Short-term memory
Properties of STM
• Capacity
• Duration/Forgetting
• Retrieval
• Code type
• Alternatives to Modal’s view of STM
Measuring STM Capacity: Digit-Span Task
• Instructions: Recall the digits in the order
presented.
• Span Defined – list length that produces
accurate performance on 50% of trials
Measuring STM Capacity: Digit-Span Task
numbers
• Digit Span Defined: # of digits accurately recalled 50% of the time
• Standard Span: 7±2 digits
• Modal Model Interpretation (Miller, 1956):
– STM Capacity: ≈ 7 “chunks”
Chunking Demo
CDHAOETGN
C D H
A O E
T G N
Chunking Demo
CDHAOETGN
C D H
A O E
T G N
Chunking Demo
CDHAOETGN
C D H
A O E
T G N
Chunking
• Chunking – “the process of combining information
so that it takes up as little as possible of the limited
space in STM”
• Chunking  capacity
• Why not a limitless STM?
– “Chunk chunked chunks?
• Required:
– Chunking scheme
– Time to apply scheme
Extraordinary Digit-Span: SF
• Materials:
– random digits
– auditory presentation
• Results:
– After 45 days of practice: span = 83
SF: Digit Span
How did he do it?
Chunk (and elaborated) groups of digits into
running times (or historical dates)
Duration & Forgetting in STM
• Brown-Peterson (1958/1959)Task:
Initial attempt to measure duration of STM
• Procedure:
– Hear target set: 3 letters
– count backwards for X s
– recall target set
• Manipulation – length of retention interval
• Assumption:
– Counting task eliminates/minimizes rehearsal
– Measure of the rate of decay
Brown-Peterson: Main Finding
• Initial interpretation:
information rapidly
decays from STM
• Note: w/ 0-delay, only
80% accuracy.
A Test of Decay Hypothesis
• Waugh & Norman (1965) – Decay or interference?
• Method:
– Auditory digit list (two speeds), followed by probe digit
– TASK: name the digit that followed the probe in the list
• Decay prediction:
– Recall better at faster presentation rate
• Interference prediction:
– Recall equal for both presentation rates
Slow fast
Slow fast
Waugh & Norman (1965)
Results:
• Recall  w/ # of
intervening items
– consistent w/ both
decay & interference
• Recall (more or less)
unaffected by
presentation rate
– Consistent mainly w/
interference
That was “retroactive” interference:
Items that follow the to-be-remembered information interfere with
remembering the earlier stuff:
That was “retroactive” interference:
Items that follow the to-be-remembered information interfere with
remembering the earlier stuff:
e.g. if you have a psych and an english test in one week, if you study psych,
then study english, all the english info will interfere with your memory for the
psyc info.
But there is a second kind of interference as well:”proactive”:
Information that was previously studied will ALSO interfere with subsequent
information you study.
Evidence for PI in Brown-Peterson Task
-Keppel & Underwood (1962): re-analyzed Brown & Peterson’s data.
Took a look at recall trial-by-trial, at various retention intervals
Notice: on trial 1, 100% accuracy at ALL RI’s, only after
3 trials do you see substantial drops, which is more
pronounced at longer RI’s
Wickens, Born & Allen (1963): release from PI by varying target stimuli
-first three trials, target consists of letters,
fourth is letters in control, numbers in
“changed category” group
Side track: This issue of decay versus interference can be
seen in other areas: e.g. Ebbinghaus’ serial position curve
Side track: This issue of decay versus interference can be
seen in other areas: e.g. Ebbinghaus’ serial position curve
What might a “decay-based” perspective say about primacy/recency?
(e.g. Atkinson & Shiffrin’s Multistore/modal model)
What would the “interference” camp have to say?
So who’s right?
So who’s right?
These days, more evidence of forgetting based on interference
than decay, although decay isn’t ruled out entirely, since it does
have supporting evidence that an interference hypothesis wouldn’t
anticipate (anterograde amnesics fail to show a primacy effect)
STM Retrieval/searching: 3 Possiblities
• Issue:
Retrieval
Models
Parallel
SelfTerminating
– How do we access
information in STM?
– Is Item X in STM?
Serial
Exhaustive
• Three possibilities:
– Parallel –
simultaneous access
to all items.
– Serial – consider 1
item at a time.
Testing between these alternatives:
The Sternberg Task
– Set Size: 1 to 6 letters (well within limit for accurate memory in most
people
– Probe Type:
• positive (in memory set)
• negative (not it set)
Competing Retrieval Model Predictions
Sternberg Task: Results
• RT  w/ set size
Implication: serial
• Negative = Positive
Implication: exhaustive
Codes of short-term memory
Codes of short-term memory
-semantic code
Codes of short-term memory
-semantic code
Wickens & Morisano (1972): release from PI by varying category of
studied items
Codes of short-term memory
-semantic code
Wickens & Morisano (1972): release from PI by varying category of
studied items
-first three trials, target consists of fruit, fourth is fruit in control,
other things in other conditions
-varied the similarity of the new target to the old category
Codes of short-term memory
-semantic code: Wickens & Morisano (1972)
-visual code: Shepard and associates
Shepard’s rotational stimuli
A.
A.
45
B.
B.
N/A
B.
B.
110
D.
D.
180
Codes of short-term memory
-semantic code: Wickens & Morisano (1972):
-visual code: Shepard and associates
Codes of short-term memory
-semantic code: Wickens & Morisano (1972):
-visual code: Shepard and associates
-other codes (e.g. Shand, 1982)
Working Memory: somewhat of an elaboration of the traditional STM description