close

Вход

Забыли?

вход по аккаунту

?

The constructive view of scene perception

код для вставкиСкачать
Scene perception after those first
few hundred milliseconds
Jeremy Wolfe
Brigham and Women's Hospital and Harvard Medical
School
The constructive view of scene perception
The constructive view of scene perception
The constructive view of scene perception
The constructive view of scene perception
Appealing…but wrong
The transient binding view of scene perception
The transient binding view of scene perception
The transient binding view of scene perception
The transient binding view of scene perception
The transient binding view of scene perception
The transient binding view of scene perception
If you want to know what is here
?
You will have to check
What is the evidence for this claim?
The repeated search task
A
a
D
G
A
B
F
Y
f
D
G
A
B
F
Y
h
D
G
A
B
F
Y
B
y
D
G
F
Y
• The capital letters are the search array.
• The lower case letter at the center tell you
what to look for on this trial.
• In repeated search, the search array does not
change from trial to trial.
Repeated Search
K
e
E
R
T
V
Repeated Search
K
c
E
R
T
V
Repeated Search
K
t
E
R
T
V
Repeated Search
K
r
E
R
T
V
Repeated Search
1
T
T
A
b
Q
G
B
T
A
p
Q
350
3
T
A
f
Q
G
B
2
t
Q
G
B
G
B
Standard Search
T
S
G
B
J
F
A
b
Q
2
1
H
f
B
M
V
p
R V
350
3
Z
L
N
t
T
Memory Search
1
A
Q
B
G
f
p
S
E
350
2
T
A
t
Repeated search never becomes efficient.
S lo pe (m sec /ite m )
80
60
40
20
0
-20
0
100
200
300
Trial number
REPEATED SEARCH
Slope of the RT x set size function is the measure of search efficiency
R e pe a te d se a r c h is up at the “c e iling ”
80
60
S
L
O
P
E
40
20
0
-20
0
100
200
300
T r ial num be r
R E PE A T E D SE A R C H
C E IL IN G
M e m or y se ar c h is ac tually m or e e ffic ie nt
80
60
S
L
O
P
E
40
20
0
-20
0
100
200
300
T r ial num be r
R E PE A T E D SE A R C H
C E IL IN G
M E M OR Y SE A R C H
Actually, only 2AFC memory search is efficient
Localization responses
100
80
Slope
60
msec /item
40
20
100
200
300
400
500
Trials
Memory
Repeated
Unrepeated
Melina Kunar’s mouseclick experiments. (Kunar, Flusberg, & Wolfe)
But this is a meeting about scenes
So Aude and I did a version with scenes like this
Same basic result
Unrepeated
Repeated
Note: Mean RT is faster, but search efficiency is unchanged.
We also did “panoramic” search
Panorama Experiments
Panorama Experiments
Panorama Experiments
Panorama Experiments
Panorama Experiments
Panorama Experiments
Panorama Experiments
You can search for a visible target
You still get a slope
If the same target is hidden
The slope can vanish
Subjects make a pragmatic choice
Search if the target is visible
The “IsWas” Paradigm
Minimal Change Blindness
Behold, I bring you a mystery.
Does the cued dot change color?
No
No
Yes
No
So…to summarize
E a rly V isio n
V isu a l e x p e rie n c e
A selective pathway
E a rly V isio n
and
subsequent
awareness
with an
attentional
bottleneck
feeding
object
recognition
B in d in g
R eco g n itio n
T h e se le c tiv e p a th
Access to the bottleneck is controlled by
V isu a l e x p e rie n c e
guiding representation.
E a rly V isio n
C olor
O rien tation
S ize
M o tion
D ep th
etc.
T h e G u id in g
R e p re se n ta tio n
B in d in g
R eco g n itio n
T h e se le c tiv e p a th
V isu a l e x p e rie n c e
A non-selective pathway can
fill in the rest of the experience
E a rly V isio n
C olor
O rien tation
S ize
M o tion
D ep th
etc.
T h e G u id in g
R e p re se n ta tio n
B in d in g
R eco g n itio n
T h e se le c tiv e p a th
Документ
Категория
Презентации
Просмотров
5
Размер файла
5 298 Кб
Теги
1/--страниц
Пожаловаться на содержимое документа