2. Outline
Processes that underlie change detection
Memory retention
Memory retrieval
Comparison and decision
I will talk about four series of experiments
3. Homework
After the talk, you have to write 1,000-words
summary, including:
What you have learned
The unsolved issues
New idea and ongoing experiments
4. Change Detection (CD)
Change detection is a fundamental function of
human vision
Detecting changes is important
Successful change detection enables us to perform
proper action
e.g., dealing with traffic, etc
5. However,
People may often not detect a large change in the
visual environment
Change Blindness
Demo
7. 改變盲的意涵
The absence of visual representation
Grand illusion (Noe, 2002)
“…seeing constitutes an active process of probing the
external environment as though it were a continuously
available external memory.” (O’Regan, 1992)
Do people not preserve any visual representation in
their brain?
8. Multiple stages in CD
Is there any change?
Change or No change?
Encoding Retention Retrieval and Comparison
Encoding
Time
10. In summary, …
Change blindness may occur because
Representation-failure: related to encoding and
retention
Retrieval- and comparison-failure: related to retrieval,
comparison and decision
Change blindness does not necessarily mean the
absence of visual representation
11. Recent studies showed that …
Pre-change representation is preserved under
change blindness
It is more likely that failure of retrieval and
comparison may cause change blindness
12. Overview of the studies
Issue 1: Is the pre-change representation preserved
under change blindness?
Issue 2: Does retrieval failure cause change
blindness?
Issue 3: What is the nature of the retrieval
mechanism?
Issue 4: What is the nature of the comparison and
decision mechanism?
13. Overview of the studies
Issue 1: Is the pre-change representation preserved
under change blindness?
Issue 2: Does retrieval failure cause change
blindness?
Issue 3: What is the nature of the retrieval
mechanism?
Issue 4: What is the nature of the comparison and
decision mechanism?
14. Evidence of the dissociation between
change detection and recognition memory
Examine the existence of the pre-change
representation under change blindness
Task: change detection task and perceptual
identification task
Index: repetition priming effect (RPE)
identifying viewed object is faster and more accurate
than new object
Yeh & Yang (2008)
15. Procedure
1000 ms
2000 ms +
350 ms blank
2000 ms +
350 ms blank
2000 ms
Change
Detection
Perceptual identification task
Experimenter
recorded the
response
Time category
16. Results
RPE is observed under CD and CB
Pre-change representation is preserved
17. Dissociation effect as a function of similarity
between pre- and post-change objects
Manipulating the similarity between pre- and post-
change objects to investigate the effect of similarity
on change detection and recognition memory
Task: change detection task and recognition
memory task
Yeh & Yang (2008)
18. Procedure and Stimuli
1000 ms
2000 ms +
350 ms blank
2000 ms +
350 ms blank
2000 ms
Change
Detection
2AFC recognition
memory test
20. In summary, …
Pre-change representation is preserved under
change blindness
The dissociation effect
Similarity ↑ CD↓ RM ↑
Similarity does not affect the pre-change
representation, it only affects RM
Similarity benefit occurs because participants use the
similarity as retrieval cues to retrieve the pre-change
representaiton
21. Overview of the studies
Issue 1: Is the pre-change representation preserved
under change blindness?
Issue 2: Does retrieval failure cause change
blindness?
Issue 3: What is the nature of the retrieval
mechanism?
Issue 4: What is the nature of the comparison and
decision mechanism?
22. Dual processes in recognition memory
In the literature of recognition memory: two
processes underlie memory recognition
Recollection and familiarity (Atkinson & Juola, 1973,
1974; Jacoby, 1991; Yonelinas, 1994)
Recollection: high-threshold model
Slow, control process and conscious awareness
Familiarity: signal detection model
Fast, automatic, and unconscious
23. Dual processes in change detection
In the context of change detection, it seems that
both processes play roles in detecting changes
I can recollect information about the pre-change
object: including location, perceptual details, …
I know that there is something changed, but I cannot
recall any detail about the changed object
24. Purpose
To examine if retrieval-failure may cause CB
Retrieval process = recollection process?
25. Process dissociation procedure
Jacoby (1991) and Yonelinas (1994):
Study phase (list method)
List A (interested list); List B (non-interested list)
Test phase
Inclusion condition: Is the word from List A?
Exclusion condition: Is the word from List B?
26. Inclusion: Is the word from List A?
p(yes | interested list) = P ( R ) + P ( F ) − P ( R ) × P ( F )
Exclusion: Is the word from List B?
p (yes | interested list) = P ( F ) − P( R) × P( F )
where P ( F ) =φ d '− )
( c
R= Pr(Ex) - Pr(Inc)
F = Pr (Ex) / (1-R)
27. When recognizing the pre-change
object?
Inclusion condition Exclusion condition
“Pre” “Pre/New”
R R
F “Pre” F “Post”
1-R 1-R
“Post/New” “Pre/New”
1-F 1-F
Response
28. Inclusion condition Exclusion condition
Pre-change Pre-change Post-change
Post-
object or New object
change or
New object object
32. In summary, …
Although the pre-change representation is
preserved, participants may fail to detect the change
as a result of retrieval failure
Successful retrieval requires participants to recollect
the pre-change details
33. Overview of the studies
Issue 1: Is the pre-change representation preserved
under change blindness?
Issue 2: Does retrieval failure cause change
blindness?
Issue 3: What is the nature of the retrieval
mechanism?
Issue 4: What is the nature of the comparison and
decision mechanism?
34. Questions
Given that both CD and RM require successful
retrieval, why do I observe the dissociation effect?
What is the nature of the retrieval mechanism?
35. Retrieval blocking hypothesis
(Yang & Yeh, 2009)
The retrieval of the post-change object may block
the retrieval of the pre-change object
A location-based retrieval
Example
36. To verify the retrieval blocking hypothesis, I used a false
memory paradigm
Change blindness may occur because:
Comparison failure: pre-change representation is intact.
Participants should falsely recognized the pre-change lure as the
pre-change target
Retrieval blocking hypothesis: : the retrieval of the post-change
object may block the retrieval of the pre-change object.
Participants should falsely recognized the post-change lure as
the pre-change target
37. Pre-change array 2000 ms
ISI 350 ms
Post-change array 2000 ms
ISI 350 ms
Pre-4-AFC
post-4-AFC
38. Presentation duration for 2 S
Support the retrieval blocking hypothesis
Does this effect occur because pre-change objects
are not completely encoded?
0.8 Target
0.7 Similar to the target
Similar to the alternative target
0.6 Unchanged object
Proportion
0.5
0.4
0.3
0.2
0.1
0
Pre-change Post-change
39. Presentation duration for 3.6 S
Also support the retrieval blocking hypothesis
0.8 Target
Similar to the target
0.7 Similar to the alternative target
Unchanged object
0.6
Proportion
0.5
0.4
0.3
0.2
0.1
0
Pre-change Post-change
40. Discussion
Retrieval blocking hypothesis
This effect is not caused by the incompleteness of
the pre-change representation
However, the result may occur because the
attention is divided when viewing the pre-change
display; the post-change object is under focal
attention
Divided attention
The manipulation of the spatial cuing
41. Under divided attention
Memory error of the post-change lure increases,
supporting the retrieval blocking hypothesis
0.7 Target
Similar to the target
0.6
Similar to the alternative target
0.5 Unchanged object
Proportion
0.4
0.3
0.2
0.1
0
Pre-change Post-change
42. The manipulation of the spatial cuing
Memory bias occurs in the INVCD and VCB condition
0.7 Target
Similar to the target
Similar to the alternative target
0.6 Unchanged object
Forget
Proportion
0.5
0.4
0.3
0.2
0.1
0
INVCB INVCD VCB VCD
V: valid cue; INV: invalid cue; CD: correct detection; CB: change blindness
43. The retrieval mechanism
Adopting a location-based retrieval strategy, the
retrieval of the post-change object may block the
retrieval of the pre-change object for change
detection
However, this strategy may aid the recognition
memory because similarity between the pre-change
post-change objects can be used as a retrieval cue
This benefit may be a product of false recognition
44. Overview of the studies
Issue 1: Is the pre-change representation preserved
under change blindness?
Issue 2: Does retrieval failure cause change
blindness?
Issue 3: What is the nature of the retrieval
mechanism?
Issue 4: What is the nature of the comparison and
decision mechanism?
45. In the visual environment, …
A change usually contains changes of multiple
features
E.g., change of hair style (color and length)
Usually, these changes are not equally salient
E.g., hair coloring is more salient than hair shortening
46. Questions
How do people process the concurrent feature-
changes for a decision?
How do relative salience affect the comparison and
decision processes?
47. Signal detection framework
Signal detection process in
Pre-change Post-change each feature
feature feature
d’
N S+N
Pre-change
object
Feature 1
c
d’
N S+N
∑ D
Feature 2
c
d’
N S+N
Feature 3
Post-change
c
object
...
...
...
...
Relative salience affects the weighting process, rather than the process
characteristics
49. The present study
Using the systems factorial technology (SFT) to
study the comparison and decision processes
Examining how relative silence affects the
processes
Exp1: two features are equally salient
Exp2: one feature is more salient than the other
50. The SFT
Double factorial design
Redundant-target condition
51. Five possibilities of ICSF
Serial self-terminating model (SS)
Serial exhaustive model (SE)
Parallel self-terminating model (PS)
Parallel exhaustive model (PE)
Channel summation model
(coactive model) (C)
See Townsend & Nozawa (1995)
52. Inference of the process characteristics
Process architecture: serial, parallel, coactive
Stopping rule: self-terminating, exhaustive
53. Exp 1
To examine the processes when two features were equally
salient
Type of changes: frequency/ orientation-change
Ambiguity: Unambiguous change (100% detection threshold);
Ambiguous change (75 % detection threshold)
57. Summary
ICM > 0, ICSF > 0
When the detection thresholds were equated, all
participants adopted parallel self-terminating
processing to detect redundant changes
58. Exp 2
To examine the processes when one feature was
more salient than the other
Detection threshold F O
Unambiguous change 100% 75%
Ambiguous change 75% 50%
62. Summary
3 observers: ICM = 0, ICSF = 0; Observer TY: ICM > 0,
ICSF > 0
When relative salience exists, 3 observers changed
their decision strategy to serial self-terminating
processing; Observer TY remained adopting parallel
self-terminating processing to detect redundant
changes
63. Subjective self-reports
Exp1: all of them considered that two features were
equally important for a decision
Exp2
3 observers detected that frequency was more salient
than orientation
Although Observer TY detected the relative salience,
she still considered that two features were equally
important for a decision
64. In summary, …
Relative salience may affect the subjective expected
utility (SEU)
Participants may change their decision strategy to
detect redundant changes according to the SEU
65. Discussion
Many factors may cause CB. Therefore, CB does
not necessarily mean the absence of the pre-
change representation
Failure of retrieval, comparison, and decision may
cause CB
Change detection and recognition memory are
dissociable
66. The nature of the retrieval mechanism
A location-based retrieval strategy
The nature of the comparison and decision
mechanism
Not necessarily a coactive process
Flexible
Can shift according to the subjective expected utility