PSY 368 Human
Memory
Short Term Memory cont.
Working Memory
Reminders
• Exam 1 one week from today (Feb 15)
• Review sheet posted, linked to syllabus
• Experiment 1 report also due 1 week from today
Experiment 1 assignment
• Experiment 1 - Primacy and Recency Effects in Short-term
Memory (modified from Neath & Surprenant book pg. 61)
• Download from BB assignments page. Find 3 friends willing to participate
• Report (Due Wednesday Feb. 15): The results for all subjects will be reported in class.
Your assignment is to write a 2-3 page report that includes the following:
• brief description of the purpose and design of the experiment, including independent and
dependent variables
• brief description of the participants, materials, and procedure of the study, written in your own
words
• description of results and line graph of mean percentage recall by serial position
• discussion of conclusions that can be made from the results. Include answers to these questions:
•
How long does short-term memory appear to last?
•
How important is attention to retrieval from short-term memory?
•
What kinds of everyday tasks in life use short-term memory?
Structural Model
• Memory composed of storage structures that hold memories
for a period of time
• Sensory memory
• Short-term memory (STM)
• Long-term memory (LTM)
Structural Model
• Brief review from last time
• Capacity:
• Can hold 7 + or - 2 “chunks”(capacity)
• Chunking increases capacity of STM
• Encoding:
• Info mostly stored in auditory form (but some visual,
some semantic)
• Duration:
• Brown/Peterson task: decay, gone by 15-20 sec
• Keppel & Underwood; Waugh and Norman suggest
interference as better explanation
• Retrieval:
STM Duration
Wickens, Born, & Allen (1963)
• Changing the nature of the items to be remembered
reverses the decline in performance due to proactive
interference- release from proactive interference
•
Proactive interference: already learned info interferes with
new information
•
Retroactive interference: new information interferes with old
information (more on this in chapter 6)
• Two groups of subjects given 3 trials following the BrownPeterson task (letters) - Memory performance declined with
each trial
• Control group given a 4th trial using letters
• Experimental group switched to remembering digits
STM Duration
Wickens, Born, & Allen (1963)
• Changing the nature of the items to be remembered
reverses the decline in performance due to proactive
interference- release from proactive interference
• Experimental group,
but not control group,
performed perfectly;
they were released
from proactive
interference
7
STM Duration
Wickens (1970, 1972, 1976)
• Proactive interference occurring as a result of semantic coding in STM
• 5 groups of subjects given 3 trials of lists of 3 words each all from the
same category where all list contained names of fruit
•
Group 1 – names of fruit
•
Group 2 – vegetable names
•
Group 3 – flower names
•
Group 4 - names of meats
•
Group 5 – names of different
professions
• Then all groups given a 4 trial where all list contained names of fruit
8
STM Duration
Wickens (1970, 1972, 1976)
• Results:
1st trial all
groups about
90% correct
2nd trial more
words in same
category
all groups about
50%
3rd trial words
still in same
category
all groups 35–
45 %
4th trial , shift to fruit
category
professions 80%, meat 50%,
flowers 47%, vegetables 40%
and fruit 32%
STM Duration
Wickens (1970, 1972, 1976)
• Results:
• Conclusion:
• Interference rather than decay causes the drop in performance
• Information was coded using semantic information causing groups
to confuse current list with previous lists
STM Duration
• Duration:
• STM is short store of about 15-20 s
• Loss of information initially thought to be due
to decay.
• More recent work suggests interference more
likely reason
STM Retrieval
How do we get information out of STM?
• Retrieval from STM appears to operate by searching
STM contents one at a time (serial search)
• Sternberg (1966)
• Serial position recall curves: primacy and recency
STM Retrieval
Sternberg (1966)
• study short list of 1 to 6 items followed by test probe - must
decide if probe item was in list, measured time to make Y/N
response
•
Two important variables were involved
• The number of letters in each list
• The location of the letter in the memory probe – in the beginning,
middle, or end
STM Retrieval
Sternberg (1966)
• Parallel processing
• Simultaneous handling
of multiple operations
• Response times should
be the same, regardless
of the size of the set of
items, because all
comparisons would be
done at once
• Serial processing
• Operations being done one after another
• It should take longer to retrieve four
digits than to retrieve two digits
• Exhaustive serial processing – the
participant always checks the test digit
against all digits in the set, even if a match
were found partway through the list
• Self-terminating serial processing – the
participant would check the test digit
against only those digits needed to make a
response
STM Retrieval
Sternberg (1966)
• Results
• Response times increased
linearly with set size but were
the same regardless of serial
position
• It indicates that serial
exhaustive model seems to be
right
• Conclusion: people search all items
• Subjects take longer to respond by
probe (by 40 ms) when an
additional item is added to the list •
• Same results for probes that were
in the list and probes that were not
in STM when asked to retrieve an item
(happens very fast)
Automatic process - fast and efficient,
done for every item – doesn’t stop once
a match is found
STM Retrieval
Serial Position Curve: Free recall
• Primacy:
• better recall for items in
the beginning of the list
than those in the middle
• Recency:
• better recall for items at
the end of the list than
those in the middle
STM Retrieval
Free recall Curve
• Primacy:
• better recall for items in
the beginning of the list
than those in the middle
• Recency:
• better recall for items at
the end of the list than
those in the middle
From Murdock (1962)
• Models of STM propose that
• Primacy is due to more rehearsal for items at beginning of list - LTM
• Recency due to immediate dumping of items from STM
STM Retrieval
Free recall Curve vs. serial recall curves
From Klien et al.
(2005)
• Primacy:
• better recall for items in
the beginning of the list
than those in the middle
• Recency:
• better recall for items at
the end of the list than
those in the middle
• due to retrieval from
STM
• Models of STM propose that
• Primacy is due to more rehearsal for items at beginning of list - LTM
• Recency due to immediate dumping of items from STM
• Recency is stronger effect than primacy in free recall
Modal Model of STM: Summary
• Duration: STM is short store of about 20 s
• Capacity:
• Can hold 7 + or - 2 “chunks”(capacity)
• Chunking increases capacity of STM
• Encoding: Info mostly stored in auditory form
• Retrieval: Modal models suggest recency
effects mostly due to STM retrieval
Problems with the Modal Models
• When distractor task is done after every list item
preventing items from staying in STM, recency effect
still occurs
• Primacy effects have also been shown to disappear
when rehearsal is prevented
• Long-term recency effects
• Baddeley and Hitch (1977) Rugby Study
• Question: “Which teams have you played this past season?”
• Results:
• Recency effect: Recent games were recalled best.
• The total number of games played, not the amount of time gone by,
best predicted forgetting.
More Recent Models
• Current models focus more on processing (than structures)
and that include STM in some form (storage and manipulation
of information)
• There are multiple models that have been called “Working Memory,”
designed to account for similar sets of data
STM
LTM
• Models of WM assume:
• That STM is a storage
unit and that WM =
processes involved in
completing a task
WM
Vision
Touch
• That WM requires
Hearing
consciousness - includes
info we are currently
attending to
Rehearsal
Working Memory
• Today we will talk about 3 theories of WM
• Baddeley’s model of WM (Baddeley, 1986,
2000)
• Cowan’s activation model (Cowan, 1988,
1995)
• Nairne’s feature model (Nairne, 1988, 2001)
Baddeley’s Model
• Currently the most influential model
• Baddeley and Hitch (1974)
• Components of the model:
• Central executive controls the focus of attention
• Three subsystems under its control
• Visuospatial sketchpad
• Phonological loop
• Episodic buffer
Baddeley’s Model
Visual scribe
Articulatory
Control
Baddeley’s Model
Visual scribe
Articulatory
Control
Central Executive
• Thought to be an attentional controller, with two main
modes of operation (Norman & Shallice, 1986):
• A semi-automatic conflict-resolution system, based on existing
habits and requiring little attention.
• The supervisory attentional system (SAS), based on an
attentionally limited executive.
• Major Functions:
• Direct attention to the task at hand
• Divide attention between two or more tasks
Baddeley’s Model
Visual scribe
Articulatory
Control
Visuo-spatial Sketchpad
• Stores and manipulates visual and spatial information of
images
location
color
• May be separate parts for visual info (color) and spatial info
(orientation, location)
• Info enters through perception or internally generated
visual images
• Controls tasks like mental rotation and geographical
search of a visual or mental image
Visuo-spatial Sketchpad
Baddeley (1992)
• found that subjects’ memory for chess patterns was more
disrupted by a visual distractor task than an auditory one
Shepard and Feng (1972)
• Imagine folding the shapes below to create a solid, with the
shaded area as the base. Will the arrows meet head on?
• found that the time it
takes to answer the
question depends on
the number of folds
required
Baddeley’s Model
Visual scribe
Articulatory
Control
Episodic Buffer
• Back-up storage - allows for recall when other
systems are engaged with other tasks
• A storage system with a capacity of around 4 chunks of
information in a multidimensional code.
• Multiple dimensions permit links between the subsystems, as well
as with LTM & perception.
• Information is retrieved through conscious awareness –
consciousness pulls info together.
• Allows for the binding of previously unrelated concepts -disrupting the executive does not impair binding, so it may be
automatic/passive.
• Newest component of the model, not much research yet
Baddeley’s Model
Visual scribe
Articulatory
Control
Phonological Loop
• Two parts: Phonological Store (PS) and
Articulatory Control Process (ACP)
• PS - stores auditory info for 1-2 s and then it starts to decay
• ACP - recodes visual info into auditory code for storage and
controls rehearsal
• 4 Main Effects in Serial Recall Task to account for
•
•
•
•
Phonological similarity effect
Articulatory suppression effect
Irrelevant speech effect
Word length effect
Phonological Loop
• Demos
Listen to list,
recall words in
order
Rhinoceros
Zinc
Gorilla
Tuberculosis
Measles
Calcium
Uranium
Carbon
Hippopotamus
Mumps
Listen to list, Listen to list,
recall words in recall words in
order
order, while I
read the words
say ‘the’ aloud
Planet
Musician
Land
Property
Trumpet
House
Star
Comet
Orchestra
Moon
Bronze
Book
Magazine
Bike
Copper
Dress
Copier
Soda
Shoe
Rock
Listen to list,
recall words in
order, while I
read the words
say ‘the’ aloud
Block
Brick
Stick
Blue
Chew
Trick
Prick
Clue
Click
Blimp
Read list, recall
words in order,
whileRecall
I read the
words say ‘the’
aloud
Gold
Code
Bold
Hold
Told
Cold
Mode
Slowed
Hope
Goad
Phonological Loop
• Demos
Listen to list,
recall words in
order
Rhinoceros
Zinc
Gorilla
Tuberculosis
Measles
Calcium
Uranium
Carbon
Hippopotamus
Mumps
Listen to list,
Listen to list,
recall words in recall words in
order
order, while I
read the words
say ‘the’ aloud
Planet
Musician
Land
Property
Trumpet
House
Star
Comet
Orchestra
Moon
Bronze
Book
Magazine
Bike
Copper
Dress
Copier
Soda
Shoe
Rock
Listen to list,
recall words in
order, while I
read the words
say ‘the’ aloud
Block
Brick
Stick
Blue
Chew
Trick
Prick
Clue
Click
Blimp
Read list, recall
words in order,
while I read the
words say ‘the’
aloud
Gold
Code
Bold
Hold
Told
Cold
Mode
Slowed
Hope
Goad
Phonological Loop
• Phonological Similarity Effect
• Memory worse for items that
sound alike than those that look
alike or have similar meanings
• Visual items are recoded to
auditory for storage and rehearsal
by ACP
• Works for both auditory
presentation and visual
presentation of the letters.
• List 1 (Easy to remember/dissimilar
phonology and semantics):
• PIT, DAY, COW, PEN, HOT
• List 2 (Only slightly harder than List
#1/similar semantics) :
• HUGE, WIDE, BIG, LONG,
TALL
• List 3 (Much harder than List
#1/similar phonology) :
• CAT, MAP, MAN, CAP, MAD
• What happens if you prevent the recoding of visual
information into auditory information?
Phonological Loop
• Articulatory Suppression Effect
• Engaging in an auditory task after study removes phonological
similarity effect for visual items
• Procedure: Say “the” aloud over and over
• No re-coding of visual info by ACP
• Phonological info gets in directly, doesn’t need re-coding
Peterson & Johnson (1971)
Auditory presentation: PGTCD
(similar sounding) harder to recall
than RHXKW (different sounding)
Visual presentation: PGTCD
(similar sounding) recalled equally as
RHXKW (different sounding)
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
Dissimilar
Similar
Visual
Auditory
Item Type
Phonological Loop
• Irrelevant Speech Effect
• Background speech presented during study decreases memory for
visual items
• Irrelevant speech interferes with recoding of visual info to auditory
• Visual info weak in WM
• Something stored in the visuospatial sketchpad, but this system not
as efficient as the phonological loop
Phonological Loop
• Word-length Effect
Baddeley, Thomson, and Buchanan
(1975)
Results
• Recall decreases as the length
of time it takes to say a word
increases.
• Rehearsal takes longer for
longer words - can’t rehearse
as many times
• Retrieval from PS also takes longer due to auditory coding of items
• Reading rate correlated with memory ability
• Digit span depends on language - how long it takes to say numbers
Baddeley’s Model
• Potential Problems with the model
• Some of the supportive results can’t be replicated
(e.g., irrelevant speech effect)
• Model can’t explain all results:
• why word-length effect is larger for visual than auditory items
• why it differs based on serial list position
• Model is not precise in explanation of effects
Cowan’s Activation Model
• Cowan (1999)
• WM = info that is
currently highly activated
from STM or LTM
• Focus of attention
• Emphasizes attention’s role in
activation
• Activation of info when
attention is oriented to it
• Activation will decay to
cause loss of info from WM
WM
STM
Cowan’s Activation Model
• Central Executive
• Focuses attention and other
control processes
• Capacity of about 4 chunks
• Duration of 20s without
reactivation
• STM
• activated items that are
just outside of attention passive store
• Things within attentional
focus are available to
consciousness
WM
STM
Cowan’s Activation Model
• Potential problems with the model
• Only general descriptions so specific predictions are
hard to make
• Activation is not operationally defined very well when is something is “activated”?
• What causes decay? Passage of time isn’t causal
Nairne’s Feature Model
• Items represented in WM as individual features
(e.g., color, length, etc.)
• Features indicate
• presentation info (e.g., font, size, gender of voice, etc.)
• meaning info (e.g., what the item means, category, etc.)
• Features represented by -1 or +1 when studied
(yes or no for a feature, 0 if no info for feature)
• Interference: Later items with same features
overwrite feature info for previous items
Nairne’s Feature Model
Bold
Lower Upper Blue
SCHOOL +1
-1
+1
-1
+1
+1
-1
-1
fish
• “fish” presented after “SCHOOL”
- features in common can be overwritten - SCHOOL
can become 0, -1, +1, 0
- interference
During retrieval, item features are compared with items in
memory - lost features can be updated and restored
Nairne’s Feature Model
• Quantitative model - numerical predictions are
possible - can simulate data to generate
predictions for studies
• Simulations show that model can predict:
1) Recency effect
2) Suffix effect
3) Phonological similarity effect
4) Word length effect
Nairne’s Feature Model
Criticisms of the model
•
•
•
Jones and Macken (1995) showed that irrelevant
speech similarity had no effect
Model is complex - lack of parsimony
Summary of WM
(1) Focus on processing (vs. storage)
(2) Three main modern models
-
Baddeley model
-
-
Central executive controls VS, PL, EB
Cowan activation model
-
-
WM = attention focus, STM = activated
Nairne feature model (quatitative)
-
Items coded as features with overwriting interference