Transcript document

Forgetting
1
Memory
 Internal record or representation of past experience
 Not necessarily the same as the original experience
 History & metaphors
 Slate  Filing cabinet  Computer
2
Types of Memory
 Many different types of memory
 2 are important for our purposes:
 Working memory: short-term, no need to store each
instance for future reference
 e.g. matching to sample: need to remember what the
sample was only until you make the choice
 Samples change from trial to trial
 Reference memory: long-term, remember specific
information for future reference
 e.g. maze training: remember lay-out of the maze,
doesn’t change across trials
3
Working Memory
Trial #1
Trial #2
Trial #3
Sample:
Retention
Interval:
Remember
“green”
Remember
“red”
Remember
“red”
Choice:
4
Reference Memory
Goal is always in the same place… remember over time!
Trial #1
Trial #2
Start
Food
Trial #3
Food
Start
Food
Start
5
Behaviorist View of Memory
 No need to discuss “representation”
 No focus on storage & retrieval
 Experience’s ability to change an organism’s behaviour
under certain conditions
 Stimulus control
6
Forgetting
 Deterioration in learned behaviour following a period
without practice
 Defined behaviourally
 Performance vs Description
 Note: extinction is not the same as forgetting
7
Measuring Forgetting
 Working memory
 Sample (training)
 Retention interval (usually short… seconds/minutes/hours)
 Test
 Next sample is different
 Reference memory
 Training
 Retention interval (can be much longer… days/weeks)
 Test
 Samples (training) are always the same
8
Free Recall Method
 Train, wait, test
 See how much deterioration in performance
 “All-or-nothing” test of behaviour
 May not be appropriate for complex tasks
 Some elements remembered, others not
9
Free Recall
 Learn:
 Banana
 Interesting
 Annoy
 Book
 Computer
 Recall:
 _______________
 _______________
 _______________
 _______________
 _______________
10
Prompted (Cued) Recall
 Give prompts to increase likelihood of behaviour
 Measure number of prompts needed to produce
behaviour
11
Cued Recall:
 Learn:
 Banana
 Interesting
 Annoy
 Book
 Computer
 Recall:
 Ba_________
 In_________
 An_________
 Bo_________
 Co_________
12
Relearning Method
 Reinstall original training procedure after retention
period
 How many trials (or time) needed compared to
original training to return to initial level of
proficiency?
 Reacquisition
13
Relearning
Trial #3
#1
#2
 Learn :
 Banana
 Interesting
 Annoy
 Book
 Computer
 Recall:
 Banana
 Interesting
__________
 Annoy
 Book
__________
 Computer
__________
Total Trials Score
onScore
Initial
= =2/5
3/5
5/5
Learning = 3
How many trials to relearn after a break (retention interval)?
Difference = amount of forgetting
Recognition Method
 Subject only has to identify material previously
learned
 E.g., distinguish between original stimulus and a
number of distracter stimuli
15
Recognition
 Learn:
 Banana
 Interesting
 Annoy
 Book
 Computer
 Which words were on
the list?
 Banana
 Orange
 Interesting
 Annoy
 Ugly
 Computer
 Table
 Apple
16
Delayed Matching to Sample
 Show S+
Sample
 Wait (Delay =
Retention Interval)
 Choose from S+ and S Working memory only
Delay
Matching
17
Extinction Method
 Train two subjects
 Put both on extinction, but one has delay between
training and extinction and the other doesn’t
 Compare rate of extinction for two subjects
18
Extinction methods
Group 1 & 2
Learning
Phase
Group 1
Extinction
Group 2
Extinction
break
Compare
amount of
time
19
Gradient Degradation Method
 Establish stimulus control (discrimination training)
 Measure generalization gradient
 Repeated measure gen. grad. over time
 If generalization gradient flattens, forgetting
20
Gradient degradation
No Forgetting
Training: Establish
gradient
Forgetting
21
22
Is time a variable?
 Retention interval = Time between learning and
testing
 Greater the interval, less retained (i.e., more
forgetting)
 But, time is not an event (time doesn’t account for
forgetting)
 Need causal factors
23
Variables are:
 Degree of learning (overlearning)
 Prior Learning
 Facilitation
 Interference
 Subsequent Learning
 Context
24
Overlearning
 Learn to asymptote, then keep training
 Learning list perfectly, then practice a few more times
 Better recall for longer
 Point of diminishing return
 Not a linear relationship between overlearning and
retention
 i.e. 100% overlearning is twice as good as 50%
overlearning
25
Krueger (1929)
 Adults learn 3 lists of 12 one-syllable nouns
 List 1: go through list until they remember all 12
 List 2: learn list perfectly, then go through again for half
as many trials as it took to learn

i.e. if they took 10 trials to learn perfectly, they go through list
another 5 times
 Group 3: learn list perfectly, then go through again as
many times as it took to learn

i.e. another 10 times
 Relearn after various intervals
26
Results
 Greater amount of overlearning, less forgetting
 100% overlearning better than 50% overlearning
 50% overlearning way better than 0% overlearning
 i.e. difference between 100% & 50% was LESS THAN
difference between 50% & 0%
27
Facilitation of Prior Learning
 Previous experience makes something easier to
remember
 Meaningful material easier to retain than random
material
 e.g. Easier to learn a complete sentence than 12 random
words
 Prior experience important in determining what is
meaningful (e.g., words in known or unknown
language)
28
DeGroot (1966)
 Arranged chess pieces on board as if in the middle of a
game
 Chess masters and novices; 5 seconds to observe
 Masters reproduced arrangement 90% of time, novices
only 40%
 Is this prior experience, or do chess masters forget less
than other people?
29
Chase & Simon (1973)
 Chess pieces placed randomly on board
 Masters no better than novices at recall
 Past learning of “legal” arrangements is what increased
masters’ performance in deGroot (1966) study
30
Interference of Prior Learning
 Proactive interference
 Previous learning interferes with recall of newer
learning
31
Studying proactive interference
 Paired Associate Learning (PAL) technique
 Subjects learn paired lists, tested with 1 item and must
recall second
 All learn A-C list, but some previously learned A-B list
 In testing, give A and ask to recall C
 Those with A-B learning have more difficulty recalling C
when given A
32
PAL example
 1 group first
learns:
 Both groups then
learn:
 Both groups then
RECALL:
 Red-Apple
 Red-Book
 Red- ________
 Cloud-Shoe
 Cloud-Paper
 Cloud- ________
 Cat-Shelf
 Cat-Fence
 Cat- ________
 Plate-Spoon
 Plate-Notebook
 Plate- ________
 Carpet-Tent
 Carpet-Window
 Carpet- ________
33
Levine & Murphy (1943)
 Proactive interference with attitudes
 Determine initial attitude towards communism
 Likert Scale
 Read pro- and anti-communism passages
 Students who had prior pro-communist attitudes
forgot anti-communist elements of passages but
remembered pro-elements (and vice versa)
 Proactive interference because attitudes are not
innate; effect of prior learning
34
Subsequent Learning (Interference)
 Inactivity during
100
Recall (%)
retention interval leads
to less forgetting than
activity
 Implies forgetting partly
based on learning new
material
 Jenkins & Dallenbach
(1924)
sleep
50
awake
0
2
4
6
8
Hours after learning tested
35
Retroactive Interference
 New learning interferes with ability to recall earlier
learning
 PAL technique (opposite order)
 Subjects learn A-C, but some then learn A-B
 Test by giving A and recalling C
 Subjects who learned A-B have worse recall for C
 e.g. forgetting old phone numbers, license plates,
passwords
36
PAL example
 BOTH groups
first learns:
 1 group then
learns:
 Both groups then
RECALL:
 Red-Apple
 Red-Book
 Red- ________
 Cloud-Shoe
 Cloud-Paper
 Cloud- ________
 Cat-Shelf
 Cat-Fence
 Cat- ________
 Plate-Spoon
 Plate-Notebook
 Plate- ________
 Carpet-Tent
 Carpet-Window
 Carpet- ________
Order is just “switched” from last example
37
Context
 Learning occurs in a context
 Various stimuli around the learner
 These stimuli serve as cues to evoke a behaviour
 If stimuli absent, may have cue-dependent forgetting
 Stimulus control
 e.g. forgetting names when in a different context
38
Context
SD
Colour, size, shape,
etc…
Cue set, set of SD’s, has
changed! Less cues to
signal correct response.
39
Perkins & Weyant (1958)
 Train two groups of rats in two mazes, one black, one
white
 1 minute retention interval
 Half of each group tested in original maze, half in
maze of opposite colour
 Opposite colour rats did poorly compared to original
maze tested rats
40
Kamin (1957)
learning, tested at
various retention
intervals.
 Time of day, internal
clock
 Internal physiological
state cues recall
 “internal” context
Avoidance (%)
 Gave rats avoidance
100
50
0
12
24
36
48
60
72
84
Retention Interval (hr)
41
State-Dependent Learning
 Train under a particular physiological state (e.g., drug
condition) and test under various states
 Recall best when in the same state as training
 Drug conditions: alcohol, caffeine, etc.
 Internal State: tired, level of stress, emotions, etc.
42
Application: Foraging
 Food Caching
 Cache: food store
 Retrieval of food later
 Spatial memory
 Wide variety of species
 Accuracy can be quite high for very long times
43
Application: Eyewitness Testimony
 Notoriously poor
 Basic issue of retention interval and forgetting
 Also the nature of the question used to retrieve
information
44
Loftus & Zanni (1975)
 Subjects watched film of car accident
 Asked “Did you see <the>/<a> broken headlight?”
 “the” subjects twice as likely as “a” subjects to say
“yes”
 Actually, no broken headlight shown
 Reinforcement history
 Previous conditioning: “the” (definite article)
implies presence; “a” implies possible presence
45
Loftus & Palmer (1974)
 Watch film of car accident
 “How fast were the cars going when they hit each
other?”
 Underlined word replaced with smashed, collided,
bumped, contacted
 Speed estimates varied based on wording of question
 Reports of broken glass varied based on wording
46
47
Learning to Remember
 In essence, improving learning
 Practice increases retention
 Techniques:
 Overlearning
 Mnemonics
 Context cues
 Prompts
48
Overlearning
 Practice beyond learning
 e.g. Tiger Woods putting practice
 e.g. Flash Cards (SAFMEDS)
49
Mnemonics
 Rhymes, First Letters
 HOMES, Roy G. Biv
 Method of Loci
 Associate learned items with locations on a well-known
route
 Peg Word System
 1 = “bun”, 2 = “shoe”, 3 = “tree”…
 Also works with visual
 1 looks like a pencil, 2 looks like a swan, 3 is a tricycle…
50
Mnemonic example
 1 – “coffee cup” – imagine using a coffee cup as a pencil





holder
2 – “rubber ball” – imagine a swan holding a rubber ball in
its beak
3 – “printer” – imagine a printer printing a piece of paper
with a tricycle on it
4 – “yoga mat” – imagine trying to balance in yoga
positions while on top of a table (4 legs)
Etc….
The more bizarre the image, the easier it is to remember
(lack of interference)
51
Prompts
 Creating S+’s
 Memos, notes, calendar markings
 Often don’t contain all info, so just a “reminder” of what
needs to be done
 E.g. “3-Choice meeting”
 String on finger, watch beep, cell phone reminders
52