Transcript Cognitive Psychology - University College Dublin

Cognitive Psychology
Lecture 2: Memory
September 2008
John Toner
Overview of Lecture
• Memory (definitions and examples)
• Remembering
• The Multi-Store Model
• Working Memory
What is Memory?
Information about the past…
Details about your life…
–
–
–
–
–
Identity
Goals
Faces
Current conversation
This slide
Some expert opinion
“Memory is the process of maintaining information over time.”
-- Matlin (2005)
“Memory is the means by which we draw on our past experiences
in order to use this information in the present.’
-- Sternberg (1999)
Importance of current circumstances
“… memory is also the basis for the sense of who we are that each
of us carries.”
-- Gleitman et al (2004)
Why is our sense of self so robust?
Most people are certain that they are the
same person over days/years
Things which involve memory
(Illustrating the vast amount of categories and types of information
we store!)
What is the capital of Italy?
How do I ride a bicycle?
Where was the last barbecue I was at?
How do I get from Clonskeagh to Clondalkin?
What song comes next on the album?
What does that smell remind me of?
Basic process of remembering
• Encoding
• Storage
• Retrieval
Encoding
Refers to the way in which information comes to
be represented
Hamlet: “
The spirit that I have seen
May be a devil. And the devil hath power
T’assume a pleasing shape…”
Encoding
Hamlet: “
The spirit that I have seen
May be a devil. And the devil hath power
T’assume a pleasing shape…”
• Encode the rhythm/acoustic code
• Visually encode the quote
• Semantically encode the quote
Storage
We learn many quotes from the play and
do not rehearse them continually. We
may not think of them for hours/days but
we can still recall them
Therefore they must be stored somehow
Retrieval
Question: Show evidence that Hamlet
doubted his father’s ghost
Retrieval involves finding and accessing
stored memories and processing them
so that they are in a form that is useful
in the given context
The Multi Store Model
We know that the actions of the memory
system involve encoding, storage and
retrieval.
But what does the system itself consist
of?
The Multi Store Model is one theory
It looks like this…
The Multi Store Model
Waugh & Norman (1965)
The Multi Store Model
Atkinson & Schifrin (1968)
The Multi Store Model
Features of the multi-store model
• The multi-store model proposes three distinct parts to memory.
Sensory stores, a short term store and a long term store.
• In the sensory stores, information and knowledge that comes to
us from the senses is stored momentarily. After processing,
some of this information is sent on to the short term store.
• Some of the information in the short term store is then passed
on to the long term store.
• The multi store model posits that the long term storage of
information often depends on rehearsal, with a direct
relationship between the amount of rehearsal in the short term
store and the strength of the memory in the long term store.
The Multi Store Model
Sensory Stores: The persistence of sensory information for a
moment after its perception. The information is modality specific.
Vision:
The Iconic Store
Research by Sperling (1960)
S
T
Q
H
P
C
F
W
X
Z
N
L
N
C
R
V
B
G
K
FJ
T
R
M
L
Y
JM
The Multi Store Model
• Participants in this task could usually report only 4 or 5 of the 12
letters. Sperling’s research also indicated that information in
iconic storage usually decays within less than half a second of
exposure to the stimulus.
• Problem with the experiment: memories of letters decayed
before participants could report them.
• Averback & Coriell (1961) presented two rows of 8 random
letters. A small mark appeared beside one letter’s position just
after it appeared. Participants could recall that letter with about
75% accuracy, suggesting that participants had information
about 12 of the 16 letters after just 50 milliseconds exposure to
them.
• This experiment required only one item at a time to be
remembered, which meant that information wasn't lost during
the time it took participants to report what they saw.
The Multi Store Model
Sensory Stores: The persistence of sensory information for a
moment after its perception. The information is modality specific.
Audition:
The Echoic Store
E.g. Reading & listening
Research by Triesman (1964)
Dichotic listening task
Temporal duration 2 seconds approx.
The Multi Store Model
The Short Term Store
• Atkinson and Schifrin proposed that information that had been
attended to in the sensory stores went into a short-term store
(also called short-term memory) where it had to be rehearsed
before it could go into relatively permanent long term memory.
• Short term memory is of limited capacity and is fragile
(Remembering phone no. or directions)
• Capacity of 5-7 items
– Lists of letters/numbers used in experiments
– Serial or free recall
– Chunks are groups we know already
• Last 4 digits in number are same as a familiar one
• ALIBM, ALMBI
The Multi Store Model
The Short Term Store
Free-recall memory experiments
• In free recall experiments, participants can recall the items
presented to them in the list in any ordering.
• Typically, these experiments involve people being presented
with a series of words, usually consisting of about 15 to 30
words, read a rate of about 1 per second.
• Immediately after the task, people are asked to repeat back as
many words as they can.
• Online free recall task: http://tinyurl.com/j77cq
• Example of the task: Glasner and Cunitz (1966)
Free Recall as a Function of Serial Position
(Glanzer & Cunitz, 1966)
The Multi-Store Model
Results from Glanzer and Cunitz (1966)
• The graph of probability of recall against serial position is a Ushaped curve, known as the serial position curve.
• Participants had a higher probability of recall on items that were
near the start of the list (i.e. early serial position). This is called
the primacy effect.
• Participants had a higher probability of recall on items that were
near the end of the list (i.e. late serial position). This is called the
recency effect.
The Multi-Store Model
Explaining the recency effect with the MSM
• During the presentation of the list of words, people are trying to
keep these words in their short term memory.
• Short term memory is limited in size to about 7 chunks of
information.
• Therefore, as new words come into short-term memory, older
words must be bumped out of short term memory.
• At the end of the task, the only words that are left in short term
memory are the ones that have just been heard and therefore
have not been bumped out.
• This explains why people have better recall of the more recent
items.
The Multi Store Model
Explaining the primacy effect with the MSM
• According to the multi-store model, the transfer of information
from short term memory into long term memory depends on the
amount of attention and rehearsal that the information receives.
• Suppose the first word in the list is “doctor”. Short-term memory
can give the word the full attention of the rehearsal mechanism.
• Suppose that the second word is “sandwich”. Then short term
memory must give half its attention to the first word and half its
attention to the second word.
• When the third word is presented, short term memory will only
be able to assign it one third of the attention available.
• Words experienced earlier in the list will have more rehearsal,
and therefore will have a greater chance of making it to long
term memory, and therefore will be more likely to be recalled.
The Multi Store Model
• Glanzer and Cunitz (1966) also describes a modified version of
the free recall experiment.
• In the modified version, after the full list of items had been
presented to the experimental participants, but before they were
asked to recall them, the participants were asked to count
backwards from 10.
• Counting backwards from 10 involves the use of short term
memory, and therefore, according to the multi-store model, the
last few words in the list would be bumped out of short term
memory
• Therefore, the multi-store model would predict no recency effect
in this task.
• This was found to be the case (see graph on next slide)
The Multi Store Model
Free Recall as a Function of Serial Position
(Glanzer & Cunitz, 1966)
The Multi Store Model
Forgetting Over Time in Short Term Memory
(Peterson & Peterson, 1959)
100
Letter Recall (%)
90
80
70
60
50
40
30
20
10
0
0
5
10
Retention Interval
15
20
The Neurology of Memory
• Case reported by Milner (1959). A patient ‘H.M.’
suffered on average 10 epileptic seizures per day
• Evidence suggested that hippocampus was source of
epileptic disorders, so they decided to remove it
The Neurology of Memory
After surgery H.M.’s intellect and language abilities
remained intact.
His personality remained largely the same
However he suffered massive anterograde amnesia
(loss of all memories since surgery)
He suffered mild retrograde amnesia (events prior to
surgery)
The Bourne Identity v Memento
Evaluating the Multi Store Model
Strengths
On a conceptual level the model
makes sense. We can
understand the different
functions and recognise the
different capabilities of the
systems
Each store differs in a number
of ways, suggesting that they
are separate entities
• Temporal duration
• Capacity
• Forgetting mechanisms
• Brain injury evidence
Weaknesses
Evidence that the Short term
Memory is not unitary
(Warrington & Shallice, 1972)
Long term memory store holds
different types of memories
• Declarative
– Episodic
– Semantic
Remembering yesterday
2+2=4
• Non-Declaritive
– Implicit
Songs on radio
How to cycle a bike
Evaluating Multi Store Model
The MSM is a good place to start modeling,but
it does cause some problems
– Leads to the assumption that forgetting
rates are constant
– Assumes that rehearsal is the only route to
storage. Is this our experience..?
• We can be ready to receive certain info
• Person’s mood, wandering imagination
Levels of Processing theory
Proposed by Craik & Lockhart (1972)
Key Idea: The amount of cognitive
processing carried out on an item
determines its memorability, not the
amount of rehearsal.
Levels of Processing theory
The multi-store model makes two proposals regarding
the memorability of a stimulus:
• The memorability of a stimulus depends on its
being transferred from the short term store to the
long term store.
• The probability of a stimulus being transferred to
the long term store depends rehearsal.
Level of Processing theory emphasises
Deep, meaningful analysis of a stimulus produces
longer lasting memories than superficial, perceptual
analysis of a stimulus.
Levels of Processing theory
Memory is linked with attention
Rather than being mental items purposely
constructed and stored, memories are after
effects of the processing of a stimulus/event
Levels of Processing theory
Maintenance rehearsal: Repeating previous
analysis to prevent decay
Elaborative rehearsal: Deeper, more semantic
analysis.
Only elaborative analysis improves long
term memory
Levels of Processing theory
Craik and Tulving (1975, Experiment 1): Participants
answered questions about words presented to them.
The questions were designed to engage four different
levels of processing from shallow to deep:
Graphemic: Is the word in capital letters?
Phonetic: Does the word rhyme with ‘weight’?
Semantic Is the word a type of fish?
Elaborative semantic Would the word fit in the
sentence: The man peeled the _____ ?
Levels of Processing theory
Subsequent recognition of words was better for deeper
processing tasks
Levels of Processing theory
Eysenck (1980) argued that distinctive items will
be more readily remembered than typical
items
Carried out an experiment where participants
had to process items that were either
distinctive or non-distinctive and process
them at a semantic or non-semantic level
Levels of Processing theory
Eysenck (1980) conditions
• Non-semantic / Non-distinctive: Pronounce ‘lamb’
• Non-sem / Distinctive: Pronounce ‘comb’ phonetically
• Sem. / Non-distinct.: Processed in terms of meaning
• Sem. / Distinct.: Unusual way of processing in terms
of meaning
Levels of Processing theory
Eysenck (1980) results:
Non-semantic distinctive words were
remembered almost as well as semantic nondistinctive words, highlighting the importance
of an items uniqueness in its memorability
Levels of Processing theory
Strengths
Weaknesses
Gives importance to the
circumstances at time
of learning (e.g.
environment, attention,
previous experience)
Explains fMRI data
showing increased
activity for semantic
treatment of stimuli
(Gabriele et al. 1996)
It is hard to decide the
level of processing
being used by someone
in a given real world
situation (see quote on
next slide)
It does not offer any
explanation for implicit
learning (coming up!)
Levels of Processing theory
“ There is a danger of using retention test performance
to provide information about the depth of processing,
and then using the (alleged) depth of processing to
‘explain’ the retention test performance, a selfdefeating exercise in circularity” Eysenck (1978)
Contemporary views of memory
Schacter & Tulving (1994) drew on many
previous theories and research in an
attempt to get a more accurate model of
memory. We will mow examine their ‘5
memory systems’
Sensory Store
Short Term
Systems
Long Term
Systems
Echoic
Working
Memory
Semantic
Iconic
Episodic
Other modalities
Perceptual
Representation
Procedural
Working Memory
Originally introduced by Baddeley & Hitch
(1974) as an improved version of Short
Term Memory it has received much
attention and is credited with solving
many theoretical problems that existed
with the original, simple MSM version of
what lies between sensory stores and
long term memory
Working Memory
• Baddeley & Hitch wanted to emphasize
that temporary storage was not simply a
station on the way to long term memory.
• Working memory is the way we store
information while we are working with it,
or attending to it.
• These items interact with other cognitive
processes (attention, imagery. . .)
Working Memory
Peach
AppleMelon
Blueberry Mango
Banana
Orange
Lemon
Fig
Pineapple Grape
Tangerine Raspberry Plum
Working Memory
London
Oslo
Rome
Prague
Paris Dublin
Berlin
Athens
Warsaw
Stockholm Madrid
Vienna
Lisbon
Zagreb
Working Memory
Working Memory
Stores auditory information,
including words.
It consists of an
•Phonological store (Inner ear)
•Articulatory loop(Inner voice)
Working Memory
Stores visual information
Uses and manipulates visual
images to effectively ‘draw’
pictures in the mind.
How many windows are at the
front of your house?
Working Memory
Controls both the phonological loop and the
visuo-spatial sketchpad
Directs attention toward one stimulus or
another and determines which items will be
stored in working memory
Working Memory
Episodic buffer was added to the theory of
working memory much later (Baddeley, 2000)
• It takes the modality specific information from
the phonological loop and the visuo-spatial
sketchpad and combines them into a unitary
multi-dimensional representation
Working Memory
Working Memory
Evidence for the Phonological Loop
• The word length effect. People can recall more items in free
recall tasks that use short words than in tasks that use long
words. Baddeley et al (1975) showed that people can store as
many words as can be uttered in about 2 seconds. Performance
is related to duration of utterance, not number of syllables.
• The unattended speech effect. Irrelevant concurrent speech
inhibits recall. For example, Colle and Welsh (1976): serial recall
of visually presented numbers, accompanied by auditory
presentation of German language speech. This is explained by
assuming that all auditory material enters the phonological store
and therefore the spoken German interferes with the recall task.
Working Memory
Evidence for the Phonological Loop
• The phonological similarity effect. Immediate recall of
phonologically similar items is harder than with dissimilar items.
Explained by less distinction between similar phonemes in the
phonological store. Larsen et al. (2000) found that word lists like
“FEE, HE, KNEE, LEE, ME, SHE” were remembered 25% less
than “BAY, HOE, IT, ODD, SHY, UP”
• Baddeley et al. (1998) ‘The function of the phonological loop is
not to remember familiar words, but to learn new ones’.
Papagno et al. (1991) found that disrupting the phonological
loop (using articulatory suppression) in native Italian speakers
impeded learning Russian words but not pairs of Italian words
Working Memory
Evidence for the Visuo-Spatial Sketchpad
Baddeley et al. (1975) presented participants with details of the
location of digits within a matrix. These details were either
•
easily visualised (…in the square to the left of the centre is
a digit ‘1’. Above this is a ‘3’. . .)
•
not easily visualised (…in the square to the good of centre
is a digit ‘1’. To the quick this is a
‘3’. . .)
Tracking a moving light disrupted the visualised message but
not the other
Working Memory
Evidence for the Central Executive
Baddeley (1996) had participants randomly generate
digits by pressing a keypad
The task was done on its own or in combination with
reciting the alphabet, counting from 1, or alternating
numbers and letters (A 1 B 2 C 3…)
The randomness of the digits produced reduced for the
alternating condition
The interpretation was that the central executive is
needed to constantly switch retrieval plans and so
disrupts performance
Working Memory
Evidence for the Episodic Buffer
Papadopoulou & Wresinksi (1999) studied memory
span for Arabic numerals and digit words (1, one)
They found that participants used both verbal and visual
coding while performing the task
Information from both the Phon. loop and V-S
sketchpad were being stored together in working
memory
These and other findings suggested another
mechanism
Sensory Store
Short Term
Systems
Long Term
Systems
Echoic
Working
Memory
Semantic
Iconic
Episodic
Other modalities
Perceptual
Representation
Procedural
Reading. . .
• Eyesenk & Keane - Chapter 6
• Sternberg - Chapter 6
• Current focus articles:
Baddeley, A. (2003) Working Memory: Looking Back
and Looking Forward. Nature Reviews Neuroscience
(4) 10, 829-839