Transcript No Slide Title

BUILDING MEMORIES I:
ATTENTION AND REHEARSAL
• Themes
– Learning across multiple “episodes”
– Factors that influence encoding
•
•
•
•
Presented information
What you do with it
What you know about it
The context of the encoding episode
– In general, memory will be influenced by
• The quantity of practice
• The distribution of practice
• The quality of practice
– Informal strategies for learning
• What do you do?
ATTENTION AND LEARNING
• The unimportance of being earnest
(Hyde & Jenkins, 1969)
24 words presented
instructions about
recall
Encoding task:
rate pleasantness
detect # of e’s
incidental
intentional
16.3
16.6
9.4
10.2
• The importance of being awake
Simons & Emmons (1956)
– Word lists presented during sleep
– EEG recorded to confirm sleep
– Next day: recognition d’ = 0
• Tilley (1979)
– 20 pictures of concrete objects shown
before sleep
– Ten repetitions of object names at
different sleep stages
– Next day: better recall, recognition of
named objects
– But only for “shallow” stages of sleep
• The importance of paying attention
– The classic “shadowing’ studies
(e.g., Moray 1959: 35 reps don’t help)
– Dual-task studies and divided attention
(Murdock, 1965)
– Are some attributes of events encoded
“automatically”?
• Frequency
• Recency
• Temporal & spatial distribution
– (Hasher & Zacks, 1979)
• Evidence that even these can be influenced
by attention, age, etc.
AMOUNT OF PRACTICE
• Retention increases monotonically with
amount of practice
– Repetitions across lists (Ebbinghaus, 1885)
– Repetitions within list (Rundus, 1971)
• The Power Law of Practice
log(Y) = a * (log [practice]) + b
taking the “antilog” of each side:
Y = b * (practice)a
– Ubiquitous in declarative and procedural
learning
– A number of models can generate it
– E.g., Estes’ classic Stimulus Sampling
Theory (1960’s)
THE POWER LAW OF PRACTICE
Speed and accuracy improve,
but at an ever-slower rate
(L O G
m )i n / p a g e
Task: reading inverted text (Kolers, 1975)
19
16
13
10
7
4
1
2
4
8
16
32
64 128
# of pages read (LOG)
m in /p a g e
100
10
1
2
4
8
16
32
64 128
# of pages read (LOG)
Time = 10
a
x
practice
-b
DISTRIBUTION OF PRACTICE:
THE SPACING EFFECT
• The total-time hypothesis (Bugelski
1962)
8 CVC-CVC pairs @
sec/pair
6
8
10
12
19
total trials 10.2 8.8 5.8
4.7
3.3
total time 61.2 70.1 57.9 56.1 62.2
• Spacing across days
– Spanish vocabulary (Bahrick & Phelps
(1987)
•
•
•
•
Two sessions
0, 1 or 30 days between sessions
Immediate test: no diffs
8 years later: 30-day is 2.5 times better
– Typing Skill (Baddeley & Longman,
1978)
• One- or two-hour blocks
• One or two blocks per day
• Spaced practice group learns twice as fast
• Spacing within sessions
– The “lag” effect (e.g., Melton 1962)
e.g. Underwood (1970):
42 nouns for free recall, one/sec rate
1 to 4 presentations, massed or spaced
1
2
3
massed 15%
17%
17%
19%
spaced
31%
42%
47%
16%
4
• Limits to Spacing Advantage
– Immediate tests after study
– “data-driven” encoding and memory
tasks
– Very-long lag between presentations
• The wonders of “expanded
rehearsal”
EXPLANATIONS OF THE
SPACING EFFECT
• Encoding variability and relational
processing
– Idea: increasing retrieval paths
– Spacing helps free recall > cued recall
– Forcing variability sometimes helps,
sometimes hurts, final recall
• Deficient attention (and its
consequences)
– Idea: massed presentations give
habituation, less attention and learning
e.g., Johnson & Uhle (1976):
repeat Underwood (1970), measure
“tone probe” secondary task RT:
1
spaced
2
3
4
321
330
328
238
223
206
282
massed
• Deficient Rehearsal
– Idea: less “covert” rehearsal if massed
– Spacing does increase overt rehearsal
– Spacing advantage even in incidentalmemory tasks
• Consolidation
– Idea: massed practice prevents full
consolidation
– Can it handle wide “scale” of spacing
effects?
• Retrieval practice
– Idea: spaced study gives “covert
retrieval” of prior encounter
– Forcing retrieval (test/study) gives
better memory than study only (Carrier
& Pashler, 1992)
Random Practice and
“Procedural Reinstatement”
• The importance of “procedural
reinstatement” in long-term
retention (Healy)
– Relation to ideas about encoding
specificity (Tulving), learning as
transition from declarative to procedural
representation (Anderson)
– Procedures as algorithms
– When procedures are reinstated,
learning is maintained
Fenrich et al. ’95:
• Ss read 4-digit sequences
• Enter on one of two keypads
• Recognition of sequences better if same
keypad used for test
• The contextual interference effect
– Battig (1966): high interference during
learning gives better final retention
– Schmidt & Bjork (92): “random”
scheduling of skill components gives
worse acquisition, but much better longterm retention
– Analogy to spacing effect and
immediate versus delayed tests
– Healy & Sinclair: provides better
“retrieval practice” of procedures
• Retrieval practice or relational
encoding?
– Lincoln (2001): practice on novel
arithmetic symbols and rules
• Blocked
• Random
• Blocked-with-intervening-task
– Strong contextual interference effect
– Blocked-IT equivalent to Blocked
• So not a good test of retrieval practice
– Small advantage for relational
comparison (exp. 2 vs. exp 1))
– Contextual interference effect a
– function of the number of rules per
block (exp. 3)
– Overall, suggests strong role of retrieval
practice, but procedures can be
maintained during irrelevant WM
Materials used by Lincoln (2001)
Figure 1. Kanji characters and
associated rules used in Exp
1. The characters have no
structural similarity, and the
mapping of characters to rules
is arbitrary.
Figure 2. Kanji characters and
associated rules used in Exps 2
and 3. Each pair of characters
shares a feature, which is
linked to the similarity of the
rules for that pair.
Lincoln (2001): contextual interference
And long-term retention, unrelated rules
Rule Selection Errors
during Learning (Exp 1)
70
Mean Errors
60
Random
Blocked
Blocked-IT
50
40
30
20
10
0
1
2
3
4
5
6
Learning Block
Rule Selection Errors
at 48-hr Retention (Exp 1)
60
Random
Blocked
Blocked-IT
Mean Errors
50
40
30
20
10
0
1
2
Retention Block
3
Lincoln (2001): relational cues
During acquisition
Rule Selection Errors
during Learning (Exp 2)
60
Random
Blocked
Blocked-IT
Mean Errors
50
40
30
20
10
0
1
2
3
4
5
6
Learning Block
Rule Selection Errors
at 48-hr Retention (Exp 2)
Mean Errors
50
Random
Blocked
Blocked-IT
40
30
20
10
0
1
2
Retention Block
3
Lincoln (2001): effects of number of
Rules per training block
Rule Selection Errors
during Learning (Exp 3)
60
Random
Blocked
2 Rules
3 Rules
Mean Errors
50
40
30
20
10
0
1
2
3
4
5
6
Learning Block
Rule Selection Errors
at 48-hr Retention (Exp 3)
50
Random
Blocked
2 Rules
3 Rules
Mean Errors
40
30
20
10
0
1
2
Retention Block
3