Sensory and Working Memory
Transient Memory: Transient (short-lived)
Limited capacity
Sensory Memory and Memory
Working Memory
Transient Long-term
Sensory Working Declarative Nondeclarative (Implicit)
Visual Articulatory Visuo-
(Explicit) Skills & Priming & Simple Nonassociative
loop spatial Habits Perceptual Classical Learning
Echoic Facts & Events Learning Conditioning
sketch pad
Introduction to Cognitive Psychology Haptic
Central Executive
Pick a card, any card Psychic Professor Stark?
Concentrate - remember your card ...
SHAZAM! Your card is gone...
Atkinson & Shiffrin Jevons (1871)
Rehearsal
Input from
environment Sensory Short-term Long-term
Store Store Store
Haptic
Echoic Visual STM LTM
Transfer
Decay Displacement Interference
STM vs. Primary Memory vs. Working Memory
Sperling Task Partial Report
Full Report P Q F X
K B E M G D L J
S D X O A W K O
A I L J B C P T
Z N W Y
Partial Report Sperling’s Data
B S E O
A Z X F
R L Q G
I P D N
Sensory Memory
Visual Buffer
R N P Z Echoic Buffer
O D X G Haptic Buffer
Large Capacity
A J H L Constantly being overwritten by new information
I B K V Rapid decay
Little, but probably some “higher level” information
Haber (1983) Icon Formation and Decay
Iconic memory is a laboratory curiosity What if icons were formed and started to decay
It is not and cannot be used in normal perception on onset rather than offset?
Each time we saccade, the iconic memory is Haber’s argument would go away
formed and starts to decay, but our eyes land on
another area and that visual stimulus overwrites
the old one. ...
Ergo, iconic memory can’t be used
10-100ms
Di Lollo’s Results Echoic Sensory Memory
“What? Oh, yea, I’ll be there.”
Shadow two streams that actually are the same
audio tracks. Vary the offset between the two ears.
0-2 sec, participants hear both as one. Longer, and
don’t notice it’s the same stream (Triesman, 1964)
High/low tone detection with random tones
appearing after the target. If > 250 ms following
target, no effect. (Massaro, 1970)
“Short Term Memory” Capacity
W
Q
H
D
C
U
K
R
V
LSI
Generic term: Primary Memory
Specific models: STM and Working Memory H R S Q W I D C V K L U
All posit active maintenance of information
(representations / patterns of activity) to span
relatively short delays (seconds, not minutes).
Separate process from sensory memory or long-
term memory.
Serial Position Effect Capacity 2
O
N
H
K
PSJI
JOHNSHOPKINS
Organizing information into meaningful subgroups
(chunks) aids memory
STM’s capacity ~4-6 “chunks” not items
What exactly is a chunk?
Serial Position
Postman & Phillips (1965)
STM vs. LTM: Evidence Amnesic vs. Control: STM
Primacy vs. Recency effects 1.2
Qualitative differences 1
Control
STM:Very limited, but rapid input and retrieval .8
d’
LTM:Virtually limitless, but slow input and .6
Amnesic
retrieval .4
Neuropsychological data .2
H.M. and other amnesic patients 0
0-2 6-10 25-40
Animal Models of amnesia Delay (seconds)
Buffalo et al. (1998)
Delayed Nonmatch to Sample Monkeys Rats
Nonmatch test
e lay
D
Sample
Zola et al. (2000) Clark et al., 2000
Recap so far... Working Memory
Sensory memory separate from Primary / Short Problems forming with STM model
Term / Working memory separate from Long Term Can seem to get to LTM w/o STM
memory.
Seems to be more than one kind of “short term”
Sensory:Very short duration (decay starts
Does more than rehearse
immediately), easily overwritten Iconic, Echoic, and
Haptic stores. High capacity that is largely sensory- Re-think the problem (Baddeley)
based. Central
Primary: Longer duration, but based on active Executive
maintenance of information. Smaller capacity, Phonological Visuo-Spatial
clearly linked to semantic knowledge. Loop Sketchpad
Digit Task Phonological Loop
8572681 Phonological Store (Inner Ear)
Did you hear yourself repeating the digits over and over? ~2 s of information
Input from environment (via Central
Keep silently saying “Blah blah blah” Executive) or via articulatory control
process.
6350469
Articulatory Control (Inner voice)
Was that tougher?
Literally, talking to yourself
Did you interject “Blah” between digits?
Why? Actual articulatory codes (subvocal speech)
Used to refresh the phonological store.
Predictions 4.5
Span
Recall
2.25
10
Span
250
Articulation rate (ms/digit)
9.5
# of words read / second
Reading
Rate Articulation 270
# of words recalled
4 2 9 length
# Digits recalled
Since it only lasts 2s, people who talk faster 8.5
290
should have larger capacity 3.5 1.75 8 310
7.5 330
Capacity should be smaller for longer words or, 7
350
more precisely, words that take longer to say 3 1.5
6.5
370
Should confuse similar-sounding words or 2.5 1.25
6
5.5 390
letters (e.g., “cap” to “cat”, BGTVCD) 1 2 3 4
# of syllables
5
Chinese English Welch
Should be able to interfere with it if you
Also, digit span shorter for words with long
interfere with articulation (blah, blah, blah...)
vowel sounds than short vowel sounds
Baddeley (1986)
Visuo-Spatial Sketchpad Dog Sustained activity in the
CAT
absence of external stimuli
Separate system from phonological loop.
Stores “images” and spatial information Human
Rhesus
100
Able to manipulate the rote
Monkey
90
Correct words (%)
images visual
Record
80
70
60
50
None Visual Speech
Interference type
Quinn & McConnell, 1996
Central Executive Split the Visuo-Spatial Sketchpad?
Coordinate and direct phonological loop and Visual Perception: Split of What/Where (or How)
visuo-spatial sketchpad. Maintained into frontal lobe
Control of encoding and retrieval processes.
Manipulate information in subsystems.
Responsible for strategy selection, planning and
attentional control.
Interactions with long-term memory.
These “big” roles makes it tough to study
Sala et al. (2003)
Location
+
Instruction Preparation
3 sec ITI - 3 sec +
Memory Set Memory +
3 sec Delay Test
1.Location (1 sec each) 9 sec 3 sec Response
ITI
2.Face Identity 3 sec
3.House Identity
4.Nothing WHAT vs. WHERE
Houses
Current Status
Looks pretty good.
May need to add a few bits
Episodic Buffer?
Split Visuo-Spatial sketchpad into parts?
(object identity and location/spatial)
What are the fundamental processes that the
Central Executive does? How do they work?
