Shortterm memory, STM

1
Short-term memory, STM

• COGS 551, Human Memory
• Annette Hohenberger
• 05/03/07

2
Short-term memory, STM

• Baddeley, Alan (1999), Essentials of human
  memory. Chapter 2 Short Term Memory, 21-43.
  Hove Taylor and Francis.
• Alan Baddeley (1997) Human Memory. Theory and
  Practice. Hove Psychology Press.
• Miller, George A. (1956) The Magical Number
  Seven, Plus or Minus Two. Some Limits on our
  capacity for processing information.
  Psychological Review, 63, 81-97.
• http//psyclassics.yorku.ca/Miller/

3
STM What is it?
4
What is STM?
Definition of Short-term memory in the Webster
Medical Dictionary Short-term memory A system
for temporarily storing and managing information
required to carry out complex cognitive tasks
such as learning, reasoning, and comprehension.
Short-term memory is involved in the selection,
initiation, and termination of information-process
ing functions such as encoding, storing, and
retrieving data. One test of short-term memory
is memory span, the number of items, usually
words or numbers, that a person can hold onto and
recall. In a typical test of memory span, an
examiner reads a list of random numbers aloud at
about the rate of one number per second. At the
end of a sequence, the person being tested is
asked to recall the items in order. The average
memory span for normal adults is 7. Short-term
memory is also termed recent or working memory.
http//www.medterms.com/script/main/art.asp?articl
ekey7142
5
STM How long?
6
Duration of STM?
30 s? 2 s? weeks?
Duration of short-term memory (Wikipedia) The
most important characteristic of a short-term
store is, clearly, that it is short-term that
is, it retains information for a limited amount
of time only. Most definitions of short-term
memory limit the duration of storage to less than
a minute no more than about 30 seconds, and in
some models as little as 2 seconds. Some models
limit short term memory to weeks. Memory that
exceeds short-term memory duration limits is
known as long-term memory.
Is this the same Memory system or are these
separate ones?
http//en.wikipedia.org/wiki/Short_term_memory
7
STM and limited consciousness

• Obviously, our STM is limited to
• (1) a certain time window during which we can
  integrate information
• (2) a certain amount of information we can
  integrate during that time
• What evidence is there for how limited STM is in
  terms of time and information?

8
Short Term Memory (Baddeley 1997)

• 9753
• 5625
• 8967
• 67368
• 16497
• 56328
• 390139
• 476262
• 147905
• 5391674
• 6592136
• 6859122
• 57351430
• 15871804
• 35267728
• 580618479
• 059252865
• 596012607

• Digit Span Joseph Jacobs (1887) devised the
  first digit span test to his pupils
• Criterion point at which the subject is right
  50 of the time
• About 7 items
• Increase through reading aloud or grouping
  (chunking in groups of 3)
• Further increase through LTM strategies

9
Short-term forgettingThe Peterson Paradigm

• People are extremely susceptible to forgetting
  novel material if they are briefly distracted
• The Peterson Paradigm
• 1. Subject is presented with CCC trigrams
• B L Q ...
• K Z X ...
• Q M C ...
• _______________________
• 2. Subject is presented with numbers
• 492 640 328
• Task repeat back numbers in steps of 3, move on
  to the next trigram, etc.
• ________________________
• 3. recall all trigrams after 0, 3, 6 9,12,15,18s

10
The Peterson Paradigm

• In the Peterson Paradigm, the retention interval
  (time from presentation to recall) can be varied
  systematically.
• With growing retention interval, memory for the
  trigram drops dramatically. However, recall is
• Very good retention for 1 word
• Same for three words than for three letters
• STM depends on the number of chunks, not on the
  content of the chunks (whether they consist of
  words or of letters)
• What is chunking? (? Miller 1956)

11
Trace decay or interference?

• Forgetting as the result of automatic fading of
  the memory trace
• In the Peterson Paradigm, the numbers were so
  different from the letters that they should NOT
  interfere. --gt The memory decay is spontaneous

• Forgetting as the result of the disruption of the
  memory trace by other traces
• The more similar the two traces are, the more
  likely the interference
• --gt Proactive and retroactive interference

12
Proactive (PI) and Retroactive interference (RI)

• RI
• Old material is disrupted by new material
• Yatak lt-- yanak

• PI
• New material is disrupted by old material
• Yatak --gt yanak

Forgetting of the trigrams in the Peterson
paradigm was the result of proactive interference
from earlier trigrams --gt the first trigram in a
sequence was hardly ever forgotten since there
was no material that could interfere with it
proactively
13
Release from PI

• If the nature of the items to remember is
  changed, the first one of the new group is
  relieved from PI
• COW SPINACH MILK
• DOG CELERY WATER
• SWAN POTATOE JUICE
• BUG CARROT BEER
• APE BROCCOLI WINE
• Subjects must have processed these categories
• Also works if the categories are pointed out to
  the subjects after learning but prior to recall!
• Semantic changes are most efficient
• --gt Next experiment?

14
Interference account of Release from PI

• Interference depends on similarity. Release from
  PI stems from the spontaneous recovery from
  extinction of earlier items due to dissimilarity
  of novel items.

15
Problems for decay and interference theory

• (Loess and Waugh 1967) could show that NO
  short-term forgetting takes place when the
  interval between the trials (ITI, Intertrial
  Interval) was increased to 2 minutes.
• --gt each trial became like the first trial
• Contradicts the decay theory (after 2 minutes
  material should have decayed)
• Contradicts interference theory (after 2 minutes
  interference should have taken place)
• --gt compromise there is both decay and
  interference, rise of discrimination theory

16
Trace discrimination hypothesis(Nairne 2002,
2005)

• Peterson task
• Presenting two items (P1, P2) at a certain
  interval followed by blocked delayed recall
  (recalling P2 at a certain fixed time)
• Group 1 P1 at 0s, P2 at 5 s, recall R of P2
  always after 5 s (at time 10s)
• Ratio between P2 and R12 (510)
• Group 2 P1 at 0s, P2 at 20 s, recall of P2
  always after 20s (at time 40).
• Ratio between P2 and R12 (2040)

17
Trace discrimination hypothesisFamiliarizationGr
oup 1 (R2 recall P2 always after 5 s)Group 2
(R2 recall P2 always after 20 s)
0 5 10 20 30 40 50 P1
P2 R2 Ratio 12 (5/10s) 0
10 20 30 40 50 P1 P2

R2 Ratio 12 (20/40s)
While the absolute times of presentation and
recall differ, their ratios are identical.
18
Trace discrimination hypothesisTest Both groups
receive delayed recall after 15 s
0 5 10 20 30 40 50 P1 P2
R2 Ratio 15201.33 (worse than
before) 0 10 20 30 35
40 50 P1 P2 R2 Ratio
15352.33 (Better than before) Group 1 old
ratio 12 gt new ratio 11.33 --gt harder, since
less distinct! Group 2 old ratio 12 lt new
ratio 1 2.33 --gt easier, since more
distinct!
15s
5s
20s
15s
20s
35s
19
Trace discrimination hypothesis

• Recall is not only dependent on ITI (Inter trial
  interval), but also on the retention interval of
  the preceding trials.
• Having recalled after a fixed long interval
  (20s), it is easier to switch to a shorter
  interval (15s) with a larger P2R2 ratio, as in
  group 2, as compared to having recalled after a
  fixed short interval (5s) followed upon a longer
  interval (15s) with a smaller P2R2 ratio, as in
  group 1.
• As the ratios change, subjects experience more or
  less discrimination between the stimulus time and
  the retrieval time.

20
Back to Release from PIDiscrimination account of
Release from PI

• Subjects use the nature of the target item (the
  category information) in order to exclude
  dissimilar prior items.
• If a new category starts, the new candidate can
  be discriminated better, relative to the old one
  of the old category.
• Works even ex post facto, when subjects are
  informed about a subtle category switch (wild vs.
  Cultivated flowers) AFTER learning but PRIOR to
  recall. This can only be explained by the
  discrimination hypothesis. Thecue can release a
  discrimination ex post facto.

21
One or two memory stores?Controversy in the
1960ies

• Unitary system (Melton 1963)
• Showed LTM effects in STM tasks
• Presenting an item several times enhances its
  memory in a Peterson task
• Presenting the same sequence interspersed between
  other sequences enhances its memory
• Favors a continuous view of memory

• Duplex approach (Waugh Norman 1965)
• Primary Memory (STM) and Secondary Memory (LTM)
  contribute to a STM task (e.g., immediate recall)
• Only Secondary Memory (LTM) contributes to a LTM
  task (e.g., delayed recall)
• Favors a two-component view of memory

22
ST and LT aspects of a STM task

• One is not to confuse the nature of the memory
  system with the performance in a specific memory
  task.
• A STM-Task such as immediate recall can rely on
  both STM and LTM aspects of memory.
• Since that time, one distinguishes between Memory
  System and Memory Task

23
4 arguments against a unitary view

• 1. 2-component tasks Recency effect
• 2. storing capacity
• 3. coding acoustic, semantic
• 4. Neuropsychological evidence, patients

24
1. Recency effect
Recency effect

• In an immediate recall task, the recency effect
  shows up.
• It disappears after a brief filled delay, though.
• --gt recent items are held in STM, earlier items
  are held in LTM (at least in a longer-term system)

www.uark.edu/misc/lampinen/serialpos2.gif
25
2. Storage capacity and processing speed

• STM and LTM have different storage capacities
• STM very limited (7 chunks), rapid processing
  (input and retrieval)
• Responses from recent items are faster than
  responses from earlier items
• Recency effect is unaffected by concurrent load
  (secondary task, e.g. Sorting cards), hence not
  dependent on attention.
• LTM almost unlimited, slow processing (input and
  retrieval)
• LTM is affected by concurrent load.

26
3. Acoustic and semantic coding

• STM encodes in terms of sound
• In trigrams, letters are misrecalled in terms of
  their phonological similarity
• V substitutes P but R does not substitute P
• Order of items in sequences similar in sound is
  harder to recall than in unsimilar ones PDVCT
  harder than KYZWR
• Lists of words similar in sound are harder to
  recall in a STM task (immediate recall) as
  compared to those similar in meaning

• LTM encodes in terms of meaning
• Lists of words similar in meaning are harder to
  recall in a LTM task (delayed recall)
• --gt possible next experiment
• Detection of syntactic or semantic changes in
  texts is time-sensitive both can be detected
  well immediately but only semantic changes
  survive a delay.

27
4. Neuropsychological evidence

• LTM/-STM
• Patient K.F. has impaired STM (only digit span of
  2-3 items) but could acquire new information in
  his LTM
• --gt argues against Atkinson's and Shiffrin's
  model!

• -LTM/STM
• Patient H.M. has LTM problems (could hardly
  acquire new information, only old info was
  preserved). However, immediate memory span normal

Double dissociation between LTM and STM.
28
Atkinson and Shiffrin's model
29
Problems with the AS model

• 1.
• 2.
• 3.
• 4.

30
1. Neuropsychological evidence

• Patients with STM deficits should also have LTM
  deficits. This is not necessarily the case, as
  the case of K.F. shows

31
2. Maintenance rehearsal

• The probability of an item to be transferred to
  LTM is a function of maintenance in STS,
  according to AS.
• However,mere maintenance rehearsal does not
  enhance transfer to LTM
• Also, incidental exposure to an information many
  times does not guarantee recall, e.g.,
  remembering the characteristics of a 1-YTL-coin

32
3. LT recency effects

• There exist LTM recency effects
• When a filled delay occurred after each item to
  be learned, a recency effect even after delayed
  recall was found!
• Two kinds of recency effects? Better to explain
  it with a singular concept.

33
4. Coding too simplified STM phonologicalLTM
semantic

• Depending on the nature of the task, subjects
  rely on either phonological or semantic codes.
• Semantic coherence can also be used in true STM
  tasks.
• Priest pious apple delicious
• yields powerful semantic cues for immediate recall

34
Levels of processing model(Craik Lockhart,
1972)

• STM feeding LTM was not approriate
• Type of learning is decisive, not so much the
  structure of the memory system.
• Craik Lockhart emphasized the style of coding
• Shallow processing (phonological encoding) vs.
• Deep processing (semantic encoding)
• Longer storage resulted form deeper (semantic)
  processing, not from transfer from one store to
  the other.

35
Levels of processing model(Craik Lockhart,
1972)

• Craik Lockhart distinguished between
• Maintenance rehearsal shallow recycling of
  material in the phonological mode
• Elaborative rehearsal deeper encoding of the
  material in a semantic mode
• The levels of processing theory was more
  concerned with role of coding in LTM than with
  the structure of the memory system (whether
  unitary or multiple)

36
References

• Baddeley, Alan (1997) Human Memory. Theory and
  practice. Hove Psychology Press.
• Craik, F. Lockhart, R. (1972). Levels of
  processing A framework for memory research.
  Journal of Verbal Learning Verbal Behavior, 11,
  671-684.
• Nairne, J. S. (2002). Remembering over the
  short-term The case against the standard model.
  Annual Review of Psychology, 53, 53-81.
• Nairne, J. S. (2005). Modeling distinctiveness
  Implications for general memory theory. In R. R.
  Hunt J. Worthen (Eds.), Distinctiveness and
  memory. New York Oxford University Press.