Chapter 4: Short-Term Working Memory 10/16/2012 10:34:00 AM
George Miller: “The Magical Number 7”
Rate of Forgetting
Peterson and Peterson study: tested undergrads on their ability to remember
3 consonants over a short retention interval.
E.g. subject heard the letters CHJ followed by the number 506. They would
then count backward until a light presented, which was a signal for recalling
the three consonants (the counting down from 506 by 3s disabled any verbal
rehearsal). The light went on between 3-18 sec after the subjects began
Results: probability of a correct recall declined rapidly over the 18-
second retention interval.
o Implies we must rehearse verbal info to keep it available in
o Good side about putting a phone number in STM rather than
LTM: Harder to retrieve the few numbers that you actually
Decay versus Interference
Is the loss of info from STM caused by decay or interference?
Interference theory: memory for other material or the performance of
another task interferes with memory and causes forgetting.
Decay theory: info is spontaneously lost over time, even when there is no
interference from other material.
If memory decays over time, the amt of recall should be determined by the
length of the retention interval.
If memory is disrupted by interference, then recall should be determined by
the number of interfering items.
Waugh and Norman: tested whether the loss of info from STM is caused
by decay or interference. - presented lists of 16 single digits. The last digit in every list occurred. The
task was to report the digit that had followed the probe digit.
They varied the # of interfering items by varying the location of test digit in
the list. (many interfering items if test item occurred early in the list, and
only a few if the test item occurred late in the list.
They also tested the rate of pres, to see if the prob of recalling the test difgit
would be influenced by the length of the retention interval.
Pres digits either 1 or 4 digits per second. Result: the rate of pres
had little affect.
Memory is only slightly better for the shorter retention interval.
The prob. Of recall declines rapidly as the number of interfering
Therefore, W and N’s findings suggest: Interference, rather than decay, is
the primary cause of forgetting. Although some decay may occur, the amt of
forgetting caused by decay is way less than the amt caused by interference.
Main Point: Interference is the chief cause of forgetting!
Release from Proactive Interference
Means how interference can be reduced by decreasing the similarly
Proactive: forgetting that occurs because of interference from material
encountered BEFORE learning.
Retroactive: forgetting that occurs… material encountered AFTER learning.
Release from P I will occur by having information be dissimilar from earlier
Reduction of interference thru appropriate sequencing can partly
compensate for the rapid forgetting from STM.
CAPACITY The Magic Number 7
Memory span task: the # of correct items that people can immediately
recall from a sequence of items.
Magic #7 also in an absolute judgment task: identifying stims that vary
along a single, sensory, continuum (e.g. varying by loudness, pitch,
Subject’s task is to learn to identify each stim (diff levels of loudness for
e.g.) by assigning the correct label (1 being softest, 7 loudest)
The limitation was caused by the inability to keep more than about 7 sensory
values available in STM cuz of its limited capacity.
The upper limit (7 +-2) of absolute j experiments corresponds very well with
the upper limit of memory span tasks!
Miller: putting stuff in chunks helps memory
Chunks: a cluster of items that has been stored as a unit in LTM (e.g. FBI,
Miller: The capacity of STM should be measured in chunks rather than in
Individual Differences in Chunking
Chunking relating to how chess players reproduce the pieces on a
chessboard. This study was done by de Groot a Dutch psychologist.
Main conclusion of the study: differences in skill among novice and master
players is from diffs in perception and memory as opposed to how they
planned their moves.
Master players depended on their ability to code the pieces into
o When the players viewed pieces that were placed randomly
on the board, the master players no longer had an advantage
over the weaker players, and the two groups performed about
Chase and Simon did a similar test: identified chunks that
presumably produced the superior coding ability of master players o Result: Obvee masters better cuz they had more chunks and
more pieces (chess pieces) per chunk.
o ALSO… the estimated # of chunks across the 3 skill levels are
at that 7 range.
Sweller’s cognitive load theory (includes redundancy effect, split attention
effect, expertise reversal effect, etc): our efforts to combine information can
result in cognitive overload – too much information for our STM to manage.
Can result from a split attention effect: occurs when ppl must
divide their attention b/t 2 sources, such as instructions and the
o To avoid this effect, physically integrate the info. E.g. steps
with diagrams so that the reader does not have to
continuously switch attention b/t 2 sources.
Sweller and Chandler’s experiment: two groups:
Group 1 received standard instructions and could apply them to a physical
Group 2 received MODIFIED instructions (steps as well as diagrams)
This instruction proved most effective
If for instance Group 2 received modified instructions as well as the physical
apparatus, Sweller and Chandler proposed they would suffer cog overload
cuz of the redundancy effect: if equivalent info is provided twice, then the
added info simply provides MORE rather than new, information. It would
increase time of completing task too. Result was that if Group 2 also got
physical thing, they would perform the same as those in Group 1.
Expertise reversal effect: instruction that may reduce cognitive load for a
novice, may actually increase cognitive load for an expert.
E.g. adding text to a diagram will increase the load for the expert if
he/she can understand the diagram already w/o the text.
MEMORY CODES Argument that acoustic (speech based) codes are the predominant codes
in STM, and that semantic (meaning based) codes are the predominant
codes in LTM.
Acoustic Codes and the Rehearsal
2 separate verbal-processing rates influence a person’s memory span.
1) The speed at which a person can pronounce the items on the list
used to test the memory span
2) The speed at which a person can retrieve the items from STM.
Both rates determine how many items a person can keep active in STM.
Cowan’s exp: students asked to recall strings of digits.
Pronunciation rates measured by asking students to count from 1-
10 as quick as can be.
Retrieval rates measured by seeing the amt of time b/t digits as
students attempted to recall the items during the memory span
Results: both pronunciation rates and retrieval rates are correlated
with memory span BUT they do not correlate with each other!
Acoustic confusion: an error that sounds like the correct answer.
It is very easy for acoustic confusions to occur when all letters in a sequence
sound alike. They occur often in STM tasks.
Phoneme: any of the basic sounds of a language that are combined to form
speech. (e.g. the letter C (“se”) ) The letter C has two components –
namely, the s and e sounds. These are the basic sounds of the English
Each pronunciation is represented by a diff phoneme
Phonemes can account for acoustic confusions cuz words that sound alike
usually have some phonemes in common.
Laughery’s assumption: each of the auditory components representing an
item can be independently forgotten – that if a name consists of two phonemes for e.., a person might remember one phoneme but not the
Easy to mix up letters cuz its