amp Poulin 1987) This explana-tion would be consistent with the mixed results thathave been found for various actions Vanrie andVerfaillie (2006) found that presentations with littlelateral oscillations of the dots resulted in little or nofacing bias

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 8

Impression of depth

If the stimulus looks essentially two-dimensional theobserver will tend to underestimate how much thestimulus faces out of the plane of the display (up to acertain degree when the displayed facing direction isobviously close to frontal) As implied by the largeerror bars in Figure 5A some subjects generally had agood impression of depth resulting in a high accuracywhereas others saw little depth in the stimuli leading toa systematic negative mean error

The number of responses along the cardinal axes ismore suitable to compare the impression of depthbetween the six conditions Such errors were rare forthe three conditions with correct pendular motion (cfTable 1) but frequent for the other three conditionsThis suggests that pendular motion of the points withrespect to each other is an important cue to conveydepth This is in accordance with earlier suggestions(Hoffman amp Flinchbaugh 1982 Johansson 1973Kuhlmann et al 2009) The tendency to respond alongthe two cardinal axes is also known for the kineticdepth effect and is called the repulsion effect (Hiris ampBlake 1996)

Our results are consistent with a kinetic depth effectin combination with an oblique repulsion Neverthe-less biological motion is a special case in this respectbecause the kinetic depth is typically defined for rigidobjects biological motion is nonrigid The kineticdepth in this case can only function if the observer usesimplicit knowledge of the body structure and themotions that the human body is capable of

Cues to depth

The conditions of the main experiment weredesigned to systematically change the stimulus indifferent ways to remove some information whileleaving other information intact (cf Table 1)

The human form was intact in four conditionsnormal reversed marionette and frame-scrambledThe rubber conditions were physically impossiblebecause the knees and elbows became overextendedand the limb segments stretched and compressed duringthe movement Only the normal stimuli presenteddynamically correct human movements because undernormal gravity conditions the marionette movementscannot be performed and forward walking played inreverse temporal order differs from real backwardwalking (even if the marionette and reversed stimuli arekinematically and biomechanically valid) Intact pen-dular relative movements of pairs of point-lights werepresent in the normal reversed and the marionettestimulus Finally intact point trajectories were presentin the normal the rubber and the position-scrambledstimulus whereas the local motions are dynamically

incorrect for reversed stimuli (cf human movements)Comparisons of accuracy and precision betweendifferent conditions thus allow us to estimate the roleof the different depth cues

Accuracy differed widely between subjects asexpressed by the large error bars in Figure 5A Thewithin-subject ANOVA showed a clear difference onlyfor the position-scrambled stimulus This indicates thatpoint trajectories and local motions alone give onlypoor depth information

The precision of the judgments is shown in Figure5B In this respect the subjects were very similar butthere were large differences between the conditions

The precision was significantly better in the normaland reversed conditions than in all other conditionsThis suggests that the human form was an importantcue On the other hand neither the human movementnor the local motions were important because bothwere dynamically incorrect in the reversed condition(cf Table 1)

Comparing the marionette to the normal stimulusreveals a small but significant decrease in precision Themarionette is consistent with the human form at alltimes but dynamically incorrect with respect to humanmovement point trajectories and local motions Thisreiterates the importance of human form and alsosuggests that point trajectories might be importantHowever one should take into account that theconfiguration of the human form in the marionettestimulus is less familiar than in the normal or reversedconditions

Precision for the frame-scrambled condition wassignificantly better than for the position-scrambledcondition (Figure 5B) This is consistent with thefinding that the human form is an important cueHowever the frame-scrambled condition was signifi-cantly worse than the normal and reversed conditionsalthough all three conditions contain veridical humanform information The difference between the frame-scrambled condition and the normal and reversedconditions lies in pendular motion and point trajecto-ries which are both absent in the frame-scrambledcondition (cf Table 1) The rubber and marionettestimuli help to resolve this issue Marionette has correctpendular motion but incorrect point trajectoriesRubber has incorrect pendular motion but correctpoint trajectories Since the precision was better inmarionette than in rubber stimuli the pendular motionsmust be important for judging the facing in depth

This is consistent with a recent finding that thepresence of points for the feet improves the detection ofsmall angular deviations from a frontal view of walking(Cai Yang Chen amp Jiang 2011 Kuhlmann et al2009) Indeed the notion that human movements aretypically pendular has been suggested as an importantcue for recognizing biological motion long ago (Hoff-man amp Flinchbaugh 1982 Johansson 1973 Webb amp

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 9

Aggarwal 1982) In a slightly different manner this hasalso been proposed more recently by Neri (2009)

Lastly even the precision for the position-scrambledcondition was marginally better than the expectancyvalue for giving the same response on each trial (ie2628) Thus the rhythmic oscillatory movements ofthe individual points may have provided some infor-mation about the simulated facing in depth

Altogether our results support two major cues forthe judgment of facing in depth The first is the familiarhuman form The importance of the familiarity of bodyform for judging the orientation in depth of biologicalmotion is consistent with electrophysiological findingsin monkeys (Vangeneugden et al 2011) and psycho-physical findings in humans (Bulthoff et al 1998Jackson amp Blake 2010 Kuhlmann et al 2009Verfaillie amp de Graef 2000) The second is the pendularmovement of the limbs (Hoffman amp Flinchbaugh1982) The body is an articulated structure withinherent pendular motions Since these motions arenot random but highly characteristic these are apotential additional cue The data give support to theuse of pendular motion as an additional cue for judgingthe facing in depth

Supplementary material

There is a Quicktime movie for each of the sixconditions of the main experiment Each moviepresents without interruptions one walking cycle forthe facing in depth of 08 308 608 908 1208 1508 and1808

Acknowledgments

This study is funded by the German FederalMinistry of Education and Research BMBF [Grant01EC1003A] We thank Martin Habelbarth for his helpwith collecting the data We thank the anonymousreviewers for their constructive comments

Commercial relationships noneCorresponding author Marc H E de LussanetEmail lussanetwwudeAddress Psychologisches Institut Westf Wilhelms-Universitat Munster Munster Germany

References

Albright T D (1989) Centrifugal directionality bias inthe middle temporal visual area (MT) of the

macaque Visual Neuroscience 2 177ndash188[PubMed]

Ball K amp Sekuler R (1980) Human vision favorscentrifugal motion Perception 9(3) 317ndash325[PubMed]

Beintema J A amp Lappe M (2002) Perception ofbiological motion without local image motionProceedings of the National Academy of Sciencesof the USA 99 5661ndash5663 doi101073pnas082483699 [PubMed]

Blanz V Tarr M J amp Bulthoff H H (1999) Whatobject attributes determine canonical views Per-ception 28 575ndash599

Brooks A Schouten B Troje N F Verfaillie KBlanke O amp van der Zwan R (2008) Correlatedchanges in perceptions of the gender and orienta-tion of ambiguous biological motion figuresCurrent Biology 18(17) R728ndashR729 doi101016jcub200806054 [PubMed]

Bulthoff I Bulthoff H H amp Sinha P (1998) Top-down influences on stereoscopic depth-perceptionNature Neuroscience 1 254ndash257 [PubMed]

Cai P Yang X Chen L amp Jiang Y (2011) Motionspeed modulates walking direction discriminationThe role of the feet in biological motion perceptionChinese Science Bulletin 56(19) 2025ndash2030 doi101007s11434-011-4528-6

Chang D H F amp Troje N F (2009) Accelerationcarries the local inversion effect in biologicalmotion perception Journal of Vision 9(1)19 1ndash17 httpjournalofvisionorg9119 doi1011679119 [PubMed] [Article]

Fahle M amp Wehrhahn C (1991) Motion perceptionin the peripheral visual field Graefersquos Archive forClinical and Experimental Ophthalmology 229(5)430ndash436 doi101007BF00166305 [PubMed]

Foster D H amp Gilson S J (2002) Recognizing novelthree-dimensional objects by summing signals fromparts and views Proceedings of the Royal Society BBiological Sciences 269 1939ndash1947 doi101098rspb20022119

Georgeson M A amp Harris M G (1978) Apparentfoveofugal drift of counterphase gratings Percep-tion 7 527ndash536 [PubMed]

Giese M A (2004) Neural model for biologicalmovement recognition A neurophysiologicallyplausible theory In L M Vaina S A Beardsleyamp S Rushton (Eds) Optic flow and beyond (pp443ndash470) Dordrecht NL Kluwer Academic Pub-lishers

Grill-Spector K Kushnir T Edelman S AvidanG Itzchak Y amp Malach R (1999) Differential

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 10

and 1808 To avoid a possible influence from the stimuliof the previous block only the last 50 trials of eachblock were used for this measure

A profile view neither faces into the screen nor out ofit so it has no bias Therefore only trials in which boththe facing direction of the stimulus and the responsedeviated more than 108 from a side view were included(84 for frame-scrambled 90 for position-scram-bled and 94 for the other conditions)

The accuracy or mean error

The mean error was computed as a measure of theaccuracy For this a positive error was defined in the

direction away from the plane of the display screen(grey areas in Figure 2D) Thus for a side view (ie 08

or 1808) any errors were positive by definition and inthe range 08ndash908 For a front-back view (ie 908 or2708) any errors were negative by definition and in therange 08 to 908 (white range in Figure 2D) Forexample if a stimulus at 458 was responded with 488this would be an error ofthorn38 If the same stimulus wasperceived as facing away (ie 3158) an error of thorn38

would be obtained by responding 3128 Note that thisdefinition has the advantage that it is symmetric withrespect to the midline so that errors in the twohemifields are directly comparable and can be aver-aged

Figure 2 Definition of facing angle a (A) Screen shot of the response mask The white arrow (currently pointing at 888 out of the plane of

screen) was controlled by the mouse (B) Schematic top view of the experimental setting showing the correspondence of the setting in the

response mask and the perceived depth orientation (C) Any given configuration of points corresponds with two depth orientations ndasha and

a which are mirrored in the plane of the display screen (D) The relationship between the angle of facing in depth perceived angles and

errors Correct responses are located on the diagonal grey lines Since a represents the same configuration as ndasha two responses are

correct for any facing angle (except the side views) Errors towards the midplane (908 and 2708) are positive (grayed regions see text)

Stimulus Human form Human movement Pendular motion Point trajectories Local motions

Normal Correct Correct Correct Correct Correct

Reversed Correct Dynamically incorrect Correct Correct Dynamically incorrect

Marionette Correct Dynamically incorrect Correct Dynamically incorrect Dynamically incorrect

Rubber Physically

impossible

Physically

impossible

Physically

impossible

Correct Correct

Frame-scrambled Correct ndash ndash ndash ndash

Position-scrambled ndash ndash Physically

impossible

Correct Correct

Table 1 Presence of implicit cues that might be used for judging the facing in depth in the stimuli used in the experiment Dynamically

incorrect The stimulus is physically consistent with the human body but not normally performed this way Physically incorrect The cue is

Physically impossible eg the joints bend in unnatural directions and limb segments are non-rigid A ndash symbol The cue is absent

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 4

Also note that the measure is insensitive to aconfusion of the front and backside of a stimulus Thiscan occur for example with the position-scrambledstimulus (Figure 1F) which does not have a uniquelydefined front or backside If we apply the examplewhile the observed confuses front and back he wouldnot respond at 488 but at 1808 thorn 488 frac14 2288 whichwould still be an error of 38 (with respect to 2258 theleft-oblique away direction)

The precision or standard deviation

As a measure of the precision of the settings thestandard deviation of the error was computed As areference for the reader A homogeneous randomdistribution of responses would result in a standarddeviation of 3708 whereas the expectancy for givingthe same response on each trial (eg the plane of thedisplay) is 2628

Statistics

The data were plotted with the package R (Version213 with the Rapp GUI for Mac) A locally weightedscatterplot smoothing (LOWESS) moving-average witha window of 108 was fitted to the data of all subjects foreach stimulus type

For statistical comparisons between conditionsrepeated measures analyses of variance (ANOVAs) werecomputed (with stimulus type as a within-factor) Forthe planned pairwise comparisons Fisherrsquos partial least

squares difference test (PLSD) was used In a number ofcases the results were also tested against chance level orthe expectancy level for random responses In these casessigned t tests were used with a Bonferroni-Holmcorrection for multiple comparisons (Holm 1979)

Results

Facing-the-observer bias

The percentage of facing-the-observer responses ispresented in Figure 3 An ANOVA with scramble typeas within-factor revealed a significant effect of scrambletype F(5 48) frac14 45 p frac14 0003 Fisherrsquos PLSD testsconfirmed that frame-scrambled differed significantlyfrom all other conditions (p 002 Figure 3)

T tests showed that the frame-scrambled conditionwas the only one that did not differ significantly fromthe 50 chance level (pfrac14 03) All the other conditionswere significantly above 50 (p 002 Bonferroni-Holm correction) Thus in all conditions but the frame-scrambled condition the observers had a systematic biasto see the stimuli as facing out of the plane of the screen

Error distributions

Figure 4 plots the responses for each of the stimuli asa function of the simulated facing in depth Correct

Figure 3 Facing-the-observer-bias The 100 is a maximal bias at 50 there is no bias The asterisk indicates a significant difference

from all other conditions

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 5

responses are located on the ascending diagonalwhereas the opposite facing direction is depicted bythe descending diagonal The latter kind of errors canoccur if the front and backside are confused Sucherrors were rare and most common in the reversed andmarionette conditions

A different kind of error is the tendency to respondalong the cardinal axes (08 908 1808 and 2708) Theseresponses are clustered around the horizontal 08 908and 1808 lines These errors were most common for therubber and frame-scrambled conditions with a prob-ability of a response within 38 of a cardinal axis ofabout five times the expectancy rate of a random

distribution For the normal reversed and marionette

stimuli this probability was about 25 times the

expected rate

In most conditions the moving average curves show

a slight S-shape In the range of 08ndash908 the curves are

slightly below the diagonal whereas for 908ndash1808 the

curves are somewhat higher This indicates that the

observers tended to underestimate how much the facing

in depth deviated from the plane of the display Due to

the S-shape the errors tend to be negative (cf Figure

2D) In other words the distributions indicate that the

precision tends to be negative

Figure 4 Distribution of the responses in each condition as a function of the simulated facing in depth (a) Each subject is represented by a

different symbol Note that responses between 1808ndash3608 are converted to 1808ndash08 (cf Figure 2D and the inset in the upper left panel) 08

facing right Red curves LOWESS moving-average fit with a window of 108 Ascending diagonal correct responses descending diagonal

wrong facing direction

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 6

Accuracy and precision (mean and variableerrors)

Figure 5 presents the two error measures Note thatthe two panels have the same scale The error bars aremuch larger for the mean errors (the accuracy Panel A)than for the variable errors (the precision Panel B)This reflects that the S-shape in the distributions ofFigure 4 (see above) was strongly expressed for somesubjects but not for all

The S-shape in the response distributions resulted ina negative mean error of68 for the normal stimuli (theaccuracy Figure 5A) The ANOVA on the mean errorsshowed a significant main effect of scrambling typeF(5 48) frac14 106 p 00001 Fisherrsquos PLSD post-hoctests revealed that the differences between position-scrambled and all other conditions as well as betweenframe-scrambled and the rubber and marionetteconditions were statistically significant (p 001)The normal condition differed only significantly fromthe position-scrambled condition T tests showed thatthe mean errors differed significantly from zero in thenormal reversed frame-scrambled and position-scram-bled conditions (p 01 corrected) and did not in therubber condition (p 04)

The ANOVA on the variable errors (the precisionFigure 5B) also showed a significant main effect ofscrambling type F(5 48) frac14 440 p 00001 FisherrsquosPLSD post-hoc tests showed that the only conditionsthat did not differ significantly were normal andreversed and rubber and frame-scrambled (p 05)All other differences were highly significant (p 001)All conditions differed significantly from zero (p 001 corrected t tests) As explained in the Methodsthe optimal guessing strategy ie giving the sameresponse on every trial would result in a precision of2628 The precision was significantly lower than 2628

in all conditions (p 005 corrected) except theposition-scrambled stimuli (pfrac14 006)

The mean and variable errors of subjects were notcorrelated for any of the six conditions (r 026 p 05)

Control experiment Frame duration isirrelevant for frame scrambling

The frame-scrambled condition of main experimentwas presented with a 10-ms frame duration which veryshort and may have deteriorated the performance ofparticipants The control experiment included only theframe-scrambled conditions but with a longer frameduration of 80 or 160 ms With longer frame durationsthe subsequent configurations might have been recog-nized better and thus improved performance

A new group of 11 undergraduate students per-formed in the control experiment Each still configura-tion of the frame-scrambled stimulus remained on thescreen for either eight or 16 monitor refreshes (ie 80or 160 ms)

The repeated-measures ANOVAs on the meanerrors on the standard deviation of the error and onthe percentage of toward-responses did not reveal anysignificant main effects of frame duration F(1 20) 3p 01 These measures were all very similar to thosefound for the scrambled frame-order condition in themain experiment (mean error frac14 898 standarddeviation error frac14 1708 towards responses frac14 47)Three new ANOVAs with experiment as a between-factor confirmed this There was no significant maineffect of experiment for any of the three measures F(128) 10 p 03

The results of the control experiment confirm thatthe frame duration was not a limiting factor for the

Figure 5 (A) Mean error (accuracy) and (B) the standard deviation (precision) of the error NS the only nonsignificant differences (see

text) Error bars standard error

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 7

recognition of the facing direction in the mainexperiment The results also reproduce the lack of afacing bias in the frame-scrambled condition

Discussion

The results of the present study show that non-trained human observers can accurately judge thefacing in depth from two-dimensional point-lightbiological motion displays As we will discuss belowboth the familiar structure of the body and the relativependular motions of the joints contributed to this

The results revealed four different factors to thejudgment of the facing in depth of a point-light displayThese are (a) the discrimination of the front and backside (b) the facing bias (c) the impression of depthfrom the display and (d) the precision of the judgedfacing direction

Confusion of front and back

The participants were generally very good atdiscriminating the front and back sidse (cf Figure 4in all conditions but the position-scrambled there werefar more responses around the ascending dashed linesthan around the descending dashed lines) This is in linewith earlier findings in which the task was only todiscriminate between facing left and right (Beintema ampLappe 2002 for normal walking Lange amp Lappe2007 for frame-scrambled) The position-scrambledcondition showed a substantial number of confusionsof the front side and back side consistent with thefinding of Troje and Westhoff (2006) The present studyextends these findings to facing in depth orientations

Facing bias

We found a strong facing-toward-the-observer biasin most conditions This finding corroborates earlierfindings (Brooks et al 2008 Manera et al 2012Vanrie et al 2004 Vanrie amp Verfaillie 2006) Thesestudies all investigated the depth ambiguity inherent tobiological motion without explicit depth cues In somecases such as walking this ambiguity leads to a strongfacing bias (Vanrie et al 2004 Vanrie amp Verfaillie2006) More generally this kind of bias is known asdepth inversion (Yellott amp Kaiwi 1979) A well knownexample from research on face perception is the hollowface illusion (Hill amp Johnston 2007)

Many biological motion studies on the facing biashave sought the reason for this bias in social or actionrelated explanations Vanrie and Verfaillie (2006) found

large differences in the bias for different actions Theyfound that the presence of a facing bias was neitherrelated to semantic effects nor to whether the actionwas familiar or not A gender bias was reported Brookset al (2008) but not confirmed (Schouten et al 2010)Manera et al (2012) found that the facing bias forwalking disappears when the observer walks on atreadmill but it was unclear whether this was due to theactor and observer performing the same action or tothe observerrsquos head movements

Our results point more to a low-level explanation Astrong bias was present for the position-scrambledstimulus that did not present biological motion Incontrast the facing bias was absent in the frame-scrambled stimulus although subjects did recognize thehuman form in these stimuli We see two implications ofthese data

First the facing bias was absent if the points did notmove The only condition without apparent motionframe-scrambled was the only one without a facingbias This finding was replicated in the main and thecontrol experiments This absence cannot be explainedby a bad recognition of the stimulus because it isknown that human observers interpret the frame-scrambled stimulus as human walking (Lange amp Lappe2007) Also in the main experiment the subjectsrsquoperformance was as good as for the rubberrsquo condition(Figure 5B)

Second the bias did not depend on whether a humanfigure was present The facing-the-observer bias waspresent in the position-scrambled stimulus although nohuman walking can be recognized from such ascrambled stimulus Moreover the largest bias wasfound for the rubber stimulus This stimulus isunnaturally deformed as the joint angles bend inimpossible directions and the limbs are nonrigid

We therefore argue that the facing-the-observer biasis not related to biological motion mechanisms Insteadone could speculate about different mechanisms Forexample the bias might be related to the bias forlooming patterns as opposed to compressing patterns(Ball amp Sekuler 1980 Fahle amp Wehrhahn 1991Georgeson amp Harris 1978) Such a biased system willrespond more with lsquolsquoloomingrsquorsquo than with lsquolsquocompressingrsquorsquoto a stimulus of back and forth oscillating dots Thusintegrating over the biased looming and compressingsignals such a stimulus will appear to be approachingthe observer regardless whether it is human or not(Albright 1989 Lappe amp Rauschecker 1995 Rau-schecker von Grunau amp Poulin 1987) This explana-tion would be consistent with the mixed results thathave been found for various actions Vanrie andVerfaillie (2006) found that presentations with littlelateral oscillations of the dots resulted in little or nofacing bias

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 8

Impression of depth

If the stimulus looks essentially two-dimensional theobserver will tend to underestimate how much thestimulus faces out of the plane of the display (up to acertain degree when the displayed facing direction isobviously close to frontal) As implied by the largeerror bars in Figure 5A some subjects generally had agood impression of depth resulting in a high accuracywhereas others saw little depth in the stimuli leading toa systematic negative mean error

The number of responses along the cardinal axes ismore suitable to compare the impression of depthbetween the six conditions Such errors were rare forthe three conditions with correct pendular motion (cfTable 1) but frequent for the other three conditionsThis suggests that pendular motion of the points withrespect to each other is an important cue to conveydepth This is in accordance with earlier suggestions(Hoffman amp Flinchbaugh 1982 Johansson 1973Kuhlmann et al 2009) The tendency to respond alongthe two cardinal axes is also known for the kineticdepth effect and is called the repulsion effect (Hiris ampBlake 1996)

Our results are consistent with a kinetic depth effectin combination with an oblique repulsion Neverthe-less biological motion is a special case in this respectbecause the kinetic depth is typically defined for rigidobjects biological motion is nonrigid The kineticdepth in this case can only function if the observer usesimplicit knowledge of the body structure and themotions that the human body is capable of

Cues to depth

The conditions of the main experiment weredesigned to systematically change the stimulus indifferent ways to remove some information whileleaving other information intact (cf Table 1)

The human form was intact in four conditionsnormal reversed marionette and frame-scrambledThe rubber conditions were physically impossiblebecause the knees and elbows became overextendedand the limb segments stretched and compressed duringthe movement Only the normal stimuli presenteddynamically correct human movements because undernormal gravity conditions the marionette movementscannot be performed and forward walking played inreverse temporal order differs from real backwardwalking (even if the marionette and reversed stimuli arekinematically and biomechanically valid) Intact pen-dular relative movements of pairs of point-lights werepresent in the normal reversed and the marionettestimulus Finally intact point trajectories were presentin the normal the rubber and the position-scrambledstimulus whereas the local motions are dynamically

incorrect for reversed stimuli (cf human movements)Comparisons of accuracy and precision betweendifferent conditions thus allow us to estimate the roleof the different depth cues

Accuracy differed widely between subjects asexpressed by the large error bars in Figure 5A Thewithin-subject ANOVA showed a clear difference onlyfor the position-scrambled stimulus This indicates thatpoint trajectories and local motions alone give onlypoor depth information

The precision of the judgments is shown in Figure5B In this respect the subjects were very similar butthere were large differences between the conditions

The precision was significantly better in the normaland reversed conditions than in all other conditionsThis suggests that the human form was an importantcue On the other hand neither the human movementnor the local motions were important because bothwere dynamically incorrect in the reversed condition(cf Table 1)

Comparing the marionette to the normal stimulusreveals a small but significant decrease in precision Themarionette is consistent with the human form at alltimes but dynamically incorrect with respect to humanmovement point trajectories and local motions Thisreiterates the importance of human form and alsosuggests that point trajectories might be importantHowever one should take into account that theconfiguration of the human form in the marionettestimulus is less familiar than in the normal or reversedconditions

Precision for the frame-scrambled condition wassignificantly better than for the position-scrambledcondition (Figure 5B) This is consistent with thefinding that the human form is an important cueHowever the frame-scrambled condition was signifi-cantly worse than the normal and reversed conditionsalthough all three conditions contain veridical humanform information The difference between the frame-scrambled condition and the normal and reversedconditions lies in pendular motion and point trajecto-ries which are both absent in the frame-scrambledcondition (cf Table 1) The rubber and marionettestimuli help to resolve this issue Marionette has correctpendular motion but incorrect point trajectoriesRubber has incorrect pendular motion but correctpoint trajectories Since the precision was better inmarionette than in rubber stimuli the pendular motionsmust be important for judging the facing in depth

This is consistent with a recent finding that thepresence of points for the feet improves the detection ofsmall angular deviations from a frontal view of walking(Cai Yang Chen amp Jiang 2011 Kuhlmann et al2009) Indeed the notion that human movements aretypically pendular has been suggested as an importantcue for recognizing biological motion long ago (Hoff-man amp Flinchbaugh 1982 Johansson 1973 Webb amp

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 9

Aggarwal 1982) In a slightly different manner this hasalso been proposed more recently by Neri (2009)

Lastly even the precision for the position-scrambledcondition was marginally better than the expectancyvalue for giving the same response on each trial (ie2628) Thus the rhythmic oscillatory movements ofthe individual points may have provided some infor-mation about the simulated facing in depth

Altogether our results support two major cues forthe judgment of facing in depth The first is the familiarhuman form The importance of the familiarity of bodyform for judging the orientation in depth of biologicalmotion is consistent with electrophysiological findingsin monkeys (Vangeneugden et al 2011) and psycho-physical findings in humans (Bulthoff et al 1998Jackson amp Blake 2010 Kuhlmann et al 2009Verfaillie amp de Graef 2000) The second is the pendularmovement of the limbs (Hoffman amp Flinchbaugh1982) The body is an articulated structure withinherent pendular motions Since these motions arenot random but highly characteristic these are apotential additional cue The data give support to theuse of pendular motion as an additional cue for judgingthe facing in depth

Supplementary material

There is a Quicktime movie for each of the sixconditions of the main experiment Each moviepresents without interruptions one walking cycle forthe facing in depth of 08 308 608 908 1208 1508 and1808

Acknowledgments

This study is funded by the German FederalMinistry of Education and Research BMBF [Grant01EC1003A] We thank Martin Habelbarth for his helpwith collecting the data We thank the anonymousreviewers for their constructive comments

Commercial relationships noneCorresponding author Marc H E de LussanetEmail lussanetwwudeAddress Psychologisches Institut Westf Wilhelms-Universitat Munster Munster Germany

References

Albright T D (1989) Centrifugal directionality bias inthe middle temporal visual area (MT) of the

macaque Visual Neuroscience 2 177ndash188[PubMed]

Ball K amp Sekuler R (1980) Human vision favorscentrifugal motion Perception 9(3) 317ndash325[PubMed]

Beintema J A amp Lappe M (2002) Perception ofbiological motion without local image motionProceedings of the National Academy of Sciencesof the USA 99 5661ndash5663 doi101073pnas082483699 [PubMed]

Blanz V Tarr M J amp Bulthoff H H (1999) Whatobject attributes determine canonical views Per-ception 28 575ndash599

Brooks A Schouten B Troje N F Verfaillie KBlanke O amp van der Zwan R (2008) Correlatedchanges in perceptions of the gender and orienta-tion of ambiguous biological motion figuresCurrent Biology 18(17) R728ndashR729 doi101016jcub200806054 [PubMed]

Bulthoff I Bulthoff H H amp Sinha P (1998) Top-down influences on stereoscopic depth-perceptionNature Neuroscience 1 254ndash257 [PubMed]

Cai P Yang X Chen L amp Jiang Y (2011) Motionspeed modulates walking direction discriminationThe role of the feet in biological motion perceptionChinese Science Bulletin 56(19) 2025ndash2030 doi101007s11434-011-4528-6

Chang D H F amp Troje N F (2009) Accelerationcarries the local inversion effect in biologicalmotion perception Journal of Vision 9(1)19 1ndash17 httpjournalofvisionorg9119 doi1011679119 [PubMed] [Article]

Fahle M amp Wehrhahn C (1991) Motion perceptionin the peripheral visual field Graefersquos Archive forClinical and Experimental Ophthalmology 229(5)430ndash436 doi101007BF00166305 [PubMed]

Foster D H amp Gilson S J (2002) Recognizing novelthree-dimensional objects by summing signals fromparts and views Proceedings of the Royal Society BBiological Sciences 269 1939ndash1947 doi101098rspb20022119

Georgeson M A amp Harris M G (1978) Apparentfoveofugal drift of counterphase gratings Percep-tion 7 527ndash536 [PubMed]

Giese M A (2004) Neural model for biologicalmovement recognition A neurophysiologicallyplausible theory In L M Vaina S A Beardsleyamp S Rushton (Eds) Optic flow and beyond (pp443ndash470) Dordrecht NL Kluwer Academic Pub-lishers

Grill-Spector K Kushnir T Edelman S AvidanG Itzchak Y amp Malach R (1999) Differential

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 10

Also note that the measure is insensitive to aconfusion of the front and backside of a stimulus Thiscan occur for example with the position-scrambledstimulus (Figure 1F) which does not have a uniquelydefined front or backside If we apply the examplewhile the observed confuses front and back he wouldnot respond at 488 but at 1808 thorn 488 frac14 2288 whichwould still be an error of 38 (with respect to 2258 theleft-oblique away direction)

The precision or standard deviation

As a measure of the precision of the settings thestandard deviation of the error was computed As areference for the reader A homogeneous randomdistribution of responses would result in a standarddeviation of 3708 whereas the expectancy for givingthe same response on each trial (eg the plane of thedisplay) is 2628

Statistics

The data were plotted with the package R (Version213 with the Rapp GUI for Mac) A locally weightedscatterplot smoothing (LOWESS) moving-average witha window of 108 was fitted to the data of all subjects foreach stimulus type

For statistical comparisons between conditionsrepeated measures analyses of variance (ANOVAs) werecomputed (with stimulus type as a within-factor) Forthe planned pairwise comparisons Fisherrsquos partial least

squares difference test (PLSD) was used In a number ofcases the results were also tested against chance level orthe expectancy level for random responses In these casessigned t tests were used with a Bonferroni-Holmcorrection for multiple comparisons (Holm 1979)

Results

Facing-the-observer bias

The percentage of facing-the-observer responses ispresented in Figure 3 An ANOVA with scramble typeas within-factor revealed a significant effect of scrambletype F(5 48) frac14 45 p frac14 0003 Fisherrsquos PLSD testsconfirmed that frame-scrambled differed significantlyfrom all other conditions (p 002 Figure 3)

T tests showed that the frame-scrambled conditionwas the only one that did not differ significantly fromthe 50 chance level (pfrac14 03) All the other conditionswere significantly above 50 (p 002 Bonferroni-Holm correction) Thus in all conditions but the frame-scrambled condition the observers had a systematic biasto see the stimuli as facing out of the plane of the screen

Error distributions

Figure 4 plots the responses for each of the stimuli asa function of the simulated facing in depth Correct

Figure 3 Facing-the-observer-bias The 100 is a maximal bias at 50 there is no bias The asterisk indicates a significant difference

from all other conditions

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 5

responses are located on the ascending diagonalwhereas the opposite facing direction is depicted bythe descending diagonal The latter kind of errors canoccur if the front and backside are confused Sucherrors were rare and most common in the reversed andmarionette conditions

A different kind of error is the tendency to respondalong the cardinal axes (08 908 1808 and 2708) Theseresponses are clustered around the horizontal 08 908and 1808 lines These errors were most common for therubber and frame-scrambled conditions with a prob-ability of a response within 38 of a cardinal axis ofabout five times the expectancy rate of a random

distribution For the normal reversed and marionette

stimuli this probability was about 25 times the

expected rate

In most conditions the moving average curves show

a slight S-shape In the range of 08ndash908 the curves are

slightly below the diagonal whereas for 908ndash1808 the

curves are somewhat higher This indicates that the

observers tended to underestimate how much the facing

in depth deviated from the plane of the display Due to

the S-shape the errors tend to be negative (cf Figure

2D) In other words the distributions indicate that the

precision tends to be negative

Figure 4 Distribution of the responses in each condition as a function of the simulated facing in depth (a) Each subject is represented by a

different symbol Note that responses between 1808ndash3608 are converted to 1808ndash08 (cf Figure 2D and the inset in the upper left panel) 08

facing right Red curves LOWESS moving-average fit with a window of 108 Ascending diagonal correct responses descending diagonal

wrong facing direction

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 6

Accuracy and precision (mean and variableerrors)

Figure 5 presents the two error measures Note thatthe two panels have the same scale The error bars aremuch larger for the mean errors (the accuracy Panel A)than for the variable errors (the precision Panel B)This reflects that the S-shape in the distributions ofFigure 4 (see above) was strongly expressed for somesubjects but not for all

The S-shape in the response distributions resulted ina negative mean error of68 for the normal stimuli (theaccuracy Figure 5A) The ANOVA on the mean errorsshowed a significant main effect of scrambling typeF(5 48) frac14 106 p 00001 Fisherrsquos PLSD post-hoctests revealed that the differences between position-scrambled and all other conditions as well as betweenframe-scrambled and the rubber and marionetteconditions were statistically significant (p 001)The normal condition differed only significantly fromthe position-scrambled condition T tests showed thatthe mean errors differed significantly from zero in thenormal reversed frame-scrambled and position-scram-bled conditions (p 01 corrected) and did not in therubber condition (p 04)

The ANOVA on the variable errors (the precisionFigure 5B) also showed a significant main effect ofscrambling type F(5 48) frac14 440 p 00001 FisherrsquosPLSD post-hoc tests showed that the only conditionsthat did not differ significantly were normal andreversed and rubber and frame-scrambled (p 05)All other differences were highly significant (p 001)All conditions differed significantly from zero (p 001 corrected t tests) As explained in the Methodsthe optimal guessing strategy ie giving the sameresponse on every trial would result in a precision of2628 The precision was significantly lower than 2628

in all conditions (p 005 corrected) except theposition-scrambled stimuli (pfrac14 006)

The mean and variable errors of subjects were notcorrelated for any of the six conditions (r 026 p 05)

Control experiment Frame duration isirrelevant for frame scrambling

The frame-scrambled condition of main experimentwas presented with a 10-ms frame duration which veryshort and may have deteriorated the performance ofparticipants The control experiment included only theframe-scrambled conditions but with a longer frameduration of 80 or 160 ms With longer frame durationsthe subsequent configurations might have been recog-nized better and thus improved performance

A new group of 11 undergraduate students per-formed in the control experiment Each still configura-tion of the frame-scrambled stimulus remained on thescreen for either eight or 16 monitor refreshes (ie 80or 160 ms)

The repeated-measures ANOVAs on the meanerrors on the standard deviation of the error and onthe percentage of toward-responses did not reveal anysignificant main effects of frame duration F(1 20) 3p 01 These measures were all very similar to thosefound for the scrambled frame-order condition in themain experiment (mean error frac14 898 standarddeviation error frac14 1708 towards responses frac14 47)Three new ANOVAs with experiment as a between-factor confirmed this There was no significant maineffect of experiment for any of the three measures F(128) 10 p 03

The results of the control experiment confirm thatthe frame duration was not a limiting factor for the

Figure 5 (A) Mean error (accuracy) and (B) the standard deviation (precision) of the error NS the only nonsignificant differences (see

text) Error bars standard error

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 7

recognition of the facing direction in the mainexperiment The results also reproduce the lack of afacing bias in the frame-scrambled condition

Discussion

The results of the present study show that non-trained human observers can accurately judge thefacing in depth from two-dimensional point-lightbiological motion displays As we will discuss belowboth the familiar structure of the body and the relativependular motions of the joints contributed to this

The results revealed four different factors to thejudgment of the facing in depth of a point-light displayThese are (a) the discrimination of the front and backside (b) the facing bias (c) the impression of depthfrom the display and (d) the precision of the judgedfacing direction

Confusion of front and back

The participants were generally very good atdiscriminating the front and back sidse (cf Figure 4in all conditions but the position-scrambled there werefar more responses around the ascending dashed linesthan around the descending dashed lines) This is in linewith earlier findings in which the task was only todiscriminate between facing left and right (Beintema ampLappe 2002 for normal walking Lange amp Lappe2007 for frame-scrambled) The position-scrambledcondition showed a substantial number of confusionsof the front side and back side consistent with thefinding of Troje and Westhoff (2006) The present studyextends these findings to facing in depth orientations

Facing bias

We found a strong facing-toward-the-observer biasin most conditions This finding corroborates earlierfindings (Brooks et al 2008 Manera et al 2012Vanrie et al 2004 Vanrie amp Verfaillie 2006) Thesestudies all investigated the depth ambiguity inherent tobiological motion without explicit depth cues In somecases such as walking this ambiguity leads to a strongfacing bias (Vanrie et al 2004 Vanrie amp Verfaillie2006) More generally this kind of bias is known asdepth inversion (Yellott amp Kaiwi 1979) A well knownexample from research on face perception is the hollowface illusion (Hill amp Johnston 2007)

Many biological motion studies on the facing biashave sought the reason for this bias in social or actionrelated explanations Vanrie and Verfaillie (2006) found

large differences in the bias for different actions Theyfound that the presence of a facing bias was neitherrelated to semantic effects nor to whether the actionwas familiar or not A gender bias was reported Brookset al (2008) but not confirmed (Schouten et al 2010)Manera et al (2012) found that the facing bias forwalking disappears when the observer walks on atreadmill but it was unclear whether this was due to theactor and observer performing the same action or tothe observerrsquos head movements

Our results point more to a low-level explanation Astrong bias was present for the position-scrambledstimulus that did not present biological motion Incontrast the facing bias was absent in the frame-scrambled stimulus although subjects did recognize thehuman form in these stimuli We see two implications ofthese data

First the facing bias was absent if the points did notmove The only condition without apparent motionframe-scrambled was the only one without a facingbias This finding was replicated in the main and thecontrol experiments This absence cannot be explainedby a bad recognition of the stimulus because it isknown that human observers interpret the frame-scrambled stimulus as human walking (Lange amp Lappe2007) Also in the main experiment the subjectsrsquoperformance was as good as for the rubberrsquo condition(Figure 5B)

Second the bias did not depend on whether a humanfigure was present The facing-the-observer bias waspresent in the position-scrambled stimulus although nohuman walking can be recognized from such ascrambled stimulus Moreover the largest bias wasfound for the rubber stimulus This stimulus isunnaturally deformed as the joint angles bend inimpossible directions and the limbs are nonrigid

We therefore argue that the facing-the-observer biasis not related to biological motion mechanisms Insteadone could speculate about different mechanisms Forexample the bias might be related to the bias forlooming patterns as opposed to compressing patterns(Ball amp Sekuler 1980 Fahle amp Wehrhahn 1991Georgeson amp Harris 1978) Such a biased system willrespond more with lsquolsquoloomingrsquorsquo than with lsquolsquocompressingrsquorsquoto a stimulus of back and forth oscillating dots Thusintegrating over the biased looming and compressingsignals such a stimulus will appear to be approachingthe observer regardless whether it is human or not(Albright 1989 Lappe amp Rauschecker 1995 Rau-schecker von Grunau amp Poulin 1987) This explana-tion would be consistent with the mixed results thathave been found for various actions Vanrie andVerfaillie (2006) found that presentations with littlelateral oscillations of the dots resulted in little or nofacing bias

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 8

Impression of depth

If the stimulus looks essentially two-dimensional theobserver will tend to underestimate how much thestimulus faces out of the plane of the display (up to acertain degree when the displayed facing direction isobviously close to frontal) As implied by the largeerror bars in Figure 5A some subjects generally had agood impression of depth resulting in a high accuracywhereas others saw little depth in the stimuli leading toa systematic negative mean error

The number of responses along the cardinal axes ismore suitable to compare the impression of depthbetween the six conditions Such errors were rare forthe three conditions with correct pendular motion (cfTable 1) but frequent for the other three conditionsThis suggests that pendular motion of the points withrespect to each other is an important cue to conveydepth This is in accordance with earlier suggestions(Hoffman amp Flinchbaugh 1982 Johansson 1973Kuhlmann et al 2009) The tendency to respond alongthe two cardinal axes is also known for the kineticdepth effect and is called the repulsion effect (Hiris ampBlake 1996)

Our results are consistent with a kinetic depth effectin combination with an oblique repulsion Neverthe-less biological motion is a special case in this respectbecause the kinetic depth is typically defined for rigidobjects biological motion is nonrigid The kineticdepth in this case can only function if the observer usesimplicit knowledge of the body structure and themotions that the human body is capable of

Cues to depth

The conditions of the main experiment weredesigned to systematically change the stimulus indifferent ways to remove some information whileleaving other information intact (cf Table 1)

The human form was intact in four conditionsnormal reversed marionette and frame-scrambledThe rubber conditions were physically impossiblebecause the knees and elbows became overextendedand the limb segments stretched and compressed duringthe movement Only the normal stimuli presenteddynamically correct human movements because undernormal gravity conditions the marionette movementscannot be performed and forward walking played inreverse temporal order differs from real backwardwalking (even if the marionette and reversed stimuli arekinematically and biomechanically valid) Intact pen-dular relative movements of pairs of point-lights werepresent in the normal reversed and the marionettestimulus Finally intact point trajectories were presentin the normal the rubber and the position-scrambledstimulus whereas the local motions are dynamically

incorrect for reversed stimuli (cf human movements)Comparisons of accuracy and precision betweendifferent conditions thus allow us to estimate the roleof the different depth cues

Accuracy differed widely between subjects asexpressed by the large error bars in Figure 5A Thewithin-subject ANOVA showed a clear difference onlyfor the position-scrambled stimulus This indicates thatpoint trajectories and local motions alone give onlypoor depth information

The precision of the judgments is shown in Figure5B In this respect the subjects were very similar butthere were large differences between the conditions

The precision was significantly better in the normaland reversed conditions than in all other conditionsThis suggests that the human form was an importantcue On the other hand neither the human movementnor the local motions were important because bothwere dynamically incorrect in the reversed condition(cf Table 1)

Comparing the marionette to the normal stimulusreveals a small but significant decrease in precision Themarionette is consistent with the human form at alltimes but dynamically incorrect with respect to humanmovement point trajectories and local motions Thisreiterates the importance of human form and alsosuggests that point trajectories might be importantHowever one should take into account that theconfiguration of the human form in the marionettestimulus is less familiar than in the normal or reversedconditions

Precision for the frame-scrambled condition wassignificantly better than for the position-scrambledcondition (Figure 5B) This is consistent with thefinding that the human form is an important cueHowever the frame-scrambled condition was signifi-cantly worse than the normal and reversed conditionsalthough all three conditions contain veridical humanform information The difference between the frame-scrambled condition and the normal and reversedconditions lies in pendular motion and point trajecto-ries which are both absent in the frame-scrambledcondition (cf Table 1) The rubber and marionettestimuli help to resolve this issue Marionette has correctpendular motion but incorrect point trajectoriesRubber has incorrect pendular motion but correctpoint trajectories Since the precision was better inmarionette than in rubber stimuli the pendular motionsmust be important for judging the facing in depth

This is consistent with a recent finding that thepresence of points for the feet improves the detection ofsmall angular deviations from a frontal view of walking(Cai Yang Chen amp Jiang 2011 Kuhlmann et al2009) Indeed the notion that human movements aretypically pendular has been suggested as an importantcue for recognizing biological motion long ago (Hoff-man amp Flinchbaugh 1982 Johansson 1973 Webb amp

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 9

Aggarwal 1982) In a slightly different manner this hasalso been proposed more recently by Neri (2009)

Lastly even the precision for the position-scrambledcondition was marginally better than the expectancyvalue for giving the same response on each trial (ie2628) Thus the rhythmic oscillatory movements ofthe individual points may have provided some infor-mation about the simulated facing in depth

Altogether our results support two major cues forthe judgment of facing in depth The first is the familiarhuman form The importance of the familiarity of bodyform for judging the orientation in depth of biologicalmotion is consistent with electrophysiological findingsin monkeys (Vangeneugden et al 2011) and psycho-physical findings in humans (Bulthoff et al 1998Jackson amp Blake 2010 Kuhlmann et al 2009Verfaillie amp de Graef 2000) The second is the pendularmovement of the limbs (Hoffman amp Flinchbaugh1982) The body is an articulated structure withinherent pendular motions Since these motions arenot random but highly characteristic these are apotential additional cue The data give support to theuse of pendular motion as an additional cue for judgingthe facing in depth

Supplementary material

There is a Quicktime movie for each of the sixconditions of the main experiment Each moviepresents without interruptions one walking cycle forthe facing in depth of 08 308 608 908 1208 1508 and1808

Acknowledgments

This study is funded by the German FederalMinistry of Education and Research BMBF [Grant01EC1003A] We thank Martin Habelbarth for his helpwith collecting the data We thank the anonymousreviewers for their constructive comments

Commercial relationships noneCorresponding author Marc H E de LussanetEmail lussanetwwudeAddress Psychologisches Institut Westf Wilhelms-Universitat Munster Munster Germany

References

Albright T D (1989) Centrifugal directionality bias inthe middle temporal visual area (MT) of the

macaque Visual Neuroscience 2 177ndash188[PubMed]

Ball K amp Sekuler R (1980) Human vision favorscentrifugal motion Perception 9(3) 317ndash325[PubMed]

Beintema J A amp Lappe M (2002) Perception ofbiological motion without local image motionProceedings of the National Academy of Sciencesof the USA 99 5661ndash5663 doi101073pnas082483699 [PubMed]

Blanz V Tarr M J amp Bulthoff H H (1999) Whatobject attributes determine canonical views Per-ception 28 575ndash599

Brooks A Schouten B Troje N F Verfaillie KBlanke O amp van der Zwan R (2008) Correlatedchanges in perceptions of the gender and orienta-tion of ambiguous biological motion figuresCurrent Biology 18(17) R728ndashR729 doi101016jcub200806054 [PubMed]

Bulthoff I Bulthoff H H amp Sinha P (1998) Top-down influences on stereoscopic depth-perceptionNature Neuroscience 1 254ndash257 [PubMed]

Cai P Yang X Chen L amp Jiang Y (2011) Motionspeed modulates walking direction discriminationThe role of the feet in biological motion perceptionChinese Science Bulletin 56(19) 2025ndash2030 doi101007s11434-011-4528-6

Chang D H F amp Troje N F (2009) Accelerationcarries the local inversion effect in biologicalmotion perception Journal of Vision 9(1)19 1ndash17 httpjournalofvisionorg9119 doi1011679119 [PubMed] [Article]

Fahle M amp Wehrhahn C (1991) Motion perceptionin the peripheral visual field Graefersquos Archive forClinical and Experimental Ophthalmology 229(5)430ndash436 doi101007BF00166305 [PubMed]

Foster D H amp Gilson S J (2002) Recognizing novelthree-dimensional objects by summing signals fromparts and views Proceedings of the Royal Society BBiological Sciences 269 1939ndash1947 doi101098rspb20022119

Georgeson M A amp Harris M G (1978) Apparentfoveofugal drift of counterphase gratings Percep-tion 7 527ndash536 [PubMed]

Giese M A (2004) Neural model for biologicalmovement recognition A neurophysiologicallyplausible theory In L M Vaina S A Beardsleyamp S Rushton (Eds) Optic flow and beyond (pp443ndash470) Dordrecht NL Kluwer Academic Pub-lishers

Grill-Spector K Kushnir T Edelman S AvidanG Itzchak Y amp Malach R (1999) Differential

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 10

responses are located on the ascending diagonalwhereas the opposite facing direction is depicted bythe descending diagonal The latter kind of errors canoccur if the front and backside are confused Sucherrors were rare and most common in the reversed andmarionette conditions

A different kind of error is the tendency to respondalong the cardinal axes (08 908 1808 and 2708) Theseresponses are clustered around the horizontal 08 908and 1808 lines These errors were most common for therubber and frame-scrambled conditions with a prob-ability of a response within 38 of a cardinal axis ofabout five times the expectancy rate of a random

distribution For the normal reversed and marionette

stimuli this probability was about 25 times the

expected rate

In most conditions the moving average curves show

a slight S-shape In the range of 08ndash908 the curves are

slightly below the diagonal whereas for 908ndash1808 the

curves are somewhat higher This indicates that the

observers tended to underestimate how much the facing

in depth deviated from the plane of the display Due to

the S-shape the errors tend to be negative (cf Figure

2D) In other words the distributions indicate that the

precision tends to be negative

Figure 4 Distribution of the responses in each condition as a function of the simulated facing in depth (a) Each subject is represented by a

different symbol Note that responses between 1808ndash3608 are converted to 1808ndash08 (cf Figure 2D and the inset in the upper left panel) 08

facing right Red curves LOWESS moving-average fit with a window of 108 Ascending diagonal correct responses descending diagonal

wrong facing direction

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 6

Accuracy and precision (mean and variableerrors)

Figure 5 presents the two error measures Note thatthe two panels have the same scale The error bars aremuch larger for the mean errors (the accuracy Panel A)than for the variable errors (the precision Panel B)This reflects that the S-shape in the distributions ofFigure 4 (see above) was strongly expressed for somesubjects but not for all

The S-shape in the response distributions resulted ina negative mean error of68 for the normal stimuli (theaccuracy Figure 5A) The ANOVA on the mean errorsshowed a significant main effect of scrambling typeF(5 48) frac14 106 p 00001 Fisherrsquos PLSD post-hoctests revealed that the differences between position-scrambled and all other conditions as well as betweenframe-scrambled and the rubber and marionetteconditions were statistically significant (p 001)The normal condition differed only significantly fromthe position-scrambled condition T tests showed thatthe mean errors differed significantly from zero in thenormal reversed frame-scrambled and position-scram-bled conditions (p 01 corrected) and did not in therubber condition (p 04)

The ANOVA on the variable errors (the precisionFigure 5B) also showed a significant main effect ofscrambling type F(5 48) frac14 440 p 00001 FisherrsquosPLSD post-hoc tests showed that the only conditionsthat did not differ significantly were normal andreversed and rubber and frame-scrambled (p 05)All other differences were highly significant (p 001)All conditions differed significantly from zero (p 001 corrected t tests) As explained in the Methodsthe optimal guessing strategy ie giving the sameresponse on every trial would result in a precision of2628 The precision was significantly lower than 2628

in all conditions (p 005 corrected) except theposition-scrambled stimuli (pfrac14 006)

The mean and variable errors of subjects were notcorrelated for any of the six conditions (r 026 p 05)

Control experiment Frame duration isirrelevant for frame scrambling

The frame-scrambled condition of main experimentwas presented with a 10-ms frame duration which veryshort and may have deteriorated the performance ofparticipants The control experiment included only theframe-scrambled conditions but with a longer frameduration of 80 or 160 ms With longer frame durationsthe subsequent configurations might have been recog-nized better and thus improved performance

A new group of 11 undergraduate students per-formed in the control experiment Each still configura-tion of the frame-scrambled stimulus remained on thescreen for either eight or 16 monitor refreshes (ie 80or 160 ms)

The repeated-measures ANOVAs on the meanerrors on the standard deviation of the error and onthe percentage of toward-responses did not reveal anysignificant main effects of frame duration F(1 20) 3p 01 These measures were all very similar to thosefound for the scrambled frame-order condition in themain experiment (mean error frac14 898 standarddeviation error frac14 1708 towards responses frac14 47)Three new ANOVAs with experiment as a between-factor confirmed this There was no significant maineffect of experiment for any of the three measures F(128) 10 p 03

The results of the control experiment confirm thatthe frame duration was not a limiting factor for the

Figure 5 (A) Mean error (accuracy) and (B) the standard deviation (precision) of the error NS the only nonsignificant differences (see

text) Error bars standard error

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 7

recognition of the facing direction in the mainexperiment The results also reproduce the lack of afacing bias in the frame-scrambled condition

Discussion

The results of the present study show that non-trained human observers can accurately judge thefacing in depth from two-dimensional point-lightbiological motion displays As we will discuss belowboth the familiar structure of the body and the relativependular motions of the joints contributed to this

The results revealed four different factors to thejudgment of the facing in depth of a point-light displayThese are (a) the discrimination of the front and backside (b) the facing bias (c) the impression of depthfrom the display and (d) the precision of the judgedfacing direction

Confusion of front and back

The participants were generally very good atdiscriminating the front and back sidse (cf Figure 4in all conditions but the position-scrambled there werefar more responses around the ascending dashed linesthan around the descending dashed lines) This is in linewith earlier findings in which the task was only todiscriminate between facing left and right (Beintema ampLappe 2002 for normal walking Lange amp Lappe2007 for frame-scrambled) The position-scrambledcondition showed a substantial number of confusionsof the front side and back side consistent with thefinding of Troje and Westhoff (2006) The present studyextends these findings to facing in depth orientations

Facing bias

We found a strong facing-toward-the-observer biasin most conditions This finding corroborates earlierfindings (Brooks et al 2008 Manera et al 2012Vanrie et al 2004 Vanrie amp Verfaillie 2006) Thesestudies all investigated the depth ambiguity inherent tobiological motion without explicit depth cues In somecases such as walking this ambiguity leads to a strongfacing bias (Vanrie et al 2004 Vanrie amp Verfaillie2006) More generally this kind of bias is known asdepth inversion (Yellott amp Kaiwi 1979) A well knownexample from research on face perception is the hollowface illusion (Hill amp Johnston 2007)

Many biological motion studies on the facing biashave sought the reason for this bias in social or actionrelated explanations Vanrie and Verfaillie (2006) found

large differences in the bias for different actions Theyfound that the presence of a facing bias was neitherrelated to semantic effects nor to whether the actionwas familiar or not A gender bias was reported Brookset al (2008) but not confirmed (Schouten et al 2010)Manera et al (2012) found that the facing bias forwalking disappears when the observer walks on atreadmill but it was unclear whether this was due to theactor and observer performing the same action or tothe observerrsquos head movements

Our results point more to a low-level explanation Astrong bias was present for the position-scrambledstimulus that did not present biological motion Incontrast the facing bias was absent in the frame-scrambled stimulus although subjects did recognize thehuman form in these stimuli We see two implications ofthese data

First the facing bias was absent if the points did notmove The only condition without apparent motionframe-scrambled was the only one without a facingbias This finding was replicated in the main and thecontrol experiments This absence cannot be explainedby a bad recognition of the stimulus because it isknown that human observers interpret the frame-scrambled stimulus as human walking (Lange amp Lappe2007) Also in the main experiment the subjectsrsquoperformance was as good as for the rubberrsquo condition(Figure 5B)

Second the bias did not depend on whether a humanfigure was present The facing-the-observer bias waspresent in the position-scrambled stimulus although nohuman walking can be recognized from such ascrambled stimulus Moreover the largest bias wasfound for the rubber stimulus This stimulus isunnaturally deformed as the joint angles bend inimpossible directions and the limbs are nonrigid

We therefore argue that the facing-the-observer biasis not related to biological motion mechanisms Insteadone could speculate about different mechanisms Forexample the bias might be related to the bias forlooming patterns as opposed to compressing patterns(Ball amp Sekuler 1980 Fahle amp Wehrhahn 1991Georgeson amp Harris 1978) Such a biased system willrespond more with lsquolsquoloomingrsquorsquo than with lsquolsquocompressingrsquorsquoto a stimulus of back and forth oscillating dots Thusintegrating over the biased looming and compressingsignals such a stimulus will appear to be approachingthe observer regardless whether it is human or not(Albright 1989 Lappe amp Rauschecker 1995 Rau-schecker von Grunau amp Poulin 1987) This explana-tion would be consistent with the mixed results thathave been found for various actions Vanrie andVerfaillie (2006) found that presentations with littlelateral oscillations of the dots resulted in little or nofacing bias

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 8

Impression of depth

If the stimulus looks essentially two-dimensional theobserver will tend to underestimate how much thestimulus faces out of the plane of the display (up to acertain degree when the displayed facing direction isobviously close to frontal) As implied by the largeerror bars in Figure 5A some subjects generally had agood impression of depth resulting in a high accuracywhereas others saw little depth in the stimuli leading toa systematic negative mean error

The number of responses along the cardinal axes ismore suitable to compare the impression of depthbetween the six conditions Such errors were rare forthe three conditions with correct pendular motion (cfTable 1) but frequent for the other three conditionsThis suggests that pendular motion of the points withrespect to each other is an important cue to conveydepth This is in accordance with earlier suggestions(Hoffman amp Flinchbaugh 1982 Johansson 1973Kuhlmann et al 2009) The tendency to respond alongthe two cardinal axes is also known for the kineticdepth effect and is called the repulsion effect (Hiris ampBlake 1996)

Our results are consistent with a kinetic depth effectin combination with an oblique repulsion Neverthe-less biological motion is a special case in this respectbecause the kinetic depth is typically defined for rigidobjects biological motion is nonrigid The kineticdepth in this case can only function if the observer usesimplicit knowledge of the body structure and themotions that the human body is capable of

Cues to depth

The conditions of the main experiment weredesigned to systematically change the stimulus indifferent ways to remove some information whileleaving other information intact (cf Table 1)

The human form was intact in four conditionsnormal reversed marionette and frame-scrambledThe rubber conditions were physically impossiblebecause the knees and elbows became overextendedand the limb segments stretched and compressed duringthe movement Only the normal stimuli presenteddynamically correct human movements because undernormal gravity conditions the marionette movementscannot be performed and forward walking played inreverse temporal order differs from real backwardwalking (even if the marionette and reversed stimuli arekinematically and biomechanically valid) Intact pen-dular relative movements of pairs of point-lights werepresent in the normal reversed and the marionettestimulus Finally intact point trajectories were presentin the normal the rubber and the position-scrambledstimulus whereas the local motions are dynamically

incorrect for reversed stimuli (cf human movements)Comparisons of accuracy and precision betweendifferent conditions thus allow us to estimate the roleof the different depth cues

Accuracy differed widely between subjects asexpressed by the large error bars in Figure 5A Thewithin-subject ANOVA showed a clear difference onlyfor the position-scrambled stimulus This indicates thatpoint trajectories and local motions alone give onlypoor depth information

The precision of the judgments is shown in Figure5B In this respect the subjects were very similar butthere were large differences between the conditions

The precision was significantly better in the normaland reversed conditions than in all other conditionsThis suggests that the human form was an importantcue On the other hand neither the human movementnor the local motions were important because bothwere dynamically incorrect in the reversed condition(cf Table 1)

Comparing the marionette to the normal stimulusreveals a small but significant decrease in precision Themarionette is consistent with the human form at alltimes but dynamically incorrect with respect to humanmovement point trajectories and local motions Thisreiterates the importance of human form and alsosuggests that point trajectories might be importantHowever one should take into account that theconfiguration of the human form in the marionettestimulus is less familiar than in the normal or reversedconditions

Precision for the frame-scrambled condition wassignificantly better than for the position-scrambledcondition (Figure 5B) This is consistent with thefinding that the human form is an important cueHowever the frame-scrambled condition was signifi-cantly worse than the normal and reversed conditionsalthough all three conditions contain veridical humanform information The difference between the frame-scrambled condition and the normal and reversedconditions lies in pendular motion and point trajecto-ries which are both absent in the frame-scrambledcondition (cf Table 1) The rubber and marionettestimuli help to resolve this issue Marionette has correctpendular motion but incorrect point trajectoriesRubber has incorrect pendular motion but correctpoint trajectories Since the precision was better inmarionette than in rubber stimuli the pendular motionsmust be important for judging the facing in depth

This is consistent with a recent finding that thepresence of points for the feet improves the detection ofsmall angular deviations from a frontal view of walking(Cai Yang Chen amp Jiang 2011 Kuhlmann et al2009) Indeed the notion that human movements aretypically pendular has been suggested as an importantcue for recognizing biological motion long ago (Hoff-man amp Flinchbaugh 1982 Johansson 1973 Webb amp

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 9

Aggarwal 1982) In a slightly different manner this hasalso been proposed more recently by Neri (2009)

Lastly even the precision for the position-scrambledcondition was marginally better than the expectancyvalue for giving the same response on each trial (ie2628) Thus the rhythmic oscillatory movements ofthe individual points may have provided some infor-mation about the simulated facing in depth

Altogether our results support two major cues forthe judgment of facing in depth The first is the familiarhuman form The importance of the familiarity of bodyform for judging the orientation in depth of biologicalmotion is consistent with electrophysiological findingsin monkeys (Vangeneugden et al 2011) and psycho-physical findings in humans (Bulthoff et al 1998Jackson amp Blake 2010 Kuhlmann et al 2009Verfaillie amp de Graef 2000) The second is the pendularmovement of the limbs (Hoffman amp Flinchbaugh1982) The body is an articulated structure withinherent pendular motions Since these motions arenot random but highly characteristic these are apotential additional cue The data give support to theuse of pendular motion as an additional cue for judgingthe facing in depth

Supplementary material

There is a Quicktime movie for each of the sixconditions of the main experiment Each moviepresents without interruptions one walking cycle forthe facing in depth of 08 308 608 908 1208 1508 and1808

Acknowledgments

This study is funded by the German FederalMinistry of Education and Research BMBF [Grant01EC1003A] We thank Martin Habelbarth for his helpwith collecting the data We thank the anonymousreviewers for their constructive comments

Commercial relationships noneCorresponding author Marc H E de LussanetEmail lussanetwwudeAddress Psychologisches Institut Westf Wilhelms-Universitat Munster Munster Germany

References

Albright T D (1989) Centrifugal directionality bias inthe middle temporal visual area (MT) of the

macaque Visual Neuroscience 2 177ndash188[PubMed]

Ball K amp Sekuler R (1980) Human vision favorscentrifugal motion Perception 9(3) 317ndash325[PubMed]

Beintema J A amp Lappe M (2002) Perception ofbiological motion without local image motionProceedings of the National Academy of Sciencesof the USA 99 5661ndash5663 doi101073pnas082483699 [PubMed]

Blanz V Tarr M J amp Bulthoff H H (1999) Whatobject attributes determine canonical views Per-ception 28 575ndash599

Brooks A Schouten B Troje N F Verfaillie KBlanke O amp van der Zwan R (2008) Correlatedchanges in perceptions of the gender and orienta-tion of ambiguous biological motion figuresCurrent Biology 18(17) R728ndashR729 doi101016jcub200806054 [PubMed]

Bulthoff I Bulthoff H H amp Sinha P (1998) Top-down influences on stereoscopic depth-perceptionNature Neuroscience 1 254ndash257 [PubMed]

Cai P Yang X Chen L amp Jiang Y (2011) Motionspeed modulates walking direction discriminationThe role of the feet in biological motion perceptionChinese Science Bulletin 56(19) 2025ndash2030 doi101007s11434-011-4528-6

Chang D H F amp Troje N F (2009) Accelerationcarries the local inversion effect in biologicalmotion perception Journal of Vision 9(1)19 1ndash17 httpjournalofvisionorg9119 doi1011679119 [PubMed] [Article]

Fahle M amp Wehrhahn C (1991) Motion perceptionin the peripheral visual field Graefersquos Archive forClinical and Experimental Ophthalmology 229(5)430ndash436 doi101007BF00166305 [PubMed]

Foster D H amp Gilson S J (2002) Recognizing novelthree-dimensional objects by summing signals fromparts and views Proceedings of the Royal Society BBiological Sciences 269 1939ndash1947 doi101098rspb20022119

Georgeson M A amp Harris M G (1978) Apparentfoveofugal drift of counterphase gratings Percep-tion 7 527ndash536 [PubMed]

Giese M A (2004) Neural model for biologicalmovement recognition A neurophysiologicallyplausible theory In L M Vaina S A Beardsleyamp S Rushton (Eds) Optic flow and beyond (pp443ndash470) Dordrecht NL Kluwer Academic Pub-lishers

Grill-Spector K Kushnir T Edelman S AvidanG Itzchak Y amp Malach R (1999) Differential

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 10

Accuracy and precision (mean and variableerrors)

Figure 5 presents the two error measures Note thatthe two panels have the same scale The error bars aremuch larger for the mean errors (the accuracy Panel A)than for the variable errors (the precision Panel B)This reflects that the S-shape in the distributions ofFigure 4 (see above) was strongly expressed for somesubjects but not for all

The S-shape in the response distributions resulted ina negative mean error of68 for the normal stimuli (theaccuracy Figure 5A) The ANOVA on the mean errorsshowed a significant main effect of scrambling typeF(5 48) frac14 106 p 00001 Fisherrsquos PLSD post-hoctests revealed that the differences between position-scrambled and all other conditions as well as betweenframe-scrambled and the rubber and marionetteconditions were statistically significant (p 001)The normal condition differed only significantly fromthe position-scrambled condition T tests showed thatthe mean errors differed significantly from zero in thenormal reversed frame-scrambled and position-scram-bled conditions (p 01 corrected) and did not in therubber condition (p 04)

The ANOVA on the variable errors (the precisionFigure 5B) also showed a significant main effect ofscrambling type F(5 48) frac14 440 p 00001 FisherrsquosPLSD post-hoc tests showed that the only conditionsthat did not differ significantly were normal andreversed and rubber and frame-scrambled (p 05)All other differences were highly significant (p 001)All conditions differed significantly from zero (p 001 corrected t tests) As explained in the Methodsthe optimal guessing strategy ie giving the sameresponse on every trial would result in a precision of2628 The precision was significantly lower than 2628

in all conditions (p 005 corrected) except theposition-scrambled stimuli (pfrac14 006)

The mean and variable errors of subjects were notcorrelated for any of the six conditions (r 026 p 05)

Control experiment Frame duration isirrelevant for frame scrambling

The frame-scrambled condition of main experimentwas presented with a 10-ms frame duration which veryshort and may have deteriorated the performance ofparticipants The control experiment included only theframe-scrambled conditions but with a longer frameduration of 80 or 160 ms With longer frame durationsthe subsequent configurations might have been recog-nized better and thus improved performance

A new group of 11 undergraduate students per-formed in the control experiment Each still configura-tion of the frame-scrambled stimulus remained on thescreen for either eight or 16 monitor refreshes (ie 80or 160 ms)

The repeated-measures ANOVAs on the meanerrors on the standard deviation of the error and onthe percentage of toward-responses did not reveal anysignificant main effects of frame duration F(1 20) 3p 01 These measures were all very similar to thosefound for the scrambled frame-order condition in themain experiment (mean error frac14 898 standarddeviation error frac14 1708 towards responses frac14 47)Three new ANOVAs with experiment as a between-factor confirmed this There was no significant maineffect of experiment for any of the three measures F(128) 10 p 03

The results of the control experiment confirm thatthe frame duration was not a limiting factor for the

Figure 5 (A) Mean error (accuracy) and (B) the standard deviation (precision) of the error NS the only nonsignificant differences (see

text) Error bars standard error

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 7

recognition of the facing direction in the mainexperiment The results also reproduce the lack of afacing bias in the frame-scrambled condition

Discussion

The results of the present study show that non-trained human observers can accurately judge thefacing in depth from two-dimensional point-lightbiological motion displays As we will discuss belowboth the familiar structure of the body and the relativependular motions of the joints contributed to this

The results revealed four different factors to thejudgment of the facing in depth of a point-light displayThese are (a) the discrimination of the front and backside (b) the facing bias (c) the impression of depthfrom the display and (d) the precision of the judgedfacing direction

Confusion of front and back

The participants were generally very good atdiscriminating the front and back sidse (cf Figure 4in all conditions but the position-scrambled there werefar more responses around the ascending dashed linesthan around the descending dashed lines) This is in linewith earlier findings in which the task was only todiscriminate between facing left and right (Beintema ampLappe 2002 for normal walking Lange amp Lappe2007 for frame-scrambled) The position-scrambledcondition showed a substantial number of confusionsof the front side and back side consistent with thefinding of Troje and Westhoff (2006) The present studyextends these findings to facing in depth orientations

Facing bias

We found a strong facing-toward-the-observer biasin most conditions This finding corroborates earlierfindings (Brooks et al 2008 Manera et al 2012Vanrie et al 2004 Vanrie amp Verfaillie 2006) Thesestudies all investigated the depth ambiguity inherent tobiological motion without explicit depth cues In somecases such as walking this ambiguity leads to a strongfacing bias (Vanrie et al 2004 Vanrie amp Verfaillie2006) More generally this kind of bias is known asdepth inversion (Yellott amp Kaiwi 1979) A well knownexample from research on face perception is the hollowface illusion (Hill amp Johnston 2007)

Many biological motion studies on the facing biashave sought the reason for this bias in social or actionrelated explanations Vanrie and Verfaillie (2006) found

large differences in the bias for different actions Theyfound that the presence of a facing bias was neitherrelated to semantic effects nor to whether the actionwas familiar or not A gender bias was reported Brookset al (2008) but not confirmed (Schouten et al 2010)Manera et al (2012) found that the facing bias forwalking disappears when the observer walks on atreadmill but it was unclear whether this was due to theactor and observer performing the same action or tothe observerrsquos head movements

Our results point more to a low-level explanation Astrong bias was present for the position-scrambledstimulus that did not present biological motion Incontrast the facing bias was absent in the frame-scrambled stimulus although subjects did recognize thehuman form in these stimuli We see two implications ofthese data

First the facing bias was absent if the points did notmove The only condition without apparent motionframe-scrambled was the only one without a facingbias This finding was replicated in the main and thecontrol experiments This absence cannot be explainedby a bad recognition of the stimulus because it isknown that human observers interpret the frame-scrambled stimulus as human walking (Lange amp Lappe2007) Also in the main experiment the subjectsrsquoperformance was as good as for the rubberrsquo condition(Figure 5B)

Second the bias did not depend on whether a humanfigure was present The facing-the-observer bias waspresent in the position-scrambled stimulus although nohuman walking can be recognized from such ascrambled stimulus Moreover the largest bias wasfound for the rubber stimulus This stimulus isunnaturally deformed as the joint angles bend inimpossible directions and the limbs are nonrigid

We therefore argue that the facing-the-observer biasis not related to biological motion mechanisms Insteadone could speculate about different mechanisms Forexample the bias might be related to the bias forlooming patterns as opposed to compressing patterns(Ball amp Sekuler 1980 Fahle amp Wehrhahn 1991Georgeson amp Harris 1978) Such a biased system willrespond more with lsquolsquoloomingrsquorsquo than with lsquolsquocompressingrsquorsquoto a stimulus of back and forth oscillating dots Thusintegrating over the biased looming and compressingsignals such a stimulus will appear to be approachingthe observer regardless whether it is human or not(Albright 1989 Lappe amp Rauschecker 1995 Rau-schecker von Grunau amp Poulin 1987) This explana-tion would be consistent with the mixed results thathave been found for various actions Vanrie andVerfaillie (2006) found that presentations with littlelateral oscillations of the dots resulted in little or nofacing bias

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 8

Impression of depth

If the stimulus looks essentially two-dimensional theobserver will tend to underestimate how much thestimulus faces out of the plane of the display (up to acertain degree when the displayed facing direction isobviously close to frontal) As implied by the largeerror bars in Figure 5A some subjects generally had agood impression of depth resulting in a high accuracywhereas others saw little depth in the stimuli leading toa systematic negative mean error

The number of responses along the cardinal axes ismore suitable to compare the impression of depthbetween the six conditions Such errors were rare forthe three conditions with correct pendular motion (cfTable 1) but frequent for the other three conditionsThis suggests that pendular motion of the points withrespect to each other is an important cue to conveydepth This is in accordance with earlier suggestions(Hoffman amp Flinchbaugh 1982 Johansson 1973Kuhlmann et al 2009) The tendency to respond alongthe two cardinal axes is also known for the kineticdepth effect and is called the repulsion effect (Hiris ampBlake 1996)

Our results are consistent with a kinetic depth effectin combination with an oblique repulsion Neverthe-less biological motion is a special case in this respectbecause the kinetic depth is typically defined for rigidobjects biological motion is nonrigid The kineticdepth in this case can only function if the observer usesimplicit knowledge of the body structure and themotions that the human body is capable of

Cues to depth

The conditions of the main experiment weredesigned to systematically change the stimulus indifferent ways to remove some information whileleaving other information intact (cf Table 1)

The human form was intact in four conditionsnormal reversed marionette and frame-scrambledThe rubber conditions were physically impossiblebecause the knees and elbows became overextendedand the limb segments stretched and compressed duringthe movement Only the normal stimuli presenteddynamically correct human movements because undernormal gravity conditions the marionette movementscannot be performed and forward walking played inreverse temporal order differs from real backwardwalking (even if the marionette and reversed stimuli arekinematically and biomechanically valid) Intact pen-dular relative movements of pairs of point-lights werepresent in the normal reversed and the marionettestimulus Finally intact point trajectories were presentin the normal the rubber and the position-scrambledstimulus whereas the local motions are dynamically

incorrect for reversed stimuli (cf human movements)Comparisons of accuracy and precision betweendifferent conditions thus allow us to estimate the roleof the different depth cues

Accuracy differed widely between subjects asexpressed by the large error bars in Figure 5A Thewithin-subject ANOVA showed a clear difference onlyfor the position-scrambled stimulus This indicates thatpoint trajectories and local motions alone give onlypoor depth information

The precision of the judgments is shown in Figure5B In this respect the subjects were very similar butthere were large differences between the conditions

The precision was significantly better in the normaland reversed conditions than in all other conditionsThis suggests that the human form was an importantcue On the other hand neither the human movementnor the local motions were important because bothwere dynamically incorrect in the reversed condition(cf Table 1)

Comparing the marionette to the normal stimulusreveals a small but significant decrease in precision Themarionette is consistent with the human form at alltimes but dynamically incorrect with respect to humanmovement point trajectories and local motions Thisreiterates the importance of human form and alsosuggests that point trajectories might be importantHowever one should take into account that theconfiguration of the human form in the marionettestimulus is less familiar than in the normal or reversedconditions

Precision for the frame-scrambled condition wassignificantly better than for the position-scrambledcondition (Figure 5B) This is consistent with thefinding that the human form is an important cueHowever the frame-scrambled condition was signifi-cantly worse than the normal and reversed conditionsalthough all three conditions contain veridical humanform information The difference between the frame-scrambled condition and the normal and reversedconditions lies in pendular motion and point trajecto-ries which are both absent in the frame-scrambledcondition (cf Table 1) The rubber and marionettestimuli help to resolve this issue Marionette has correctpendular motion but incorrect point trajectoriesRubber has incorrect pendular motion but correctpoint trajectories Since the precision was better inmarionette than in rubber stimuli the pendular motionsmust be important for judging the facing in depth

This is consistent with a recent finding that thepresence of points for the feet improves the detection ofsmall angular deviations from a frontal view of walking(Cai Yang Chen amp Jiang 2011 Kuhlmann et al2009) Indeed the notion that human movements aretypically pendular has been suggested as an importantcue for recognizing biological motion long ago (Hoff-man amp Flinchbaugh 1982 Johansson 1973 Webb amp

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 9

Aggarwal 1982) In a slightly different manner this hasalso been proposed more recently by Neri (2009)

Lastly even the precision for the position-scrambledcondition was marginally better than the expectancyvalue for giving the same response on each trial (ie2628) Thus the rhythmic oscillatory movements ofthe individual points may have provided some infor-mation about the simulated facing in depth

Altogether our results support two major cues forthe judgment of facing in depth The first is the familiarhuman form The importance of the familiarity of bodyform for judging the orientation in depth of biologicalmotion is consistent with electrophysiological findingsin monkeys (Vangeneugden et al 2011) and psycho-physical findings in humans (Bulthoff et al 1998Jackson amp Blake 2010 Kuhlmann et al 2009Verfaillie amp de Graef 2000) The second is the pendularmovement of the limbs (Hoffman amp Flinchbaugh1982) The body is an articulated structure withinherent pendular motions Since these motions arenot random but highly characteristic these are apotential additional cue The data give support to theuse of pendular motion as an additional cue for judgingthe facing in depth

Supplementary material

There is a Quicktime movie for each of the sixconditions of the main experiment Each moviepresents without interruptions one walking cycle forthe facing in depth of 08 308 608 908 1208 1508 and1808

Acknowledgments

This study is funded by the German FederalMinistry of Education and Research BMBF [Grant01EC1003A] We thank Martin Habelbarth for his helpwith collecting the data We thank the anonymousreviewers for their constructive comments

Commercial relationships noneCorresponding author Marc H E de LussanetEmail lussanetwwudeAddress Psychologisches Institut Westf Wilhelms-Universitat Munster Munster Germany

References

Albright T D (1989) Centrifugal directionality bias inthe middle temporal visual area (MT) of the

macaque Visual Neuroscience 2 177ndash188[PubMed]

Ball K amp Sekuler R (1980) Human vision favorscentrifugal motion Perception 9(3) 317ndash325[PubMed]

Beintema J A amp Lappe M (2002) Perception ofbiological motion without local image motionProceedings of the National Academy of Sciencesof the USA 99 5661ndash5663 doi101073pnas082483699 [PubMed]

Blanz V Tarr M J amp Bulthoff H H (1999) Whatobject attributes determine canonical views Per-ception 28 575ndash599

Brooks A Schouten B Troje N F Verfaillie KBlanke O amp van der Zwan R (2008) Correlatedchanges in perceptions of the gender and orienta-tion of ambiguous biological motion figuresCurrent Biology 18(17) R728ndashR729 doi101016jcub200806054 [PubMed]

Bulthoff I Bulthoff H H amp Sinha P (1998) Top-down influences on stereoscopic depth-perceptionNature Neuroscience 1 254ndash257 [PubMed]

Cai P Yang X Chen L amp Jiang Y (2011) Motionspeed modulates walking direction discriminationThe role of the feet in biological motion perceptionChinese Science Bulletin 56(19) 2025ndash2030 doi101007s11434-011-4528-6

Chang D H F amp Troje N F (2009) Accelerationcarries the local inversion effect in biologicalmotion perception Journal of Vision 9(1)19 1ndash17 httpjournalofvisionorg9119 doi1011679119 [PubMed] [Article]

Fahle M amp Wehrhahn C (1991) Motion perceptionin the peripheral visual field Graefersquos Archive forClinical and Experimental Ophthalmology 229(5)430ndash436 doi101007BF00166305 [PubMed]

Foster D H amp Gilson S J (2002) Recognizing novelthree-dimensional objects by summing signals fromparts and views Proceedings of the Royal Society BBiological Sciences 269 1939ndash1947 doi101098rspb20022119

Georgeson M A amp Harris M G (1978) Apparentfoveofugal drift of counterphase gratings Percep-tion 7 527ndash536 [PubMed]

Giese M A (2004) Neural model for biologicalmovement recognition A neurophysiologicallyplausible theory In L M Vaina S A Beardsleyamp S Rushton (Eds) Optic flow and beyond (pp443ndash470) Dordrecht NL Kluwer Academic Pub-lishers

Grill-Spector K Kushnir T Edelman S AvidanG Itzchak Y amp Malach R (1999) Differential

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 10

recognition of the facing direction in the mainexperiment The results also reproduce the lack of afacing bias in the frame-scrambled condition

Discussion

The results of the present study show that non-trained human observers can accurately judge thefacing in depth from two-dimensional point-lightbiological motion displays As we will discuss belowboth the familiar structure of the body and the relativependular motions of the joints contributed to this

The results revealed four different factors to thejudgment of the facing in depth of a point-light displayThese are (a) the discrimination of the front and backside (b) the facing bias (c) the impression of depthfrom the display and (d) the precision of the judgedfacing direction

Confusion of front and back

The participants were generally very good atdiscriminating the front and back sidse (cf Figure 4in all conditions but the position-scrambled there werefar more responses around the ascending dashed linesthan around the descending dashed lines) This is in linewith earlier findings in which the task was only todiscriminate between facing left and right (Beintema ampLappe 2002 for normal walking Lange amp Lappe2007 for frame-scrambled) The position-scrambledcondition showed a substantial number of confusionsof the front side and back side consistent with thefinding of Troje and Westhoff (2006) The present studyextends these findings to facing in depth orientations

Facing bias

We found a strong facing-toward-the-observer biasin most conditions This finding corroborates earlierfindings (Brooks et al 2008 Manera et al 2012Vanrie et al 2004 Vanrie amp Verfaillie 2006) Thesestudies all investigated the depth ambiguity inherent tobiological motion without explicit depth cues In somecases such as walking this ambiguity leads to a strongfacing bias (Vanrie et al 2004 Vanrie amp Verfaillie2006) More generally this kind of bias is known asdepth inversion (Yellott amp Kaiwi 1979) A well knownexample from research on face perception is the hollowface illusion (Hill amp Johnston 2007)

Many biological motion studies on the facing biashave sought the reason for this bias in social or actionrelated explanations Vanrie and Verfaillie (2006) found

large differences in the bias for different actions Theyfound that the presence of a facing bias was neitherrelated to semantic effects nor to whether the actionwas familiar or not A gender bias was reported Brookset al (2008) but not confirmed (Schouten et al 2010)Manera et al (2012) found that the facing bias forwalking disappears when the observer walks on atreadmill but it was unclear whether this was due to theactor and observer performing the same action or tothe observerrsquos head movements

Our results point more to a low-level explanation Astrong bias was present for the position-scrambledstimulus that did not present biological motion Incontrast the facing bias was absent in the frame-scrambled stimulus although subjects did recognize thehuman form in these stimuli We see two implications ofthese data

First the facing bias was absent if the points did notmove The only condition without apparent motionframe-scrambled was the only one without a facingbias This finding was replicated in the main and thecontrol experiments This absence cannot be explainedby a bad recognition of the stimulus because it isknown that human observers interpret the frame-scrambled stimulus as human walking (Lange amp Lappe2007) Also in the main experiment the subjectsrsquoperformance was as good as for the rubberrsquo condition(Figure 5B)

Second the bias did not depend on whether a humanfigure was present The facing-the-observer bias waspresent in the position-scrambled stimulus although nohuman walking can be recognized from such ascrambled stimulus Moreover the largest bias wasfound for the rubber stimulus This stimulus isunnaturally deformed as the joint angles bend inimpossible directions and the limbs are nonrigid

We therefore argue that the facing-the-observer biasis not related to biological motion mechanisms Insteadone could speculate about different mechanisms Forexample the bias might be related to the bias forlooming patterns as opposed to compressing patterns(Ball amp Sekuler 1980 Fahle amp Wehrhahn 1991Georgeson amp Harris 1978) Such a biased system willrespond more with lsquolsquoloomingrsquorsquo than with lsquolsquocompressingrsquorsquoto a stimulus of back and forth oscillating dots Thusintegrating over the biased looming and compressingsignals such a stimulus will appear to be approachingthe observer regardless whether it is human or not(Albright 1989 Lappe amp Rauschecker 1995 Rau-schecker von Grunau amp Poulin 1987) This explana-tion would be consistent with the mixed results thathave been found for various actions Vanrie andVerfaillie (2006) found that presentations with littlelateral oscillations of the dots resulted in little or nofacing bias

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 8

Impression of depth

If the stimulus looks essentially two-dimensional theobserver will tend to underestimate how much thestimulus faces out of the plane of the display (up to acertain degree when the displayed facing direction isobviously close to frontal) As implied by the largeerror bars in Figure 5A some subjects generally had agood impression of depth resulting in a high accuracywhereas others saw little depth in the stimuli leading toa systematic negative mean error

The number of responses along the cardinal axes ismore suitable to compare the impression of depthbetween the six conditions Such errors were rare forthe three conditions with correct pendular motion (cfTable 1) but frequent for the other three conditionsThis suggests that pendular motion of the points withrespect to each other is an important cue to conveydepth This is in accordance with earlier suggestions(Hoffman amp Flinchbaugh 1982 Johansson 1973Kuhlmann et al 2009) The tendency to respond alongthe two cardinal axes is also known for the kineticdepth effect and is called the repulsion effect (Hiris ampBlake 1996)

Our results are consistent with a kinetic depth effectin combination with an oblique repulsion Neverthe-less biological motion is a special case in this respectbecause the kinetic depth is typically defined for rigidobjects biological motion is nonrigid The kineticdepth in this case can only function if the observer usesimplicit knowledge of the body structure and themotions that the human body is capable of

Cues to depth

The conditions of the main experiment weredesigned to systematically change the stimulus indifferent ways to remove some information whileleaving other information intact (cf Table 1)

The human form was intact in four conditionsnormal reversed marionette and frame-scrambledThe rubber conditions were physically impossiblebecause the knees and elbows became overextendedand the limb segments stretched and compressed duringthe movement Only the normal stimuli presenteddynamically correct human movements because undernormal gravity conditions the marionette movementscannot be performed and forward walking played inreverse temporal order differs from real backwardwalking (even if the marionette and reversed stimuli arekinematically and biomechanically valid) Intact pen-dular relative movements of pairs of point-lights werepresent in the normal reversed and the marionettestimulus Finally intact point trajectories were presentin the normal the rubber and the position-scrambledstimulus whereas the local motions are dynamically

incorrect for reversed stimuli (cf human movements)Comparisons of accuracy and precision betweendifferent conditions thus allow us to estimate the roleof the different depth cues

Accuracy differed widely between subjects asexpressed by the large error bars in Figure 5A Thewithin-subject ANOVA showed a clear difference onlyfor the position-scrambled stimulus This indicates thatpoint trajectories and local motions alone give onlypoor depth information

The precision of the judgments is shown in Figure5B In this respect the subjects were very similar butthere were large differences between the conditions

The precision was significantly better in the normaland reversed conditions than in all other conditionsThis suggests that the human form was an importantcue On the other hand neither the human movementnor the local motions were important because bothwere dynamically incorrect in the reversed condition(cf Table 1)

Comparing the marionette to the normal stimulusreveals a small but significant decrease in precision Themarionette is consistent with the human form at alltimes but dynamically incorrect with respect to humanmovement point trajectories and local motions Thisreiterates the importance of human form and alsosuggests that point trajectories might be importantHowever one should take into account that theconfiguration of the human form in the marionettestimulus is less familiar than in the normal or reversedconditions

Precision for the frame-scrambled condition wassignificantly better than for the position-scrambledcondition (Figure 5B) This is consistent with thefinding that the human form is an important cueHowever the frame-scrambled condition was signifi-cantly worse than the normal and reversed conditionsalthough all three conditions contain veridical humanform information The difference between the frame-scrambled condition and the normal and reversedconditions lies in pendular motion and point trajecto-ries which are both absent in the frame-scrambledcondition (cf Table 1) The rubber and marionettestimuli help to resolve this issue Marionette has correctpendular motion but incorrect point trajectoriesRubber has incorrect pendular motion but correctpoint trajectories Since the precision was better inmarionette than in rubber stimuli the pendular motionsmust be important for judging the facing in depth

This is consistent with a recent finding that thepresence of points for the feet improves the detection ofsmall angular deviations from a frontal view of walking(Cai Yang Chen amp Jiang 2011 Kuhlmann et al2009) Indeed the notion that human movements aretypically pendular has been suggested as an importantcue for recognizing biological motion long ago (Hoff-man amp Flinchbaugh 1982 Johansson 1973 Webb amp

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 9

Aggarwal 1982) In a slightly different manner this hasalso been proposed more recently by Neri (2009)

Lastly even the precision for the position-scrambledcondition was marginally better than the expectancyvalue for giving the same response on each trial (ie2628) Thus the rhythmic oscillatory movements ofthe individual points may have provided some infor-mation about the simulated facing in depth

Altogether our results support two major cues forthe judgment of facing in depth The first is the familiarhuman form The importance of the familiarity of bodyform for judging the orientation in depth of biologicalmotion is consistent with electrophysiological findingsin monkeys (Vangeneugden et al 2011) and psycho-physical findings in humans (Bulthoff et al 1998Jackson amp Blake 2010 Kuhlmann et al 2009Verfaillie amp de Graef 2000) The second is the pendularmovement of the limbs (Hoffman amp Flinchbaugh1982) The body is an articulated structure withinherent pendular motions Since these motions arenot random but highly characteristic these are apotential additional cue The data give support to theuse of pendular motion as an additional cue for judgingthe facing in depth

Supplementary material

There is a Quicktime movie for each of the sixconditions of the main experiment Each moviepresents without interruptions one walking cycle forthe facing in depth of 08 308 608 908 1208 1508 and1808

Acknowledgments

This study is funded by the German FederalMinistry of Education and Research BMBF [Grant01EC1003A] We thank Martin Habelbarth for his helpwith collecting the data We thank the anonymousreviewers for their constructive comments

Commercial relationships noneCorresponding author Marc H E de LussanetEmail lussanetwwudeAddress Psychologisches Institut Westf Wilhelms-Universitat Munster Munster Germany

References

Albright T D (1989) Centrifugal directionality bias inthe middle temporal visual area (MT) of the

macaque Visual Neuroscience 2 177ndash188[PubMed]

Ball K amp Sekuler R (1980) Human vision favorscentrifugal motion Perception 9(3) 317ndash325[PubMed]

Beintema J A amp Lappe M (2002) Perception ofbiological motion without local image motionProceedings of the National Academy of Sciencesof the USA 99 5661ndash5663 doi101073pnas082483699 [PubMed]

Blanz V Tarr M J amp Bulthoff H H (1999) Whatobject attributes determine canonical views Per-ception 28 575ndash599

Brooks A Schouten B Troje N F Verfaillie KBlanke O amp van der Zwan R (2008) Correlatedchanges in perceptions of the gender and orienta-tion of ambiguous biological motion figuresCurrent Biology 18(17) R728ndashR729 doi101016jcub200806054 [PubMed]

Bulthoff I Bulthoff H H amp Sinha P (1998) Top-down influences on stereoscopic depth-perceptionNature Neuroscience 1 254ndash257 [PubMed]

Cai P Yang X Chen L amp Jiang Y (2011) Motionspeed modulates walking direction discriminationThe role of the feet in biological motion perceptionChinese Science Bulletin 56(19) 2025ndash2030 doi101007s11434-011-4528-6

Chang D H F amp Troje N F (2009) Accelerationcarries the local inversion effect in biologicalmotion perception Journal of Vision 9(1)19 1ndash17 httpjournalofvisionorg9119 doi1011679119 [PubMed] [Article]

Fahle M amp Wehrhahn C (1991) Motion perceptionin the peripheral visual field Graefersquos Archive forClinical and Experimental Ophthalmology 229(5)430ndash436 doi101007BF00166305 [PubMed]

Foster D H amp Gilson S J (2002) Recognizing novelthree-dimensional objects by summing signals fromparts and views Proceedings of the Royal Society BBiological Sciences 269 1939ndash1947 doi101098rspb20022119

Georgeson M A amp Harris M G (1978) Apparentfoveofugal drift of counterphase gratings Percep-tion 7 527ndash536 [PubMed]

Giese M A (2004) Neural model for biologicalmovement recognition A neurophysiologicallyplausible theory In L M Vaina S A Beardsleyamp S Rushton (Eds) Optic flow and beyond (pp443ndash470) Dordrecht NL Kluwer Academic Pub-lishers

Grill-Spector K Kushnir T Edelman S AvidanG Itzchak Y amp Malach R (1999) Differential

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 10

Impression of depth

If the stimulus looks essentially two-dimensional theobserver will tend to underestimate how much thestimulus faces out of the plane of the display (up to acertain degree when the displayed facing direction isobviously close to frontal) As implied by the largeerror bars in Figure 5A some subjects generally had agood impression of depth resulting in a high accuracywhereas others saw little depth in the stimuli leading toa systematic negative mean error

The number of responses along the cardinal axes ismore suitable to compare the impression of depthbetween the six conditions Such errors were rare forthe three conditions with correct pendular motion (cfTable 1) but frequent for the other three conditionsThis suggests that pendular motion of the points withrespect to each other is an important cue to conveydepth This is in accordance with earlier suggestions(Hoffman amp Flinchbaugh 1982 Johansson 1973Kuhlmann et al 2009) The tendency to respond alongthe two cardinal axes is also known for the kineticdepth effect and is called the repulsion effect (Hiris ampBlake 1996)

Our results are consistent with a kinetic depth effectin combination with an oblique repulsion Neverthe-less biological motion is a special case in this respectbecause the kinetic depth is typically defined for rigidobjects biological motion is nonrigid The kineticdepth in this case can only function if the observer usesimplicit knowledge of the body structure and themotions that the human body is capable of

Cues to depth

The conditions of the main experiment weredesigned to systematically change the stimulus indifferent ways to remove some information whileleaving other information intact (cf Table 1)

The human form was intact in four conditionsnormal reversed marionette and frame-scrambledThe rubber conditions were physically impossiblebecause the knees and elbows became overextendedand the limb segments stretched and compressed duringthe movement Only the normal stimuli presenteddynamically correct human movements because undernormal gravity conditions the marionette movementscannot be performed and forward walking played inreverse temporal order differs from real backwardwalking (even if the marionette and reversed stimuli arekinematically and biomechanically valid) Intact pen-dular relative movements of pairs of point-lights werepresent in the normal reversed and the marionettestimulus Finally intact point trajectories were presentin the normal the rubber and the position-scrambledstimulus whereas the local motions are dynamically

incorrect for reversed stimuli (cf human movements)Comparisons of accuracy and precision betweendifferent conditions thus allow us to estimate the roleof the different depth cues

Accuracy differed widely between subjects asexpressed by the large error bars in Figure 5A Thewithin-subject ANOVA showed a clear difference onlyfor the position-scrambled stimulus This indicates thatpoint trajectories and local motions alone give onlypoor depth information

The precision of the judgments is shown in Figure5B In this respect the subjects were very similar butthere were large differences between the conditions

The precision was significantly better in the normaland reversed conditions than in all other conditionsThis suggests that the human form was an importantcue On the other hand neither the human movementnor the local motions were important because bothwere dynamically incorrect in the reversed condition(cf Table 1)

Comparing the marionette to the normal stimulusreveals a small but significant decrease in precision Themarionette is consistent with the human form at alltimes but dynamically incorrect with respect to humanmovement point trajectories and local motions Thisreiterates the importance of human form and alsosuggests that point trajectories might be importantHowever one should take into account that theconfiguration of the human form in the marionettestimulus is less familiar than in the normal or reversedconditions

Precision for the frame-scrambled condition wassignificantly better than for the position-scrambledcondition (Figure 5B) This is consistent with thefinding that the human form is an important cueHowever the frame-scrambled condition was signifi-cantly worse than the normal and reversed conditionsalthough all three conditions contain veridical humanform information The difference between the frame-scrambled condition and the normal and reversedconditions lies in pendular motion and point trajecto-ries which are both absent in the frame-scrambledcondition (cf Table 1) The rubber and marionettestimuli help to resolve this issue Marionette has correctpendular motion but incorrect point trajectoriesRubber has incorrect pendular motion but correctpoint trajectories Since the precision was better inmarionette than in rubber stimuli the pendular motionsmust be important for judging the facing in depth

This is consistent with a recent finding that thepresence of points for the feet improves the detection ofsmall angular deviations from a frontal view of walking(Cai Yang Chen amp Jiang 2011 Kuhlmann et al2009) Indeed the notion that human movements aretypically pendular has been suggested as an importantcue for recognizing biological motion long ago (Hoff-man amp Flinchbaugh 1982 Johansson 1973 Webb amp

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 9

Aggarwal 1982) In a slightly different manner this hasalso been proposed more recently by Neri (2009)

Lastly even the precision for the position-scrambledcondition was marginally better than the expectancyvalue for giving the same response on each trial (ie2628) Thus the rhythmic oscillatory movements ofthe individual points may have provided some infor-mation about the simulated facing in depth

Altogether our results support two major cues forthe judgment of facing in depth The first is the familiarhuman form The importance of the familiarity of bodyform for judging the orientation in depth of biologicalmotion is consistent with electrophysiological findingsin monkeys (Vangeneugden et al 2011) and psycho-physical findings in humans (Bulthoff et al 1998Jackson amp Blake 2010 Kuhlmann et al 2009Verfaillie amp de Graef 2000) The second is the pendularmovement of the limbs (Hoffman amp Flinchbaugh1982) The body is an articulated structure withinherent pendular motions Since these motions arenot random but highly characteristic these are apotential additional cue The data give support to theuse of pendular motion as an additional cue for judgingthe facing in depth

Supplementary material

There is a Quicktime movie for each of the sixconditions of the main experiment Each moviepresents without interruptions one walking cycle forthe facing in depth of 08 308 608 908 1208 1508 and1808

Acknowledgments

This study is funded by the German FederalMinistry of Education and Research BMBF [Grant01EC1003A] We thank Martin Habelbarth for his helpwith collecting the data We thank the anonymousreviewers for their constructive comments

Commercial relationships noneCorresponding author Marc H E de LussanetEmail lussanetwwudeAddress Psychologisches Institut Westf Wilhelms-Universitat Munster Munster Germany

References

Albright T D (1989) Centrifugal directionality bias inthe middle temporal visual area (MT) of the

macaque Visual Neuroscience 2 177ndash188[PubMed]

Ball K amp Sekuler R (1980) Human vision favorscentrifugal motion Perception 9(3) 317ndash325[PubMed]

Beintema J A amp Lappe M (2002) Perception ofbiological motion without local image motionProceedings of the National Academy of Sciencesof the USA 99 5661ndash5663 doi101073pnas082483699 [PubMed]

Blanz V Tarr M J amp Bulthoff H H (1999) Whatobject attributes determine canonical views Per-ception 28 575ndash599

Brooks A Schouten B Troje N F Verfaillie KBlanke O amp van der Zwan R (2008) Correlatedchanges in perceptions of the gender and orienta-tion of ambiguous biological motion figuresCurrent Biology 18(17) R728ndashR729 doi101016jcub200806054 [PubMed]

Bulthoff I Bulthoff H H amp Sinha P (1998) Top-down influences on stereoscopic depth-perceptionNature Neuroscience 1 254ndash257 [PubMed]

Cai P Yang X Chen L amp Jiang Y (2011) Motionspeed modulates walking direction discriminationThe role of the feet in biological motion perceptionChinese Science Bulletin 56(19) 2025ndash2030 doi101007s11434-011-4528-6

Chang D H F amp Troje N F (2009) Accelerationcarries the local inversion effect in biologicalmotion perception Journal of Vision 9(1)19 1ndash17 httpjournalofvisionorg9119 doi1011679119 [PubMed] [Article]

Fahle M amp Wehrhahn C (1991) Motion perceptionin the peripheral visual field Graefersquos Archive forClinical and Experimental Ophthalmology 229(5)430ndash436 doi101007BF00166305 [PubMed]

Foster D H amp Gilson S J (2002) Recognizing novelthree-dimensional objects by summing signals fromparts and views Proceedings of the Royal Society BBiological Sciences 269 1939ndash1947 doi101098rspb20022119

Georgeson M A amp Harris M G (1978) Apparentfoveofugal drift of counterphase gratings Percep-tion 7 527ndash536 [PubMed]

Giese M A (2004) Neural model for biologicalmovement recognition A neurophysiologicallyplausible theory In L M Vaina S A Beardsleyamp S Rushton (Eds) Optic flow and beyond (pp443ndash470) Dordrecht NL Kluwer Academic Pub-lishers

Grill-Spector K Kushnir T Edelman S AvidanG Itzchak Y amp Malach R (1999) Differential

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 10

Aggarwal 1982) In a slightly different manner this hasalso been proposed more recently by Neri (2009)

Lastly even the precision for the position-scrambledcondition was marginally better than the expectancyvalue for giving the same response on each trial (ie2628) Thus the rhythmic oscillatory movements ofthe individual points may have provided some infor-mation about the simulated facing in depth

Altogether our results support two major cues forthe judgment of facing in depth The first is the familiarhuman form The importance of the familiarity of bodyform for judging the orientation in depth of biologicalmotion is consistent with electrophysiological findingsin monkeys (Vangeneugden et al 2011) and psycho-physical findings in humans (Bulthoff et al 1998Jackson amp Blake 2010 Kuhlmann et al 2009Verfaillie amp de Graef 2000) The second is the pendularmovement of the limbs (Hoffman amp Flinchbaugh1982) The body is an articulated structure withinherent pendular motions Since these motions arenot random but highly characteristic these are apotential additional cue The data give support to theuse of pendular motion as an additional cue for judgingthe facing in depth

Supplementary material

There is a Quicktime movie for each of the sixconditions of the main experiment Each moviepresents without interruptions one walking cycle forthe facing in depth of 08 308 608 908 1208 1508 and1808

Acknowledgments

This study is funded by the German FederalMinistry of Education and Research BMBF [Grant01EC1003A] We thank Martin Habelbarth for his helpwith collecting the data We thank the anonymousreviewers for their constructive comments

Commercial relationships noneCorresponding author Marc H E de LussanetEmail lussanetwwudeAddress Psychologisches Institut Westf Wilhelms-Universitat Munster Munster Germany

References

Albright T D (1989) Centrifugal directionality bias inthe middle temporal visual area (MT) of the

macaque Visual Neuroscience 2 177ndash188[PubMed]

Ball K amp Sekuler R (1980) Human vision favorscentrifugal motion Perception 9(3) 317ndash325[PubMed]

Beintema J A amp Lappe M (2002) Perception ofbiological motion without local image motionProceedings of the National Academy of Sciencesof the USA 99 5661ndash5663 doi101073pnas082483699 [PubMed]

Blanz V Tarr M J amp Bulthoff H H (1999) Whatobject attributes determine canonical views Per-ception 28 575ndash599

Brooks A Schouten B Troje N F Verfaillie KBlanke O amp van der Zwan R (2008) Correlatedchanges in perceptions of the gender and orienta-tion of ambiguous biological motion figuresCurrent Biology 18(17) R728ndashR729 doi101016jcub200806054 [PubMed]

Bulthoff I Bulthoff H H amp Sinha P (1998) Top-down influences on stereoscopic depth-perceptionNature Neuroscience 1 254ndash257 [PubMed]

Cai P Yang X Chen L amp Jiang Y (2011) Motionspeed modulates walking direction discriminationThe role of the feet in biological motion perceptionChinese Science Bulletin 56(19) 2025ndash2030 doi101007s11434-011-4528-6

Chang D H F amp Troje N F (2009) Accelerationcarries the local inversion effect in biologicalmotion perception Journal of Vision 9(1)19 1ndash17 httpjournalofvisionorg9119 doi1011679119 [PubMed] [Article]

Fahle M amp Wehrhahn C (1991) Motion perceptionin the peripheral visual field Graefersquos Archive forClinical and Experimental Ophthalmology 229(5)430ndash436 doi101007BF00166305 [PubMed]

Foster D H amp Gilson S J (2002) Recognizing novelthree-dimensional objects by summing signals fromparts and views Proceedings of the Royal Society BBiological Sciences 269 1939ndash1947 doi101098rspb20022119

Georgeson M A amp Harris M G (1978) Apparentfoveofugal drift of counterphase gratings Percep-tion 7 527ndash536 [PubMed]

Giese M A (2004) Neural model for biologicalmovement recognition A neurophysiologicallyplausible theory In L M Vaina S A Beardsleyamp S Rushton (Eds) Optic flow and beyond (pp443ndash470) Dordrecht NL Kluwer Academic Pub-lishers

Grill-Spector K Kushnir T Edelman S AvidanG Itzchak Y amp Malach R (1999) Differential

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 10

processing of objects under various viewing condi-tions in the human lateral occipital complexNeuron 24 187ndash203

Harries M H Perrett D I amp Lavender A (1991)Preferential inspection of views of 3-D modelheads Perception 20(5) 669ndash680 doi101068p200669

Hill H amp Johnston A (2007) The hollow-faceillusion Object-specific knowledge general as-sumptions or properties of the stimulus Percep-tion 36(2) 199ndash223 doi101068p5523 [PubMed]

Hiris E amp Blake R (1996) Direction repulsion inmotion transparency Visual Neuroscience 13(1) 187ndash197 doi101017S0952523800007227 [PubMed]

Hoffman D D amp Flinchbaugh B E (1982) Theinterpretation of biological motion Biological Cy-bernetics 42 195ndash204 doi101007BF00340076[PubMed]

Holm S (1979) A simple sequentially rejectivemultiple test procedure Scandinavian Journal ofStatistics 6 65ndash70

Jackson S amp Blake R (2010) Neural integration ofinformation specifying human structure from formmotion and depth The Journal of Neuroscience30(3) 838ndash848 doi101523jneurosci3116-092010[PubMed]

Jackson S Cummins F amp Brady N (2008) Rapidperceptual switching of a reversible biologicalfigure PLoS ONE 3(12) e3982 doi101371journalpone0003982

Johansson G (1973) Visual perception of biolog-ical motion and a model for its analysis Percep-tion amp Psychophysics 14 201ndash211 doi103758BF03212378

Jokisch D amp Troje N F (2003) Biological motion asa cue for the perception of size Journal of Vision3(4)1 252ndash264 httpjournalofvisionorg341doi101167341 [PubMed] [Article]

Kuhlmann S de Lussanet M H E amp Lappe M(2009) Perception of limited lifetime biologicalmotion from different viewpoints Journal ofVision 9(10)11 1ndash14 httpjournalofvisionorg91011 doi10116791011 [PubMed] [Article]

Lange J Georg K amp Lappe M (2006) Visualperception of biological motion by form Atemplate-matching analysis Journal of Vision6(8)6 836ndash849 httpjournalofvisionorg686doi101167686 [PubMed] [Article]

Lange J amp Lappe M (2006) A model of biologicalmotion perception from configural form cues TheJournal of Neuroscience 26(11) 2894ndash2906 doi101523jneurosci4915-052006 [PubMed]

Lange J amp Lappe M (2007) The role of spatial andtemporal information in biological motion percep-tion Advances in Cognitive Psychology 3(4) 419ndash429 [PubMed]

Lappe M amp Rauschecker J P (1995) Motionanisotropies and heading detection BiologicalCybernetics 72 261ndash277 [PubMed]

Lee J amp Wong W (2004) A stochastic model for thedetection of coherent motion Biological Cybernet-ics 91 306ndash314 doi101007s00422-004-0516-0[PubMed]

Manera V Cavallo A Chiavarino C Schouten BVerfaillie K amp Becchio C (2012) Are youapproaching me Motor execution influences per-ceived action orientation PLoS ONE 7(5)e37514doi101371journalpone0037514

Michels L Kleiser R de Lussanet M H E SeitzR J amp Lappe M (2009) Brain activity forperipheral biological motion in the posteriorsuperior temporal gyrus and the fusiform gyrusDependence on visual hemifield and view orienta-tion NeuroImage 45 151ndash159 doi101016jneuroimage200810063

Neri P (2009) Wholes and subparts in visualprocessing of human agency Proceedings of theRoyal Society B Biological Sciences 276(1658)861ndash869 doi101098rspb20081363 [PubMed]

Peissig J J amp Tarr M J (2006) Visual objectrecognition Do we know more now than we did 20years ago Annual Review of Psychology 58(1) 75ndash96 doi101146annurevpsych58102904190114

Perrett D I amp Harries M H (1988) Characteristicviews and the visual inspection of simple facetedand smooth objects lsquoTetrahedra and potatoesrsquoPerception 17(6) 703ndash720 doi101068p170703[PubMed]

Rauschecker J P von Grunau M W amp Poulin C(1987) Centrifugal organization of direction pref-erences in the catrsquos lateral suprasylvian visualcortex and its relation to flow field processingThe Journal of Neuroscience 7(4) 943ndash958[PubMed]

Schouten B Troje N F Brooks A van der ZwanR amp Verfaillie K (2010) The facing bias inbiological motion perception Effects of stimulusgender and observer sex Attention Perception ampPsychophysics 72(5) 1256ndash1260 doi103758APP7251256 [PubMed]

Schouten B Troje N F amp Verfaillie K (2011) Thefacing bias in biological motion perception Struc-ture kinematics and body parts Attention Per-ception amp Psychophysics 73(1) 130ndash143 doi103758s13414-010-0018-1 [PubMed]

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 11

Schouten B Troje N F Vroomen J amp VerfaillieK (2011) The effect of looming and recedingsounds on the perceived in-depth orientation ofdepth-ambiguous biological motion figures PLoSONE 6(2) e14725 doi101371journalpone0014725

Sweeny T D Haroz S amp Whitney D (2012)Reference repulsion in the categorical perception ofbiological motion Vision Research 64 26ndash34 doi101016jvisres201205008 [PubMed]

Troje N F amp Bulthoff H H (1996) Facerecognition under varying poses The role of textureand shape Vision Research 36(12) 1761ndash1771doi1010160042-6989(95)00230-8

Troje N F amp Westhoff C (2006) The inversioneffect in biological motion perception Evidence fora lsquolsquolife detectorrsquorsquo Current Biology 16 821ndash824 doi101016jcub200603022

Vangeneugden J De Maziere P A Van Hulle MM Jaeggli T Van Gool L amp Vogels R (2011)Distinct mechanisms for coding of visual actions inmacaque temporal cortex The Journal of Neuro-science 31(2) 385ndash401 doi101523jneurosci2703-102011

Vanrie J Dekeyser M amp Verfaillie K (2004)Bistability and biasing effects in the perception of

ambiguous point-light walkers Perception 33 547ndash560 [PubMed]

Vanrie J amp Verfaillie K (2006) Perceiving depth inpoint-light actions Perception amp Psychophysics68(4) 601ndash612 doi103758BF03208762 [PubMed]

Verfaillie K amp de Graef P (2000) Transsaccadicmemory for position and orientation of saccadesource and target Journal of Experimental Psy-chology Human Perception and Performance 261243ndash1259 [PubMed]

Wang G Tanifuji M amp Tanaka K (1998) Fuc-tional architecture in monkey inferotemporal cor-tex revealed by in vivo optical imagingNeuroscience Research 32 33ndash46 [PubMed]

Watson T L Johnston A Hill H C H amp TrojeN F (2005) Motion as a cue for viewpointinvariance Visual Cognition 12(7) 1291ndash1308

Webb J A amp Aggarwal J K (1982) Structure frommotion of rigid and jointed objects ArtificialIntelligence 19 107ndash130 doi1011776863

Yellott J I amp Kaiwi J L (1979) Depth inversiondespite stereopsis The appearance of random-dotstereograms on surfaces seen in reverse perspectivePerception 8(2) 135ndash142 [PubMed]

Journal of Vision (2012) 12(11)14 1ndash12 de Lussanet amp Lappe 12