Transcript Memory_Ch5_all - Arizona State University

Memory and Cognition
PSY 324
Topic 5: Short-Term & Working Memory
Dr. Ellen Campana
Arizona State University
Modal Model of
Memory
Modal Model of Memory
Atkinson & Shiffrin (1968)
Rehearsal
Input
Sensory
Memory
Shortterm
Memory
Longterm
Memory
Output


Three stages of memory
Input, Output, Rehearsal (a control process)
Modal Model

Structural Features of the Model
Sensory memory: initial stage, holds info for
seconds/fractions of seconds. Large capacity.
 Short-term Memory (STM): holds 5-7 items for
15-30 seconds. Control processes can extend this.
 Long-term Memory (LTM): holds a large amount
of information for years, even decades


Control processes: active memory strategies
controlled by individual (example: rehearsal)
Control Processes

Some control processes maintain info in STS
Rehearsal (repeat the items over and over)
 Chunking (make connections between items)
 Visualization


Some control processes affect transfer between
STS and LTS (storage and retrieval)
Memorization
 Recall


You can only process information in STS
Sensory Memory
Sensory Memory

Sensory memory is very short
Allows you to see the “trail” of a sparkler
 Allows you to see movies (flipbook, tachistoscope)


Auditory Sensory Memory is also called


Echo
Visual Sensory Memory is also called
Persistence of vision
 Iconic Memory / Visual Icon

Sperling (1960): Iconic Memory
XMLT
AFNB
CDZP
Whole report condition
X
F
D
Z C
Sperling (1960): Iconic Memory
XMLT
AFNB
CDZP
Partial report condition
X
L
M
T
Sperling (1960): Iconic Memory

What’s the point?
Sperling was studying visual sensory memory
 Before his study, people thought that visual sensory
memory could only hold 4-5 items (full report cond)
 The other conditions in his study showed that

It’s true that people can only report 4-5 items before
memory decays (or fades away)
 BUT sensory memory actually encodes the whole scene


Conclusion: Sensory Memory has a large
capacity, but fast decay
Sperling (1960): Iconic Memory
XMLT
AFNB
CDZP
Partial report delayed
condition
M
Sperling (1960): Iconic Memory

Summary of conditions

Whole report condition


Partial report condition


All 12 letters flash on/off -> 1s. delay -> report any
All 12 letters flash on/off -> auditory cue to row ->
report just that row
Partial report delayed condition

All 12 letters flash on/off -> 1s. Delay -> auditory cue to
row -> report just that row
Sperling (1960): Timing of Decay

What’s the point?

Sperling wanted to get a clearer picture of just how
fast sensory information decays


Stronger support of his hypothesis that sensory memory
has large capacity and fast decay
Conclusion: Within just 1 second, most of sensory
memory decays, leaving only what was moved to
STS via attention.
Moray, Bates & Barnett (1965)

Sternberg looked at visual memory, Moray,
Bates & Barnett were interested in echoic
memory (auditory sensory memory)
Same task for audio domain
 “four-eared listening”
 Similar effects (advantage for partial reporting)


Work after that showed

Echoic memory has larger capacity and slower decay
than visual iconic memory
Modalities of Sensory Memory

Modality: the “channel” (Broadbent) that
different inputs come in through


Auditory, visual, tactile, etc.
Sensory memory is modality specific
Saying “ba, ba, ba” while receiving auditory input
messes up echoic, but not iconic memory
 A visual mask messes up visual memory, but not
auditory memory


Mask: for control in experiments (as in demo)
Change Blindness
Change Blindness
Change Blindness
Sensory Memory Now


In Sternberg’s day, this evidence supported the
Modal Model (which has since been replaced)
Sensory memory is still important and seems to
be separate from other forms of memory


Still thought to have large capacity and fast decay
Thought to be important for
Collecting input
 Holding input during initial processing
 Filling in “blanks” (movies, static, etc.)

Short-Term Memory
(STM)
Modal Model of Memory
Atkinson & Shiffrin (1968)
Rehearsal
Input
Sensory
Memory
Shortterm
Memory
Longterm
Memory
Output


Three stages of memory
Input, Output, Rehearsal (a control process)
Short-Term Memory

Short-term memory allows you to:
Understand sentences
 Do arithmetic
 Dial a phone number
 Navigate from one place to another
 Know where we are and what we’re doing right now


Memory for current tasks, last few minutes
Momento: main character had STM, not LTM
 Clive Wearing: Real-world case in book

Issues with STM

Just as with Sensory Memory, two important
issues are
Duration (how long things stay in memory)
 Capacity (how many things fit in memory at a time)


Studying Short-Term-Memory is complicated
because people use control processes a lot
Rehearsal seems to extend duration
 Chunking seems to extend capacity

Duration of Short-term Memory

Brown (1958) / Peterson & Peterson (1959)





Same studies at the same time, same results
Step 1: three letters + one number given
Step 2: count backward from number
Step 3: 3-18 s. delay (while counting backward)
Step 4: recall three letters
Duration of Short-term Memory

Brown (1958) / Peterson & Peterson (1959)

Same studies at the same time, same results
Percent
Recalled
FIRST TRIAL
ONLY
3
18
Delay
Duration of Short-term Memory

Brown (1958) / Peterson & Peterson (1959)

Same studies at the same time, same results
THIRD TRIAL
ONLY
Percent
Recalled
3
18
Delay
Duration of Short-term Memory

Brown (1958) / Peterson & Peterson (1959)

Same studies at the same time, same results
Percent
Recalled
MANY
TRIALS
LATER
3
18
Delay
Duration of Short-term Memory

Brown (1958) / Peterson & Peterson (1959)

Same studies at the same time, same results
AVERAGE
OVER ALL
TRIALS
Percent
Recalled
3
18
Delay
Duration of Short-term Memory

The studies by Brown and Peterson & Peterson
show that the percentage of letters recalled
decreases with longer delays, BUT this pattern
interacts with where in the series of trials the
individual trail occurs

Recall of letters after long delays decreases as the
series of trials gets longer
Duration of Short-term Memory

What’s the point?
Peterson & Peterson / Brown were interested in
decay of short term memory
 It turns out, their studies demonstrate that another
type of forgetting that happens in STM: proactive
interference



What is already in STM affects ability to add new things
Larger point is that forgetting in STM occurs
through both decay and interference (proactive
and other types too) – effective duration is 15-20s.
Capacity of Short-term Memory


Capacities can vary from person-to-person,
measured by digit span
Get out some paper and something to write
with, we’re going to calculate your digit span
Capacity of Short-term Memory

Directions: Make sure you are running the next
slides in presentation mode. You will see a list of
single-digit numbers. Remember them. When
you see “go” (but not before), write them down
from memory, in order. When you are done
writing, click to get the next set of digits.
Capacity of Short-term Memory
2149
39678
649784
7382015
84261432
482392807
5852981637
GO!
GO!
GO!
GO!
GO!
GO!
GO!
Capacity of Short-term Memory
2149
39678
649784
7382015
84261432
482392807
5852981637
How many digits were in
the longest row that you
got completely right?
That’s your digit span.
Capacity of Short-term Memory
“My problem is that I have been persecuted by an
integer. For seven years this number has
followed me around, has intruded in my most
private data, and has assaulted me from the
pages of our most public journals. This number
assumes a variety of disguises, being sometimes
a little larger and sometimes a little smaller than
usual, but never changing so much as to be
unrecognizable…..
Capacity of Short-term Memory
…. The persistence with which this number
plagues me is far more than a random accident.
There is, to quote a famous senator, a design
behind it, some pattern governing its
appearances. Either there really is something
unusual about this number or else I am suffering
from delusions of persecution.”
George Miller (1956)
The Magical Number 7 (plus or minus 2)
Capacity of Short-term Memory

Miller (1956): People can remember 7±2 ….
Digits
 Words
 Numbers (with multiple digits)
 Phrases


We can remember more if it’s organized

Chunking is combining smaller units into larger
meaningful units, to improve capacity
Chunking


Chunking involves using Long-term memories
to organize information in Short-term memory
Ericcson and coworkers (1980)

College student had digit-span of 79 after training


Chunked digits into meaningful times for running, a sport
he was familiar with
Chase and Simon (1973)

Chess players chunk information based on
meaningful points within a game of chess
Chunking

Chase & Simon (1973)
Meaningful
Arrangements
Random
Arrangements
Master Beginner
Master Beginner
Correct
Piece
Placements
Chunking & Information Coding

What’s the point of all these chunking studies?


Capacity is related to how information is represented
Recall our last discussion of how information is
represented, during “Cognition and the Brain”
Specificity coding vs. Distributed coding
 Dealt with how information is represented by
neurons’ firing rates
 This is called a physiological approach to coding


We can also take a mental approach to coding
Information Coding

Mental approach to coding
More abstract than physiological approach
 Deals with how things are represented in the mind /
thoughts


Three Types of Coding
Auditory Coding – represented as a sound
 Visual Coding – represented as an image
 Semantic Coding –represented through meaning

Auditory Coding

Conrad (1964)
Participants saw target letters (quickly flashed)
 Then they wrote them down
 Mistakes were made

Not likely to replace with something that looked like the
target (E for F)
 Likely to replace with something that SOUNDED like the
target (E for B)


Suggests that letters are represented by sound
information (auditory coding)
Semantic Coding

Wickens and Coworkers (1976)

Participants divided into groups

Groups heard lists with different meanings (fruits,
professions, meats, etc.)
Proactive Iterferecen for same list-TYPE repeated
 Category switch caused release from proactive
interference



Effect was larger for categories that were less similar
Evidence for semantic (meaning) coding
Short-term Memory Today

The Modal Model had a nice clean vision of
Short-term Memory
All-purpose store with 15-20s duration and capacity
of 7±2
 Simply holds information
 How information is coded affects how much
information fits in STM, but not much else


This view of STM turned out to be too simple,
so it has been replaced with working memory
Working Memory
Modal Model of Memory
Atkinson & Shiffrin (1968)
Rehearsal
Input
Sensory
Memory
Shortterm
Memory
Output
Longterm
Memory
Working Memory
Baddeley & Hitch (1974)
Input
Sensory
Memory
Central
Executive
Phonologal Loop
Longterm
Knowledge
Visuospatial Sketchpad
Comparing Memory Models

Short-term Memory (Attkinson & Shiffrin)
Single component for all types of info
 Mainly used for holding information for a short time


Working Memory (Baddeley & Hitch)

Three components:
Central Executive
 Visuospatial Sketchpad
 Phonological Loop


Used for manipulation of information during
complex cognition
Components of Working Memory

Phonological Loop

Holds verbal and auditory information


Visuospatial Sketch Pad


Coding or source can determine whether it’s
verbal/auditory information or not
Holds visual and spatial information
Central Executive

Pulls info from long-term memory, coordinates
other components, directs and maintains attention…
Phonological Loop
A component of working memory
Phonological Loop


The phonological loop holds verbal and auditory
information (for longer than the echo)
Sources of experimental support for a
component specialized for Language
Phonological similarity effect
 Word-length effect
 Articulatory suppression

Phonological Similarity Effect

The basic effect: words that sound similar are
confused by people
We saw that earlier today when we discussed
Auditory Coding: Condrad’s study with letters
 Another example is the Coglab “Phonological
Similarity” (which you can get extra credit for doing)

Phonological Similarity Effect

Experiment design
Half of the time the letters were similar and half of
the time they weren’t
 Half of the time you had to speak (recite numbers 14 in order) and half of the time you were to be quiet



Speaking in this experiment is called articulatory
suppression (which we’ll come back to)
The two factors were independent
Phonological Similarity Effect
The U-shape
doesn’t matter for
now
How do similar
and dissimilar
compare to each
other for the quiet
trials?
How does this
support the
phonological
similarity effect?
Phonological Similarity Effect

What was the point of the Phonological
Similarity Effect experiment?

Demonstrated the phonological similarity effect
(people confuse letters that sound similar)


Key point: even though information was presented
visually, people converted it to auditory
As we’ll see later, it also showed that the
phonological loop is necessary for the conversion
(not just holding info)
Phonological Loop


The phonological loop holds verbal and auditory
information (for longer than the echo)
Sources of experimental support for a
component specialized for Language
Phonological similarity effect
 Word-length effect
 Articulatory suppression

Word-Length Effect

The basic effect: When memorizing words, you
can remember fewer words if the words are long

Here’s another demo…. Just do what you did
earlier for the digit span test. When you see the
words try to remember them. Then, when you
see “go” (but not before) write down the words
you can remember. It can be any order this time.
Then click to go on.
Word-Length Effect
alcohol, property, amplifier, officer, gallery,
beast, bronze, wife, golf, inn, limp, dirt, star
mosquito, orchestra, bricklayer
GO!
GO!
Word-Length Effect

The basic effect: When memorizing words, you
can remember fewer words if the words are long
That was a demo of a real experiment …. People
remembered more of the short words than long
words (Baddeley & Coworkers, 1984)
 American children have a longer digit span than
Welsh children (Ellis & Hennelly, 1980)



Because welsh numbers take longer to pronounce!
Number of words you can say in 1.5-2.0 seconds is
likely to be your digit span
Phonological Loop


The phonological loop holds verbal and auditory
information (for longer than the echo)
Sources of experimental support for a
component specialized for Language
Phonological similarity effect
 Word-length effect
 Articulatory suppression

Articulatory Suppression


The basic finding: if you speak while
memorizing (which keeps the phonological loop
busy) you get worse at remembering, AND the
other two effects disappear
Coglab “Phonological Similarity Effect”
illustrates both

Remember: You can get extra credit for doing it
Articulatory Suppression
Which is less
accurate on average,
quiet or suppression
(circles or squares) ?
Is the phonological
similarity effect
(difference btwn
black & white)
stronger for quiet or
suppression (circles
or squares)?
Articulatory Suppression

The basic finding: if you speak while
memorizing (which keeps the phonological loop
busy) you get worse at remembering, AND the
other two effects disappear
Performance worse in suppression condition
 Phonological effect weaker in suppression condition



Similar findings regarding the word length effect
What is going on in these situations? Why does
this support the concept of a phonological loop?
Visuospatial Sketch Pad
Another component of working memory
Working Memory
Baddeley & Hitch (1974)
Input
Sensory
Memory
Central
Executive
Phonologal Loop
Longterm
Knowledge
Visuospatial Sketchpad
Visuospatial Sketch Pad


The visuospatial sketch pad holds visual and
spatial information
Experiments we’ll talk about show just that
visual and spatial information is separate from
phonological loop
Visuospatial Sketch Pad

Brooks (1968) – the sentence experiment
Memorize a sentence
 Indicate whether each word is / is not a noun

Condition 1: indicate by speaking
 Condition 2: indicate by pointing

Visuospatial Sketch Pad
Y
Y
N
N
N
Y
N
Y
Y
N
N
Y
N
Y
Y
Y
N
N
Visuospatial Sketch Pad

Brooks (1968) – the sentence experiment
Memorize a sentence
 Indicate whether each word is / is not a noun

Condition 1: indicate by speaking
 Condition 2: indicate by pointing



Results: pointing was easier than speaking for the
participants
Explanation: Phonological loop was busy
processing the sentence, but sketch pad was free
Visuospatial Sketch Pad

Brooks (1968) – the “F” demo
Memorize a shape (in this case an F)
 Indicate whether each corner is an “inside corner” or
an “outside corner”

Condition 1: Indicate by speaking
 Condition 2: Indicate by pointing

Visuospatial Sketch Pad
Y
Y
N
N
N
Y
N
Y
Y
N
N
Y
N
Y
Y
Y
N
N
Visuospatial Sketch Pad

Brooks (1968) – the “F” demo
Memorize a shape (in this case an F)
 Indicate whether each corner is an “inside corner” or
an “outside corner”

Condition 1: Indicate by speaking
 Condition 2: Indicate by pointing



Results: Speaking is easier than pointing (the
OPPOSITE of what happened before)
Explanation: Sketch Pad was busy with image,
but phonological loop was free
Visuospatial Sketch Pad

What is the point of these studies?

Tasks are easier when the information being held in
mind and the operation being performed on it
involve different types of short-term memory
Verbal / Phonological
 Visual / Spatial


That means that the two types of short-term
memory are somewhat independent

At the least, separate capacities
Central Executive
Another component of working memory
Central Executive

The Central Executive does the “work” of
working memory
Coordinating sketchpad and phonological loop
 Performing calculations
 Directing and maintaining attention



A lot of what we learned about in the attention topic is
part of what the central executive does
Sample source of evidence: central executive’s
ability to suppress is correlated with memory
Central Executive

Gazzaley and coworkers (2005)

Compared two versions of the task
“face-relevant”: Remember faces, ignore scenes (test: faces)
 “passive”: Just watch pictures (test: arrow right/left)

Central Executive
Central Executive

Gazzaley and coworkers (2005)

Compared two versions of the task
“face-relevant”: Remember faces, ignore scenes (test: faces)
 “passive”: Just watch pictures (test: arrow right/left)


Measures:
Accuracy at remembering faces
 Brain activity in areas used for perceiving scenes




Good suppressors: less activity in scene areas (good at ignoring)
Poor suppressors: more activity in scene areas (poor at ignoring)
Results: good suppressors remembered more faces
Back to the big
picture…
Remember the
other
components!
Working Memory
Baddeley (2000)
Episodic Buffer
Input
Sensory
Memory
Central
Executive
Phonologal Loop
Longterm
Knowledge
Visuospatial Sketchpad
Working Memory Now


The model successfully explains a lot of data
Still a useful model that is used by many


There is a newer one (Cowan) but details are beyond
this class
Still changing, though….
Baddeley was frustrated that certain things didn’t
seem to “fit” (effect sizes larger or smaller, etc)
 Episodic buffer has been added as a 5th component

Working Memory
Baddeley (2000)
Episodic Buffer
Input
Sensory
Memory
Central
Executive
Phonologal Loop
Longterm
Knowledge
Visuospatial Sketchpad
What’s this Episodic Buffer?

The episodic buffer is a “backup” that talks to
the central executive and long-term memory
Greater duration than loop & sketch pad
 Greater capacity than loop & sketch pad


Very vague, still needs to be tested

The point is that models are constantly being refined
and modified to account for new results
Working Memory and
the Brain
Working Memory and the Brain

Prefrontal cortex involved in working memory
Gets inputs from the sensory areas
 Gets inputs from areas involved in action
 Connected to areas involved in long-term memory

Working Memory and the Brain
Prefrontal Cortex
Working Memory and the Brain

Prefrontal cortex involved in working memory
Gets inputs from the sensory areas
 Gets inputs from areas involved in action
 Connected to areas involved in long-term memory


Physiological evidence
Delayed-response task in monkeys
 Single-cell recording in monkeys


Brain imaging evidence
Working Memory and the Brain

Physiological evidence based on similirities
between monkey and human brain
Delayed-response: Monkeys can remember a
location over a delay. When monkeys have PFC
removed, they can’t do that very well any more.
 Funahashi and coworkers (1989) Single-cell
recording: When monkeys have to remember a
location over time, cells in the PFC remain active

Working Memory and the Brain

Brain imaging studies with humans: PFC is
active when we use working memory

BUT it isn’t the only area that’s active!
Other areas in the frontal lobe
 Areas in the parietal lobe
 Areas in the cerebellum


Activity occurs in many areas simultaneously
Working Memory and the Brain