Transcript lectures19-20-LTMx

Long Term Memory
CLPS0020: Introduction to Cognitive
Science
Professor Dave Sobel
Fall 2012
Review of STM
• Capacity: 7 +/- 2 “Chunks”
• Representation: Auditory, but also organized
temporally:
– For the most part: FIFO (First in - First Out)
– Chunks suggests an additional conceptual representation
• Maintenance:
– Words remembered better than nonwords
• Suggests more of a conceptual representation
– Slow presentation rate, overall number of items increases
• More encoding and more elaboration
• Is elaboration effortful or automatic?
Levels of Processing
• Elaborative Rehearsal is a way of chunking
information
• Question 1: Does memory improve the more you
elaborate on information?
• Question 2: Does it matter if you know you are going
to have to remember the information?
– i.e., Automatic or Effortful
Craik & Lockhart (1972)
• Present subjects with one of three tasks
– Capital Letters: Press a button if each word has a
capital letter in it (Shallow Processing)
– Rhyming: Generate a word that rhymes with this one
(Medium Processing)
– Antonyms: Generate an antonym for this word (Deep
Processing)
• Then present a surprise recall memory test for the
words they saw
– The deeper the processing, the better subjects are at
remembering the words
Discussion
• Question 1: Does level of elaboration affect
memory?
– Answer: Yes!!!
• Question 2: If you know it’s a memory test, will that
help?
– Presenting the memory test as a surprise doesn’t matter.
– Same result if subjects know it’s a memory test (increased
performance overall, though)
How does information get into LTM?
• Some definitions first:
• Encoding: The process by which information gets
into long term memory
• Retrieval: The process by which we bring
information out of LTM
• Forgetting: The process by which we lose
information in LTM
Today: Encoding
• How does encoding work?
• Encoding Specificity (Tulving, 1983)
– Information is encoded in a context
– Change that context and you affect memory
Example #1
• Shown a list of paired associates, but you are only
asked to remember the words in capital letters
– fruit - HAND
– vase - DUCK
– strawberry – JAM
• Then gave a recognition memory test, in which the
target word is presented in an alternative context
– fruit - HAND (Yes)
– lamp - FINGER (No)
– traffic - JAM (Yes) <-- This is the Target Stimulus
Tulving (1983) Results
• If context changes, subjects are worse than if context
stays the same, but there’s a different word
– When you encode strawberry – JAM, you are more
accurate at remembering jelly – JAM than traffic – JAM
• If context changes, subjects are worse than if there is
no context
– When you encode strawberry – JAM, you are more
accurate at remembering pebble – JAM than traffic – JAM
Example 2: State Dependent Learning
• Learn information in a brightly colored room
or in an ordinary colored room (Smith et al., 1978)
• Recall test in either the same/switched room
– Match Group and Mismatch group
• Recall is better in match groups than
mismatch groups
• Generalizes to many other contexts
Explicit vs. Implicit Memory
• State Dependent Learning suggests we are
encoding context as well as information
• Does this affect recall?
– Explicit Memory: Conscious recollection of past events typically
revealed through recognition or recall tasks.
– Memory with awareness
• Contrast with Implicit memory?
– Memory without awareness (McDougall, 1924)
– Influence of past events on behavior without our consciousness
– “Subliminal”
Explicit vs. Implicit Memory
• What is Explicit memory?
– Conscious recollection of past events typically revealed
through recognition or recall tasks.
– Memory with awareness
• What is Implicit memory?
– Memory without awareness (McDougall, 1924)
– Influence of past events on behavior without our
consciousness
– “Subliminal”
Example of Implicit Memory: Stem
Completion ( Warrington & Weiskrantz, 1970)
• Two Phases
• Phase 1:
• Asked to memorize a list of 50 words. Among
those words, is a target word (e.g., “class”)
• Give an explicit memory task: Do they remember
the word “class”?
• Only include subjects who do not
Stem Completion Phase 2
• Generate any word that starts with the first three letters
CLA____
• ~50% of subjects generate “class”.
• When asked again, most subjects will not recognize that
“class” was on the list.
– No awareness that they were exposed to “class”
previously
• Control: Take a second group of subjects, and just give
them phase 2.
– Very few (< 10%) generate “class” as the word
Conclusions
• Do we encode everything?
– Probably not consciously. But we often retain the
gist of information (e. g., War of the Ghosts)
• What role does the brain play in memory
– Amnesia (Next Lecture)
Types of Amnesia 1
• Soap Opera Amnesia
– Get hit on the head, don’t remember who you are
– So rare that I will claim it doesn’t exist!
• Retrograde Amnesia
– Head Trauma often results is some memory failure for
events leading up to trauma
– Often recovered, except ~500-1000ms prior to trauma
Anterograde Amnesia
• Inability to learn new information, but retain
information already learned
• Most information comes from study of patient
HM (Milner, 1966)
• HM had surgery which removed most of his
hippocampus and MTL to relieve epilepsy.
– Similar findings in patients with Korsakoff’s
Syndrome
Anterograde Amnesia (cont)
• Inability to acquire new semantic & episodic
information.
– Did not remember the doctors who see him every
day.
• Inability to encode new information into long
term memory. LTM prior to the surgery (1965)
is intact.
• Intact performance on working memory tasks
(e.g. digit span)
Encoding or retrieval?
• Is this a difficulty with encoding or retrieval?
• Marslen-Wilson & Teuber (1975) presented HM
with pictures of famous people.
– Some were famous before his surgery.
– Others only achieved fame afterwards
• HM did not recognize people who became famous after
his surgery, but did recognize those from before the
surgery
• If given a hint about people who became famous after
surgery, his performance improved
– Something was getting in
Implicit memory
• HM’s performance on implicit memory
tasks (like stem completion) is similar to
normals
• HM’s Procedural Memory is intact
– Skill learning was not impaired
– Mirror tracing task
Mirror Tracing Task
Conclusions about Amnesia
• Anterograde Amnesia involves a failure to
encode new information
– Explicit memory fails to get in
– Implicit memory system is functional
– Hippocampus and Medial Temporal Lobes
important to encoding
• Reflects the idea that implicit and explicit
memory are performed by separate memory
systems
More types of memory errors
• Amnesia is an extreme example of memory
errors
– But amnesia is pretty rare
• Bartlett’s “War of the Ghosts”
– Details of often lost, gist is retained, based on
schemas or existing knowledge.
• Remembering things that aren’t there
Brewer & Treyens (1981) Office
• Had subjects sit in a
professor’s office
• Asked what they
remember as being
in the office
– No books, but more
than 30% remember
seeing books
– Influenced by prior
knowledge (schemes)
Semantic Influences
• Deese (1959), Roediger & McDermott
(1995)
• Recall list of words
– bed, rest, awake, tired, dream, wake, snooze,
blanket, doze, slumber, snore, nap, peace,
yawn, drowsy
Semantic Influences (cont)
• Immediate or delayed recall: people will
remember “Sleep”, even though it was not
on the list
• Even with source warning, people make
this error (McDermott & Roediger, 1998)
What causes memory errors?
• Delay
– Things erode over time
• Retrieval Failure
– Stress or change in circumstances (recall statedependent learning)
– Repisodic blurring (Where did you park your car
today)
• If everything is familiar, you lose the usefulness of
familiarity (i.e. context)
– Misinformation (What if repeated information is
subtly different?)
Verbal Misinformation
• Verbal influences can affect memory for events in
general
• Verbal Misinformation (Loftus, 1974)
– Shown a videotape of a blue car hitting a red car.
– Then adults read a verbal description of events which
states that a green car hit a red car.
– Later subjects misremember car as green, not blue in
verbal report
– Paint Chips – subjects pick a blue/green color, not a
blue color
Updating or Social Pressure
• When misinformation happens, are memories
actually changed (updating)
– Loftus’s interpretation
• Or is this a response to external factors in the
experiment (like social pressure)?
Evidence against Updating (1)
• McCloskey & Zaragoza (1985)
• Standard “Misinformation” Paradigm: Subject witnesses
a theft. Before the thief leaves, he picks up a hammer.
• Later, subject reads a verbal report of the events, which
says that the thief picked up a wrench
– Note the misinformation (wrench for hammer)
• Later, given a forced-choice question: What did the thief
pick up? A hammer or a wrench?
– Most subjects will report wrench.
Evidence against updating (2)
• Do subjects encode hammer and overwrite it with
wrench? Or does hammer stay in memory?
• Same procedure: see hammer, told wrench in
written report
• Did you see the thief pick up a hammer or a
screwdriver (old vs. novel)
• If memory is updated, then subjects should be at
chance. If not, then subjects should go back to
original item
– Most subjects choose hammer. Evidence against
updating
General Conclusion
• Memory errors are critical in the real world
• Think about Eyewitness testimony
–
–
–
–
Highly unreliable
Prone to errors based on questions (Suggestibility)
Prone to errors based on experience (Misinformation)
Prone to errors based on recall (Verbal Overshadowing, Retrieval
Based Forgetting)
– Yet, often most compelling
• Worse: False Memories
– Misinformation effects suggest that whole episodic memories be
implanted?
– What happens when those episodes are unpleasant?
– How resistant are we to these false memories?
– How resistant are children?