Binding problems and feature integration theory

1
Binding problems and feature integration theory
2
Feature detectors

• Neurons that fire to specific features of a
  stimulus
• Pathway away from retina shows neurons that fire
  to more complex stimuli
• Cells that are feature detectors
• Simple cortical cell
• Complex cortical cell
• End-stopped cortical cell

3
Single cell recording of neurons in the temporal
lobe
An electrode is inserted in this area, and
measure neural responses when stimuli are changed
gradually
4
(No Transcript)
5
Neurons in this area respond to complex stimuli
like those shown on the left.
6
Selective adaptation
7
(No Transcript)
8
Selective adaptation to size
9
(No Transcript)
10
(No Transcript)
11
1 2 3 4 5 6 7 8 9 10 11 12
12
(No Transcript)
13
1 2 3 4 5 6 7 8 9 10 11 12
14
(No Transcript)
15
(No Transcript)
16
(No Transcript)
17

• Figure 3.30 How neurons that respond best to
  narrow (N), medium (M), and wide (W) bars respond
  to the medium-bar grating on the right of Figure
  3.28. (a-b) Response before adaptation. (c)
  Response after adaptation to the wide-bar grating
  at the top left. (d) Response after adaptation to
  the narrow-bar grating on the bottom left.

18
After Image
19
(No Transcript)
20
(No Transcript)
21
After images Red ? Green Green ? Red Blue ?
Yellow Yellow ? Blue
R
G
B
Y
Color perception
22
B
Y
G
adaptation
adaptation
After image
After image
23
Binding problem
24
Two visual pathways (what where/how systems)
Image from Neuroscience, 2nd Ed. (2000).
25
Columnar organization

• Neurons that respond to the same orientation are
  packed in the same column

Image courtesy of Dr. Paul Wellman and
Neuroscience, 2nd Ed. (2000).
26
Distributed coding then what do you need?

• ? Combining input

27

• Feature integration theory by Treisman Gelade
• Attention plays a central role in solving the
  binding problem.
• Attention helps organize information.

28
Attention and Figure-ground segregation
Depending on where you look at, the figure and
the ground switch rapidly. ? Attention plays some
role in determining the figure and the ground.
29
Attention and 3D structure
Depending on where to look at, you get different
3D structures.
30

31
(No Transcript)
32
(No Transcript)
33
Damage to the parietal lobe creates binding
errors.
34
(No Transcript)
35
(No Transcript)
36
Copies of the black (A) and the white (B)
vertical contour.
Copies of the black (A) and the white (B)
diagonal contour.
37
Copies of the left sub-figure (A) The right
subfigure (B) And the central sub-figure (C)
38
Feature integration theory

• Attention is the glue that combines the
  information from the what and where systems.

39
What can you predict from this theory?

• If you cant attend, you cant combine
  information.
• ? Illusionary conjunction

40
Demonstration

• I will show you a scene quickly.
• Report first the black numbers.
• Report what you see at each of the 4 locations.

Mask

41
(No Transcript)
42

43
(No Transcript)
44
(No Transcript)
45

• Report first the black numbers.
• Report what you saw at each of the 4 locations.

46
Illusionary conjunctions

• We tend to put different features from different
  objects together.
• Some brain damaged patients (parietal lobe) show
  illusionary conjunctions even when the patients
  were allowed to view the stimuli for 10 seconds.

47
Visual search experiments

• Looking for the target

• Feature search
• This is easy because you find the target by
  looking for a single feature.
• ? you dont need attention

• Conjunction search
• For this you need to combine two or more features
  (color and orientation)
• ? you need attention

48

• Conjunction search
• For this you need to combine two or more features
  (color and orientation)
• ? you need attention
• Because you can attend an item one at a time, the
  difficulty in the conjunction search increases
  proportional to the number of items in the
  stimulus frame.
• This is not the case in the feature search.

49
Which is more difficult?
Find
50
Which is more difficult?
Find
51
Feature binding and attention (Treisman, 1988)

• Experiments
• Task
• Given a stimulus frame containing visual items,
  subjects were asked to indicate whether or not a
  target item was present in the frame.

52
Target
53
Target
54
Target
55
Target
56
Feature search vs. conjunction search

• Feature search
• The target item has a unique feature.
• Conjunction search
• You need to combine features to find the target.
• You need attention.
• Because you can attend only one item at a time,
  the conjunction search becomes more difficult
  when more items are in the stimulus frame.

57
Generality of the results
Feature search
Conjunction search
58
Experiments

• Measure accuracy and response times (conjunction
  cases vs. non-conjunction cases)

59
(No Transcript)
60
Response time
1000 ms
500ms
of distractors
61
Response time
1000 ms
500ms
of distractors
62
The physiology of attention

• How do you combine features?
• Synchrony hypothesis
• When neurons in different parts of the cortex are
  firing to the same object, the pattern of firing
  is synchronized (they fire at the same time, and
  in the same manner).
• So when neurons are firing in synchrony, the
  corresponding features are bound together.

63

• Separate neurons respond to color (green, blue,
  white), contours (orientations), textures, so on.
• Synchrony hypothesis
• When the features come from the same object
  (i.e., the woman), these neurons fire at the same
  time in the same manner.
• When the neurons fire at the same time and in the
  same manner, we perceive binding of features.
• Attention is likely to increase synchrony

64
Attention

• Selecting information
• Enhancing information
• Combining information