Transcript 5. Forgetting - USF :: College of Arts and Sciences

5 – Forgetting
Demo: On a penny, what appears to the left of Lincoln?
To the right?
We forget almost everything we once knew.
What causes forgetting?
Can forgetting be avoided or at least diminished?
Scenario
3:00 pm
4:00 pm
5:00 pm
You
study History
French Lab
History exam
Your Friend
study History
rest
History exam
Does the French Lab affect the score on History Exam?
Two possibilities:
“no”
decay
amount forgotten depends solely on time elapsed since event
“yes”
interference
learning A interferes with the learning of B
Experiment
Ss studied nonsense syllables
Then Ss slept or remained awake for 0 – 8 hours
Ss given test
Predictions
Decay:
Sleep
=
Awake
Interference: Sleep
>
Awake
Results
Test
Score
sleep
awake
0
8
Test Delay (h)
(Jenkins & Dallenbach, 1924)
(Jenkins & Dallenbach, 1924)
Follow-up Experiment
Roaches learned to avoid shock.
Then placed on treadmill or squeezed in matchbox.
1 day later, “still” Ss showed no forgetting.
(Minami & Dallenbach, 1946)
Karl M. Dallenbach
Study
Ss played on rugby team. All Ss missed at least one game during the season.
Example
Player A
Player B
Games
1 2
 

3

4
5

6

7

8
9


Sample Test Question
After game 9, Ss were asked “Who did you play in the 1st game?”
Predictions:
Decay: accuracy depends on # of days since 1st game
A=B
Interference: accuracy depends on # of games played since 1st game
A<B
Results supported interference.
(Baddeley & Hitch, 1977)
Two kinds of interference:
Proactive Interference (PI): prior learning hinders subsequent learning
Day 1
Day 2
Day 3
PI group
Spanish
French
Test on French
Controls
------
French
Test on French
PI occurred if PI group did worse
Retroactive Interference (RI): subsequent learning hinders prior learning
Day 1
Day 2
Day 3
RI group
Spanish
French
Test on Spanish
Control Group
Spanish
------
Test on Spanish
RI occurred if RI group did worse
Example
In 10th grade, Brielle took French 1
In 11th grade, she took Spanish 1.
In 12th grade, she took Spanish 2.
On the first day of Spanish 2, she took pop quiz. For green, she wrote “vert.” (French)
What kind of interference explains her error?
Answer
PI
French
Spanish
Spanish Quiz
Prior learning interfered with what she was trying to remember
Example
Today, for the first time in her life, Jill used a numeric keypad.
She entered data for about an hour.
Then she phoned her friend.
(She has used a phone all her life.)
7 8 9
4 5 6
1 2 3
1 2 3
4 5 6
7 8 9
When she tried to press “9,” she mistakenly pressed “3”.
What kind of interference explains her error?
RI
Use phone (before today)
use keypad
Test on phone
subsequent learning interfered with what she was trying to remember
Example
A Brit flies to Tampa for his first trip outside the UK.
He rents a car and drives around Tampa for several days.
Then, while in Tampa, and while driving on narrow road at night, he sees oncoming car.
He veers left.
What kind of interference explains his error?
PI
drive in UK
drive in US
test on US learning
Associated Press- Tampa - November 29, 2002
“A driver killed in a high-speed, head-on collision Wednesday after crossing the
Sunshine Skyway on the wrong side was identified Thursday as a British citizen.”
Other real-world examples of interference?
Why does Proactive Interference occur?
Usually, PI occurs only if cue is paired with more than one target.
Example
1:00 pm
2:00 pm
study French (red-rouge) study Spanish (red-rojo)
rouge
red
rojo
This explanation of interference is called cue overload theory
(e.g., Watkins, 1977)
3:00 pm
Spanish test (red-?)
Example of cue overload theory
Today, Joe parked in the same lot for the 10th consecutive day (different spot each day)
cue = where in this lot did I park?
cue is paired with 10 targets
cue overload  cannot find car
Today, Moe parked in a lot the first time
cue = where in this lot did I park?
cue is paired with 1 target
No cue overload  can find car
Why does Retroactive Interference occur?
Sometimes, RI occurs because cue is paired with more than target.
Example
1:00 pm
2:00 pm
study French (red-rouge) study Spanish (red-rojo)
3:00 pm
French test (red-?)
But RI happens even if cue is linked to only one target
Example
1:00 pm
2:00 pm
study French (red-rouge) study History
History impairs French test score!
So cue overload theory cannot be only cause of RI.
3:00 pm
French test (red-?)
consolidation theory
1. Memory needs time to strengthen or “consolidate.”
jello must harden
2. Until consolidation is complete, memory is vulnerable.
jello can spill
3. Consolidation is disrupted by concurrent new learning.
fridge door open
Example
1:00 pm
2:00 pm
3:00 pm
Mr. X
study French
study History
Test on French
Mr. Y
study French
rest
Test on French
Mr. X does worse because studying History impaired consolidation of French
(e.g., Wixted, 2005)
Evidence for Consolidation
Observational data
After car accident, victim cannot recall last 15 minutes prior to accident.
Experiment
ECT = electroconvulsive therapy
Mr. X
study …ECT…………………………………….test
Mr. Y
study ………………………………….ECT……test
Mr. X does worse. ECT erased memory before it had a chance to consolidate.
(Ribot, 1881; Squire et al.,. 1975)
Example
One day, Ned and Fred studied History.
Then, Ned napped while Fred studied French.
Then, they took History test.
Ned
History (1492  Columbus)
Nap
History Test (1492-?)
Fred
History (1492  Columbus)
French (dog-chien)
History Test (1492-?)
According to consolidation theory, who should do WORSE?
Fred.
His nap impaired consolidation of History.
Example
One day, Holly and Spence studied French vocabulary
Then, Holly studied History while Spence studied Spanish.
Then, they took French vocabulary test.
Holly
French (dog-chien)
Spence French (dog-chien)
History
French Test (dog-?)
Spanish (dog-perro)
French Test (dog-?)
a. According to cue overload theory, who should do WORSE?
Spence because his cue (DOG) is paired with 2 targets – not just 1.
b. According to consolidation theory, who should do WORSE?
Equally poor. For each, consolidation disrupted.
Recognition
Often times, we cannot recall item but we can recognize it.
Example
You cannot recall his name, but you would recognize it if you heard it.
Other times, we can neither recall nor recognize.
How should we measure recognition?
Yes-No Recognition method
Example
Study: S is shown 4 words:
girl, wall, rope, sign
Test: S is shown 4 original words (targets) and 5 new words (foils), one at a time.
Test Item
rain
sign
boat
wire
food
girl
rope
rake
wall
Item Type
foil
target
foil
foil
foil
target
target
foil
target
H  hit rate 
Did you see it before?
no
yes
hit = said yes to target
no
no
yes
false alarm = said yes to foil
no
yes
hit
no
yes
hit
# of yeses to targets 3
  75%
# of targets
4
FA  false alarm rate 
# of yeses to foils
1
  20%
# of foils
5
Example
Hypothetical Experiment
Study phase: Each S saw 50 face photos, one at a time.
Test.
Each S saw random mix of the 50 old faces (targets) and 80 new faces (foils).
For each face, S was asked if he or she recognized face.
Jane said “yes” to 40 targets and 20 foils.
1. Find H.
# of yeses to targets
40
H  hit rate 

 80%
# of targets
50
2. Find FA.
FA false alarm rate 
# of yeses to foils 20

 25%
# of foils
80
Why measure both H and FA?
H is not sufficient. Why?
A subject could simply say “yes” to every item and achieve perfect score (100%)
FA is not sufficient. Why?
A subject could simply say “no” to every item and achieve perfect score (0%)
But measuring both Hit rate (H) and False Alarm rate (FA) can lead to ambiguity.
Example
Ann:
H = 80%
FA = 60%
Ben:
H = 50%
FA = 20%
Who did better?
Ann had better H, but Ben had better FA.
Thus, we need a single measure that combines H and FA.
A simple but crude measure is
H – FA
A better measure is d’ (“d prime”), which is beyond the scope of this course.
Example
A researcher conducts a face recognition experiment.
During the study phase, Ss saw dozens of faces, one at a time.
During the test, Ss saw a random mix of the original faces and many new faces.
For each face, S was asked, ”Do you recognize this face?”
1. Anna performed perfectly. Find H – FA.
H = 100%
FA = 0%
H – FA = 100%
FA = 100%
H – FA = 0%
FA = 0%
H – FA = 0%
2. Beth said “yes” to every face. Find H – FA.
H = 100%
3. Carol said “no” to every face. Find H – FA.
H = 0%
4. Donna flipped a coin for each test face (heads = yes). Estimate H – FA.
H = 50%
FA = 50%
H – FA = 0%
5. For Emma, H = 10% and FA = 90%. What can you conclude?
She misunderstood instructions or sabotaged your experiment.
Face Memory Demo
You’ll see several faces, one at a time.
Just look at each face.
Part 1
Study Face
Study Face
Part 2
Study Face
Study Face
Get ready for Test phase
You’ll see 14 faces.
Your page is numbered 1 – 14.
For each test face, write “yes” or “no”
Twist:
If your answer is yes, also write
1
if face appeared in Part 1 (these faces had blue border)
or
2
if face appeared in Part 2 (these faces had orange border)
Test 1
Test 2
Test 3
Test 4
Test 5
Test 6
Test 7
Test 8
Test 9
Test 10
Test 11
Test 12
Test 13
Test 14
Source Memory Demo
1
H=
/4=
FA =
/ 10 =
2
3
target
Part 2
4
H – FA =
5
6
target
Part 2
10
target
Part 1
11
target
Part 1
7
8
9
12
13
14
www.bbc.co.uk/science/humanbody/sleep/tmt/instructions_1.shtml
Source Memory
Often we can recall a fact while forgetting its source
Where I did I read that?
Who told me that?
When did I learn that?
Examples
You know that Alaska is the largest state, but you cannot recall where you learned this.
You know that Bunny broke up with Chad, but you cannot recall who told you this.
Experiment
Ss studied list of non-famous names (List 1)
½ Ss waited a minute, and ½ Ss waited a day
Ss given List 2 and asked to circle famous names. List 2 includes
famous names
new non-famous names (not seen before)
old non-famous names (from List 1)
Results
minute delay
day delay
circled famous name (correct)
often
often
circled new non-famous name (error)
rare
rare
circled old non-famous name (error)
rare
often
Conclusion:
After a day, Ss forgot the source of their memory
(Jacoby, Kelley, Brown, & Jsechko, 1989)
Observational Study
In 1974, John Dean gave senate testimony about Watergate Scandal
His memory of oval office conversations was amazingly detailed.
Later, he, senate, and public learned that president secretly recorded conversations.
Detailed comparison of testimony and tapes revealed
Dean correctly recalled the gist of what was said
Dean was often wrong about who said it and when it was said
In other words, he made source memory errors.
(Neisser, 1981)
Common Criticism of Forgetting Research
Too many lab studies use procedures that are irrelevant to real world.
Examples
learning a list of unrelated words
testing subjects just 20 seconds after they studied the material
In other words, most memory research lacks ecological validity.
In recent decades, more forgetting studies use ecologically valid tasks.
Example
Remember to take medicine
(Einstein, 2000)
Forgetting Study
Ss were tested on their memory of names and faces from high school class
Time since graduation = 2 weeks - 57 years
Six tests, including cued name recall (see face photo, give name)
Sample Result
Time Since Graduation
Cued Name Recall Accuracy
3 months
70%
25 years
50%
In general, all 6 tests revealed surprisingly little forgetting.
Conclusion
Heavy, repeated exposure produces less forgetting than lab studies suggest.
(Bahrick, Bahrick, & Wittlinger, 1975)
Yearbook study
Bahrick, Bahrick, and Wittlinger (1975)
Can you recall high school classmate names if you saw their yearbook photos?
Author Demo
Try to recall the author of each book.
Initials
Name
Pride and Prejudice
JA
Jane Austen
War and Peace
LT
Leo Tolstoy
Canterbury Tales
GC
Geoffrey Chaucer
The Origin of Species
CD
Charles Darwin
Gone with the Wind
MM
Margaret Mitchell
Les Miserables
VH
Victor Hugo
blank
retrieval failure
Memory cannot be recalled but it is not forgotten
Examples
I could not recall her name, but I recognized it as soon as I heard it.
I could not recall who first sailed around world, but his name came to me later.
I cannot recall who assassinated Robert Kennedy, but his name begins with an S.
Tip-of-the-Tongue experience
TOT Demo
1
Experiment
2
Ss heard definitions of rare words
3
“If you don’t know word, can you recall anything (e.g., 1st letter)?
4
5
Question: Was accuracy greater than chance?
6
7
Results:
Yes
8
9
10
11
(Brown, 1991; Brown & McNeill, 1966)
12
13
TOT
In a survey of 51 languages, 45 use tongue metaphor.
On the Tongue
On the Tip of the Tongue
On the Point of the Tongue
On the Head of the Tongue
On the Front of the Tongue
Sparkling at the End of the Tongue (Korean)
(Schwartz, 1999)
1. Three-line verse with 5, 7, and 5 syllables per line
1. haiku
2. the three periods (...) that indicate an omission
2. ellipsis
3. create odd-shaped voting districts for political benefit
3. gerrymander
4. word that reads same forward or backward (e.g., tot)
4. palindrome
5. Japanese art of paper folding
5. origami
6. term for mammals with pouches (e.g., kangaroo)
6. marsupials
7. tree that sheds leaves every year (unlike evergreens)
7. deciduous
8. flying reptile that became extinct
8. pterodactyl
9. word that is pronounced like its meaning (e.g., buzz)
9. onomatopoeia
10. mental disorder that causes people to steal
10. kleptomania
11. instrument used to measure wind speed
11. anemometer
12. animal which eats plants and animals
12. omnivore
13. person who explores a cave
13. spelunker
Demo -
tree
palm
You’ll see 10 word pairs, very quickly.
leader
king
baseball
pitcher
monkey
banana
tennis
racket
leather
saddle
soccer
ball
time
noon
river
bridge
fire
hot
Distractor task
Demonstration
For each cue, write target
1. queen
2. fruit
3. horse
4. lunch
5. stove
6. wrist
7. lemonade
8. noise
9. dance
10. cards
Place check by each correct answer
1. queen
king
tree
palm
2. fruit
banana
leader
king
3. horse
saddle
baseball
pitcher
4. lunch
noon
monkey
banana
5. stove
hot
tennis
racket
6. wrist
palm
leather
saddle
7. lemonade
pitcher
soccer
ball
8. noise
racket
time
noon
9. dance
ball
river
bridge
10. cards
bridge
fire
hot
Results: 6 – 10 harder
blank
encoding specificity
recall improved if study context = test context
Experiment
Ss study related pairs (strawberry – JAM)
Test includes wrong-context cue (traffic - ?) or no cue
Cue doesn’t help.
Why not? Because words were “encoded” in a “specific” context.
(Tulving, 1970s)
Internal State Dependence
You forget less if your mind or body is in the same state during study and test.
Typical Experiment
Sad or happy film clip
Study phase
Delay
Sad or happy film clip
Test
Four groups: sad-sad, sad-happy, happy-sad, happy-happy
Results
Mood-matched Ss do slightly better on test.
Another Experiment on Internal State Dependence
Ss pedal or sit still on stationary bike while studying and while taking test.
Results
Words recalled at test
Test state
pedal
still
pedal
6
4
still
3
7
Study state
Practical Implications?
(Miles & Hardman, 1998)
Does chewing vs. not chewing gum produce context effects?
Miles: More work needs to be done…
2008
Miles, C., Charig, R., & Eva, H. (2008). Chewing gum as context: Effects in long-term
memory. Journal of Behavioural and Neuroscience Research, Vol. 1 (6), 1-5.
Johnson, A., & Miles, C. (2008). Chewing gum and context-dependent memory: The
independent roles of chewing gum and mint flavour. British Journal Of Psychology, 99,
293-306.
2007
Johnson, A. J., & Miles, C. (2007). Evidence against memorial facilitation and contextdependent memory effects through the chewing of gum. Appetite, 48(3), 394-396.
Miles, C., & Johnson, A. J. (2007). Chewing gum and context-dependent memory
effects: A re-examination. Appetite, 48(2), 154-158.
How about learning while asleep or while awake?
External State Dependence we forget less if study environment = test environment
Common Claims
You should study in the room where you will take the test
If you smell odor X while studying, smelling odor X will help you remember
Data:
Bunk
How would you design an experiment to test these claims?
Experiment
Scuba divers studied a list of words underwater or on pool deck
½ Ss took recall test underwater, and ½ Ss took recall test on pool deck
Results
Test Environment
dry
wet
dry
14
9
wet
9
Study Environment
11
Conclusion
In extreme scenarios, matching external states improves recall.
(Godden & Baddeley, 1975)
The End
Drawback of multiple-choice measure
Example
Ss studied 10 words:
girl, wall, rope, sign, shoe, head, moon, tree, mass, lion
One week later, Ss took test. “For each question, circle the word you studied earlier.”
1. a. zeal
b. lair
c. sign
d. whim
2. a. rope
b. aloe
c. clod
d. duel
etc.
Ss can choose correct answer by eliminating foils (incorrect choices).
Thus, S can exhibit perfect recognition memory without having any memory of words!
Multiple-choice is not a good measure….
Demo – Divide class into 2 groups
1. Who designed first VW?
a) Klein
b) Benz
c) Speer
d) Porsche
c) Ford
d) Porsche
2. Who designed first VW?
a) Marconi
b) Edison
Encoding specificity has been used to improve eyewitness memory
“First recall everything about crime scene. Now …”
(Geiselman et al., 1985)
Another experiment demonstrating encoding specificity
Ss read sentences, and the last word in each sentence was a thing.
For each thing, Ss saw one of two sentences.
Example
½ Ss read "The man lifted the piano." (weight context)
½ Ss read “The man tuned the piano." (music context)
Later, Ss given test: “Try to recall the last word of each sentence.”
During test, Ss received a cue for each word they were trying to recall.
Example
For piano, cue was “something heavy” or “something with a nice sound”
Results: Ss recalled 3 times as much if study context matched test cue.
(Barclay et al., 1974)
“An obvious explanation for this interesting phenomenon is that the reduced
rate of memory formation while one is under the influence of the drug
protects recently formed memories during a period of time when they are
especially vulnerable.” (Wixted, 2005)
October 30, 2005
Dear Abby:
I am engaged to an otherwise great guy I'll call
"Wayne," who has a bad habit. He calls me by
his ex-wife's name. The first couple of times it
happened, I called it a mistake. But now it
happens habitually, and I'm at my wit's end.
...
I've had nightmares about it happening at the
altar. I don't think I'd be big enough to forgive
that. What do you think?
Canadian toddler Erika Nordby wandered
outside at night and nearly froze to death in
2001. She wore only a diaper and T-shirt. It was
minus 11 Fahrenheit (-24 Celsius).
When found, her heart had stopped beating for
two hours and her body temperature was 61
degrees. She suffered severe frostbite but
required no amputations and otherwise
recovered.
Interference causes more forgetting when learning material is similar.
Example
Joe takes French (vert-green) and Spanish (verde-green)
Moe takes French (vert-green) and German (grün-green)
French Test: What is French word for green?
Joe suffers more interference than Moe because vert and verde are similar.
Some terminology.
If you learn an association between items (face and name). to pair two items
together two items “A-B” and later see “A” and try to recall “B” then
A is the cue and B is the target
Example
You meet fellow student.
Later, you see student and try to recall his or her name.
The face is the cue, and the name is the target.
Study
Test
face-name
face - ?
Part 2. learning new memory disrupts consolidation of recently-formed memory
Experiment
Ss voluntarily take benzodiazepine, which impairs memory for subsequent info.
Group
first 5 min
Benzo Ss
learn list A
Placebo Ss
learn list A
next 5 min
last 5 min
take benzo
learn list B
Test
take placebo
learn list B
Test
Results
Test on List B list: Placebo Ss did better
Benzo impeded subsequent learning (this is just a biology fact)
Test on List A: Benzo Ss did better
Poor learning of list B  more resources for consolidation of list A
(e.g., Fillmore et al., 2001; Hinrichs et al., 1984; Weingartner et al., 1995)
Most common measure of yes-no recognition performance is …
d’ (“d prime”)
a measure of a subject’s ability to discriminate between two kinds of items
(such as old and new faces)
---------------------------------------------------Students are not responsible for material below
---------------------------------------------------d’ can be found by consulting table or using following formula.
d’ = Z(TP) – Z(FP) where Z(p) = z-score of cumulative area under normal curve
Larger d’ value indicates greater discriminability.
This approach is known as signal detection theory
Other scenarios in which one must make a discrimination.
Is that an enemy bomber or a flock of geese on the radar screen?
Should we hire/admit this applicant?
Is he schizophrenic?
Raise your finger if you hear a beep.
Is the defendant guilty?
Should I reject H0?
Is the man in this photo the same man who robbed you?
Example
Ss saw 50 words during a study session.
Next day: Ss saw test with 50 original words (targets) and 100 new words (foils).
For each word, S asked if he or she recognized it from previous day (“yes” or “no”).
Abby said “yes” to 35 targets and 30 foils.
Beth responded correctly to 30 targets and 60 foils.
Find Abby’s H and FA.
H = 35/50 = 70%
FA = 30/100 = 30%
Find Beth’s H and FA.
H = 30/50 = 60%
FA = 40/100 = 40%
By the measure H – FA, who did better?
Abby (40%) outscored Beth (20%)