Long-term Memory

Long-term Memory
• Tests of long-term memory
– single item recognition
• how to measure performance?
– increased “hits” (saying old to targets) could be a bias to
always say old
– compare the hit rate to the false alarm rate (saying old to
distractors / foils / lures)
» assuming that recognition is performed based on
familiarity, bias and memory can be separated
» various techniques for separating bias and memory
performance
– free recall
– cued-recall (paired associate learning)
• study FLAG-SPOON, DRAWER-SWITCH, etc.
• FLAG-______
– results are similar to free recall

Theories
• levels of processing: the only difference between STM
and LTM is whether encoding is “shallow” or “deep”
– memory is a continuum; not different components
– deep processing is more semantic and self-referential and
results in stronger, longer lasting memories
– shallow processing is concerned with physical characteristics
and results in weaker, temporary memories
– Craik and Tulving (1975) had participants study adjectives with
the following four tasks (listed from worst to best memory)
• Is it in upper case?
• Does it rhyme with ?
• Does it fit in the sentence ?

• Modal view of memory (qualitative differences
between STM and LTM)
– Capacity
• STM around 7 chunks, LTM nearly unlimited
– Duration
• STM around 30 seconds, LTM can last a lifetime
–Coding
• STM is acoustic
– confuse “b” and “g”
• LTM is semantic
– Baddeley (1966a) – 20 minutes after studying a list
» poor recall for semantically similar lists (HUGE, BIG, SIZE,
etc.)
» normal recall from acoustically similar lists (MAD, MAP,
MAN, etc.)

Forgetting: Storage versus Retrieval
• interference at retrieval
– memory is like a cluttered desk
– all memory traces are preserved in LTM, but it’s hard
to retrieve them
• can’t find the right retrieval cues
– if all traces compete
• both proactive and retroactive interference

• interference at storage (neural networks)
– memory is like many pictures drawn on top of each
other so its hard to see any individual picture
– when memories are added, the representation is
permanently changed
• new memories overwrite old memories (retroactive
interference)
• old memories make it more difficult to encode new memories
(proactive interference)
– an example – adding memories to a Hopfield network
• when memories are stored, links between the features of a
memory are created
• at retrieval, activated features serve as the retrieval cues and
the system cycles until the activation no longer changes

Nurse
Sick
Study List
Lawyer
Medicine
Health
Hospital
Dentist
Physician
cure
Patient
Office
Surgeon
Bed
Rest
Awake
Tired
Dream
Snooze
Blanket
Doze
Slumber
Snore
Nap
Yawn

Results
• Forgetting Curves
– Ebbinghaus’s (1885) nonsense syllable experiment
• over the course of days, he kept on relearning lists of nonsense
syllables (e.g., RUR, HAL, BEIS, etc.)
• forgetting was measured as percent savings (a comparison of
immediate testing versus testing after a delay)
• information is rapidly forgotten at first (non-linear)

• Bahrick’s (1983) Delaware, Ohio experiment
– real-life demonstration of forgetting curves
– students and alumni of Ohio Wesleyan were asked
• free recall of streets, buildings, and landmarks
• using a map, write in street names, buildings, and landmarks
(visually cued)
• recognition of streets, buildings, and landmarks (matching)
– current students who had been there different
amounts of time provided learning curves
• buildings and landmarks learned faster
– alumni of different ages provided forgetting curves
• buildings and landmarks forgotten more slowly

Test List
Lawyer
Doctor
Rest
Sleep

• false memory
effect (DRM)
– higher false alarm
rate for critical
lures (sleep)
Test List
Lawyer
Doctor
Rest
Sleep
Nurse
Sick
Lawyer
Medicine
Health
Hospital
Dentist
Physician
cure
Patient
Office
Surgeon
Bed
Rest
Awake
Tired
Dream
Snooze
Blanket
Doze
Slumber
Snore
Nap
Yawn

• memory is worse when similar items are studied
– A,B,C, vs. A,A’,A’’ (false memory effect)
• memory is better when context is kept the same
– encoding specificity
• study “spelling-BEE” or “honey-BEE” – cue as context
• Golden and Baddeley (1975) – environmental context
– words learned on land or underwater
– Words tested on land or underwater
memory
performance
LTM Effects
study
on land
study
underwater
test
underwater
test
on land

– state-dependent learning
• same design as scuba experiment but using
pharmacological states as a context (Eich, 1980)
memory
performance
study
drunk
study
sober
test
sober
test
drunk

• memory for longer lists is worse (length effect)
– A,B,C vs. A,B,C,D,E,F
• memory is better when stimuli are studied longer
(strength effect)
– A,B,C vs. A,A,A,B,B,B,C,C,C
– distributed study is better than massed study (i.e., cramming)
• spacing effect
– A,A,A,B,B,B,C,C,C vs. C,A,BA,B,C,B,A,C
– Encoding variability
» More cues (many hooks one memory)
» Cue overload (one hook many memories)
• differences between recognition and recall
– low frequency (obscure) items are recognized better but harder
to recall
– high frequency (common) items are harder to recognized but
easier to recall

Narrative and Autobiographical Memory
• Do laboratory studies of memory apply to real life
memory for stories and personal events?
– Bartlett (1932) thought memory is often reconstructed based on
schemata (constructive view of LTM).
• serial reproduction: recall the same story on more than once
– “War of the Ghosts” became more distorted with each reproduction
– Autobiographical memory for ordinary events
• Marjorie Linton (1975, 1982) spent six years recording the daily
events in her life in a systematic fashion

• each day short descriptions of events
– the date, how distinctive, how emotional, importance to
life goals
• tested herself at delays up to three years
– order two events (which happened first)
– provide exact dates of events
• performance was good at all delays
– real life memories are more durable
» testing is similar to recognition and cued-recall
– she often used problem solving to arrive at the dates of
events
» constructivist view of memory
• events that she could not recall where often similar
to other events she could not recall
– repeated similar events can start to form a schema

– Barsalou (1988) interviewed undergrads for specific
autobiographical events from the summer
• only 21% of recollections were specific
• the rest were “summarized events” (e.g., “I went to the beach
everyday”)
– Brewer (1988) used beepers to test randomly
selected events
• Linton studied the most important events of the day instead
of the ability to recollect any particular event within the day
• at random times (~2 hours), participants were alerted by a
beeper and would write down:
– location, time, actions, and thoughts
– also a number of ratings
» how often this event occurred
» how pleasant
» how trivial or significant
– he also had them write down the most memorable event of
each day

– memory was pretty good overall
» did not fall off rapidly like Ebbinghaus’ studies
– memory was better for actions than for thoughts
– better for the most memorable events of the day
– better for infrequent events and locations
– better memory for rare actions
» the more distinctive the event, the better the memory
for that event

Flashbulb Memories
(Brown & Kulik, 1977)
• personal events at time of disaster
– JFK assassination, Challenger explosion, start of gulf
war, 9/11, etc.
• physiological response
– emotional responses cause hormone release
• better encoding
– Study of Ronald Reagan assassination attempt
• stronger emotional responses = more details

• Neisser (1982b)
– link to history
• Story retelling
– memory strengthening (rehearsal)
– progressive distortion
• Bonhannon’s (1988)
– Challenger explosion
• stronger emotions = greater memory
• more retellings = greater memory
• Weaver (1993) compared intentional memory to
flashbulb memory
– remember next meeting with roommate
• Immediately, 3 months, 1 year
– gulf war started at same time (direct comparison)
• No accuracy differences
– Ebbinghaus forgetting
• Higher confidence for flashbulb
– No relationship between accuracy and confidence

Eyewitness Memory
• memory integration (malleability)
– Elizabeth Loftus (1979): series of accident slides
• Viewed stop or yield sign
• “Did another car pass the red Datsun while it was stopped at the
stop/yield sign?”
• recognition of slides
– two alternative forced choice testing (2-AFC)
» no need to compare hit and false alarm rates
– 75% if the question was consistent
– 41% if the question was inconsistent
» More influenced by the misleading question

– Loftus (1975) had people view a film (no barn)
• “How fast was the white sports car going when it passed the
barn while traveling along the country road?”
– One week later
» Barn in sentence: 17% reported barn
» No barn in sentence: 3% reported barn
– Bransford and Franks (1971) had people read simple
sentences such as:
• The ants were in the kitchen
• The jelly was on the table
• The jelly was sweet
• The ants ate the jelly
• On a later test, they recognized with highest confidence the
unseen combined sentence:
– The ants in the kitchen ate the sweet jelly that was on the table

• does eyewitness confidence indicate accuracy?
– courts place importance on confident eyewitness testimony
– 36 of 40 DNA overturns were convictions on eyewitness
identification
• what’s the best way to conduct a lineup?
– maximize identification of guilty suspects
• similar to increasing the hit rate
– minimize identification of innocent suspects
• similar to decreasing the false alarm rate
• lineup as a psychology memory experiment
– same concerns
• demand characteristics, bias to say ‘old’, etc.

• selecting distractors in a lineup (known-innocents)
– nominal size: # in the lineup
– functional size: # fit the suspect’s description
– the use of mock eyewitnesses
• if they prefer the suspect, the lineup is flawed
– match-to-description, or also resemble-suspect?
• either way, innocent suspect is protected against erroneous identification
• if the distractors resemble-suspect, identification of guilty suspect is lower
• demand characteristics
– subtle cues
– exert pressure to select (relative judgment)
– double-blind eliminates these concerns
• Not adopted by law enforcement
• relative judgments versus absolute identification
– more likely to convict innocent suspects w/ relative judgments
– “none of the above”
– sequential lineup
• reduces ID of innocent suspects
• doesn’t reduce ID of guilty suspects
• however, might enhance demand characteristics

– is confidence related to accuracy?
• weakly in real life
– meta-analysis: 6% of the variance
• reliable in in the laboratory
– within-subjects correlation
– between-subjects differences are larger
• lineups: 1 observation per subject
– can’t assess type of person
• selection factor
– people who actually select, likely to have high confidence
• in summary
– Confidence: social and personality factors
» positive feedback from investigators
– Accuracy: perceptual and memorial factors

False Memory
• “psychological problems indicate abuse or
trauma as a child”
– the self-help book, “The Courage to Heal”
• “if you have a pattern of symptoms such as low self-esteem,
depression, etc., then you were probably abused”
– Court cases of recovered abuse
– often memories arise in therapy
• questionable therapy as source
– hypnosis
– sodium amytal
– leading questions
– imagination exercises

– Implanting new memories
• Loftus and Pickrell (1995): 3 true events and one
false “lost-in-the-mall”
– interviews of relatives
– between 4 and 6 years old at time of event
– Booklet: description of the events
» Fill in remembered details
– remembered
» 68% of the true events
» 29% of false events
– still recalled 1 and 2 weeks later
– Differences
» more words used to describe true memories
» true memories rated as being more clear

• Hyman, Husband, and Billings (1995)
– hospitalization, birthday/clown, wedding/punch, evacuate/store
» 80-90% of true events
» no false events recalled in first interview
» 20% in a second interview
» some participants elaborated upon their false memories
(added details never presented)

• Forms of suggestion
– Interviewing suspected criminals
• imagine they participated in a crime
– Therapists
• imagine childhood events to recover them
• self-help books: exercises
– where would the abuse have been likely to occur?
– who is the likely perpetrator?
• Childhood Sexual Abuse (CSA)
– Clancy et al. (2000) used DRM paradigm
• Control group
• Continuous memory of CSA
• “repressed memory” of CSA
• “recovered memory” of CSA
– Recovered group showed much higher false DRM recall

• Loftus – imagination inflation
– Likely to have occurred
– imagination
– Give likelihood ratings again
– increased for imagined
• imagination makes more familiar
• familiarity is misattributed to
childhood
– source confusion

• Impossible memories (first year)
– hippocampus and frontal lobes are underdeveloped
– implanting impossible memories
• good eye coordination comes from mobile
• hypnotized and age regressed
• “guided mnemonic restructuring”
– imagination
• false memory for
– 46% of the hypnotic participants
– 56% of the guided participants

• need experience/knowledge/script Pedzek et al.
(1997)
– Catholic and Jewish participants
» 3 true memories
» 1 false Shabbat and communion
– no give-away words
– 29 Catholic participants
» 7 falsely recognized the Catholic event
» 1 falsely recognized the Jewish event
– 32 Jewish participants
» 3 falsely recognized the Jewish event
» 0 falsely recognized the Catholic event
All these results only mean that recovered
memories might be false

edspread
betrayal
donate
escapade
holly
iceberg
missile
numeral
outcast
painless
plumber
plural
priority
rancher
reversal
skeleton
trigger
trinket
waffle

Amnesia

• Bilateral damage to the medial temporal lobes
– Hippocampus
• memory for events and episodes
– Amygdala
• memory for emotions
– patient H.M. (1953)
– sources of damage
• Surgery, stroke, hypoxia, head injury
• Korsakoff’s syndrome, Alzheimer’s disease
• electroconvulsive therapy (ECT), benzodiazepines
– anterograde and retrograde amnesia
brain
damage
PAST PRESENT
retrograde
anterograde

• Anterograde amnesia
– Inability to form new long-term memories
• general knowledge and personal history are frozen
– recall, recognition, etc.
– new vocabulary (e.g. jacuzzi, granola, etc.)
• intact STM
• all modalities are affected
• Skill learning
– Implicit learning, but no explicit memory (ch. 6 235-238)
» procedural versus declarative
– motoric or perceptual (i.e., cannot learn chess)
» rotary pursuit
» mirror tracing
» mirror image reading
» new words (not meaning)
» long-term priming (stem and fragment completion)

ed__
bet__
don__
esc___
hol___
ice___
miss___
num___
out___
pain___
pl_m__r
p_ur_l
pr_o_ity
ra_c_er
re_e_sal
s_el_ton
t_ig_er
t_in_et
w_f_le

• Retrograde amnesia
– loss of old memories
• some retrograde amnesia is associated with
anterograde amnesia
– the time period just before the injury is forgotten
(temporally graded)
– depends upon abruptness
• Korsakoff’s and Alzheimer’s
–memory consolidation (Squire)
• Hippocampally dependent
– Memories are distributed throughout the cortex
– Hippocampus ties together the various components
– with time, the cortex can learn to do this on its own
– this process is called memory consolidation
• forgetting curves and amnesia

Improving Memory (ch. 8, 274-277)
• Terms
– Mnemosyne: Greek goddess
– Mnemonic: memory aid
– Mnemonist: skilled memory
• variables that strengthen LTM
– the spacing effect
• encoding specificity
• state-dependent learning
– semantic encoding (a.k.a. deep encoding)
• self-referential works best
– rare or unusual
• violate the schema
– strong emotions (flashbulb memories)
• physiological arousal
– visual imagery

• Mnemonic Techniques
– relate new to old information
• chunking
– N F L C B S I R A M T V
• the method of loci
– ordered landmarks and imagery

• technique of interacting images
– in paired-associate learning (e.g., goat-pipe, tablehorse,...),
forming a joint image
• the pegword method
– one is a bun
– two is a shoe
– three is a tree
– four is a door
– ....
• recoding of material (similar to chunking)
– HOMES = Huron, Ontario, Michigan, Erie, Superior