You are on page 1of 1

The silhouette illusion: Evidence for a

viewing-from-above bias
Matthew McAdam & Nikolaus Troje, Queen’s University, Kingston, Canada
INTRODUCTION Initial Silhouette Percept
(Internet Poll)
RESULTS

Proportion of Time Seen From Above


The well known “Silhouette Illusion” is a rotating depth-ambiguous 75 Overall CW/CCW Proportion Above Proportion

% of Responses
figure. It can be perceived by observers as rotating clockwise (CW) or .75 .75 *

Proportion of Time
counterclockwise (CCW). 50
Informal internet polls have recorded a CW biasfor the silhouette1. No
such bias has been reported for other rotating depth-ambiguous 25 .50 .50
figures2. Here, we provide evidence for a viewing-from-above bias to
account for the puzzling CW bias present in the “Silhouette Illusion”. 0
CW CCW .25 .25
Depth-Ambiguous Figures (e.g. Necker cube)
 Provide the visual system with input for two plausible interpretations. Necker Cube
 Remain physically unchanged as the observer’s percept alternates. 0 0
CW CCW 3° 10°
Kayahara’s “Silhouette Illusion”3 Camera Elevation Magnitude
Neither group is different from 0.5, ps > .05,
 Is rendered with a camera elevation of 7° relative to the horizontal or the other t(23) = 0.53, p = .60 Both groups are greater than 0.5, ps < .05,
Which
plane. As it rotates, its limbs trace an ellipse, not a straight line. do you and different from one another, t(23) = 4.58, p < .001
 Observers must assume a viewpoint from above or below. see?
 From above is paired with CW, from below is paired with CCW. Viewing-From-Above Bias
 Operationalized as the slope of the proportion of CCW viewing time as a function of camera elevation.
Silhouette – Rotating OR
From From
Above Below Example: VFA Bias for Participant DG Effects of Angular Velocity on VFA Bias
Overall Slope: 0.01deg-1 All groups are greater than 0, ps < .001
.70 No differences between groups, F(2, 46) = 2.62, p > .05
Silhouette – Still Frame
.65 0.020
25°/s

CCW Proportion
.60

VFA Bias (Slope)


100°/s
175°/s 0.015
.55 All
.50
Which 0.010
1 CW do you
.45
see? .40
7° 7° 0.005
.35
OR
0
From Above From Below -10° -3° 3° 10° 25°/s 100°/s 175°/s
+ + Camera Elevation Angular Velocity
2 CCW Facing Toward Facing Away

Instability (rev/min)

Instability (rev/min)
Perceptual Instability 5.0 * 5.0
Silhouette - Ellipse 4.0 4.0
 Operationalized as the # of
3.0 3.0
perceptual reversals/min .
Viewing-From-Above (VFA) Bias 2.0 2.0
 Perceptual reversals are flips
from CW to CCW and vice versa. 1.0 1.0
 While many are aware of it, it has only been documented sparsely . 0 0
Perhaps most prominently in association with contour line drawings4. 3° 10° 25°/s 100°/s 175°/s
Camera Elevation Magnitude Angular Velocity
t(23) = 3.13, p = .005 F(2, 46) = 0.29, p = .75

METHODS SUMMARY & DISCUSSION REFERENCES


Design – 4 x 3 (Camera Elevation x Angular Velocity) Procedure  When CW/CCW and above/below pairings were balanced, 1. Munger (2008) http://scienceblogs.com/cognitivedaily/
no CW/CCW bias was present, but there was an above bias. 2008/10/casual_ fridays_tk421_why_cant.php
 Presentation Time: 4 min  Each participant (N = 24) ran 12 trials. 2. Jackson, Cummins, & Brady (2008) PLoS ONE, 3, e3982.
 The bias present in the original “Silhouette Illusion” is
 ITI: 30 s  Task: Hold a key (‘s’ or ‘k’) to indicate dol: 10.1371/journal.pone.0003982.
actually a VFA bias, not a CW bias as previously thought.
 Stimuli:12 variations of the silhouette illusion perceived direction of rotation. 3. Kayahara (2003) http://www.procreo.jp/labo/silhouette.
 Stimuli presented at greater camera elevations (absolute)
 Camera elevations:-10, -3, 3, 10 deg IN 3D!! swf
were more stable and were seen from above more often. 4. Mamassian & Landy (1998) Vision Research 38:2817-2832
 Angular velocities:25, 100, 175 deg/s
 Angular velocity had no effect on VFA bias or stability.

Contact: troje@queensu.ca

You might also like