Get documents about "Asymmetrical face perception with in-depth rotated faces
Document Sample
Asymmetrical face perception
with in−depth rotated faces
¨
A.I. Ruppertsberg, T. Vetter and H.H. Bulthoff
TWK 99
¨ ¨
Max−Planck−Institut fur biologische Kybernetik Tubingen http://www.kyb.tuebingen.mpg.de/bu.html
1 INTRODUCTION 2 METHODS
Burt &Perrett (1997) showed that subject’s judgment of
gender, age and expression were more influenced by the
Training Testing
left than by the right side of the face (viewer’s In the learning stage, In a delayed−match−to−sample task subjects were presented
perspective). Also in similarity studies it was shown that a
left composite face was rated as more similiar to the subjects were asked a frontal view of a face for 100 ms, followed by a mask for
original than a right composite face (Gilbert et al., 1973, to study 10 frontal 500 ms, and finally a side view (+/− 30 and 60 deg) of a face
views of ANA TIM EVA ISA SAM for again 100 ms.
Rhodes, 1985).
3D−Cyberware head
Is this asymmetry an asymmetry in perception or in same
scans with their or
coding? respective names for different
person
ten minutes. MIA UTA OLI EDI ALI ?
We investigated whether recognition performance differs
for faces rotated in depth to the right or to the left. 600 100 500 500 100
Immediately after they were tested in a naming task, where a Time [ms]
face was shown on the computer screen and subjects had to
press the corresponding name key on the keyboard. When The task was to assess whether the two views depicted the
Note: Left and right are used here always from the their error rate was lower than 5% over the last 30 trials they same person or not.
viewer’s perspective.
continued with the actual experiment. At that stage they had Subjects were asked to respond as fast as possible and their
named each face at least three times. response time and errors were recorded.
3 RESULTS
Experiment 1 Normal Faces Experiment 2 Symmetrical Faces
Q: Is there a difference in performance Factors: Familiarity Orientation Q: Is the asymmetrical performance Factors: Familiarity Orientation
between left and right in depth rotated Levels: familiar −60 deg due to the slight asymmetry in the Levels: familiar −60 deg
unfamiliar −30 deg unfamiliar −30 deg
faces? +30 deg face? +30 deg
+60 deg −>We made all faces symmetrical. +60 deg
N = 15 N = 17
17 Subjects
25
Normal Faces
8
familiar
20 unfamiliar
6
Error difference [%]
5 Subjects make more
4
* errors the further the 15 For symmetrical faces subjects’
Errors [%]
Error difference
2
face is rotated away from performance show different results for
00 10
the frontal view. familiar and unfamiliar faces.
−2
Error difference [%]
−4 5
−x deg +x deg −5−6
0
For familiar faces subjects show the
−60 −30 30 60
15 Subjects −8
60 30 Subjects make more Orientation angle [deg] same behavior as for normal faces: they
25 60 30
errors when the face is Symmetrical Faces
make more errors when the face is
20 turned to the left side. rotated to the left.
Error difference [%]
55
unfamiliar
*
15 00
Errors [%]
E(−x deg) − E(+x deg) −5
−5 For unfamiliar faces the behavior is
10 60
60
30
30 reversed to the contrary: Subjects make
< 0 more errors to the right
more errors when the unfamiliar
55
familiar
5
> 0 more errors to the left symmetrical face is turned to the right.
00
0
−60 −30 30 60 −5
−5 *
Orientation angle [deg]
60 30
60 30
3 RESULTS 4 CONCLUSIONS
From Experiment 1 we conclude, that there is indeed a difference
Experiment 3 Visual Field in our ability to recognize a person depending on the side of the
face we see. In accordance with Burt and Perrett’s results that
perception of faces is biased to the left side of a face our results
Q: Is the effect due to the location of Factors: Group Familiarity Visual Field Orientation
show that people make more errors when they cannot see the left
Levels: Set 1 familiar central −60 deg
presentation in the visual field? Set 2 unfamiliar left −30 deg side of the face (−x deg rotation).
right +30 deg
−>The rotated face is presented in +60 deg
N = 18 From Experiment 2 we conclude, that the asymmetry per se in a
the left, right, or central visual field (+/− 2.6
face is not the origin of the asymmetrical performance but rather
deg).
that the analysis of familiar and unfamiliar faces differs.
There was no effect of the set
Q: Is the selection of faces biased? of faces used. From Experiment 3 we conclude that the right hemisphere is the
−>Two groups of subjects learned a specialized face processing hemisphere since presentation of
When faces are presented in
different set of ten familar faces. faces in the left visual field results in the same behavior as
the central visual field,
presentation of faces in the central visual field.
subjects make less errors than
when they are presented in Subjects make more errors for right rotated faces when they are
the left or the right visual field either unfamiliar or presented in the right visual field.
All Faces, 18 Subjects
and there is a tendency that
left VF central VF right VF
30 8 8 8 presentation in the right visual It is still not clear whethera symmertrical object is coded
left VF 6 6 6 field leads to a lower asymmetrically.
25 central VF
right VF 4 * 4 4 * performance than a
Since the effect in general is quite small can we get more
20 presentation in the left visual
Error difference
2 2 2 significant results for the observed asymmetries with more
Errors [%]
field.
15 0 0 0 subjects?
−2 −2 −2 Subjects make more errors for
10 Is there a difference in the task for familiar and unfamiliar faces
−4 −4 −4
left rotated faces when they
because the coding of the faces is different?
5
−6 −6 −6
are shown in the central and
0
left visual field.
−8 −8 −8
−60 −30 30
Orientation angle [deg]
60 60 30 60 30 60 30 In the right visual field
however subjects make more References:
errors for right rotated faces. Burt, D. M. and Perrett, D. I., 1997, Neuropsychologia 35, 685−693.
Rhodes, G.,1985, Brain and Cognition 4, 197−218.
Gilbert, C. and Bakan, P., 1973, Neuropsychologia 11, 355−362.
Related docs
