1. Cognitive Psychology C81COG 5. Memory – Structure & Processes
Dr Jonathan Stirk

2. Background Reading
Baddeley, A.D. (1999). Essentials of human memory. Hove, UK: Psychology Press.
Chapters 2 & 3

3. Overview Dichotomies in the description of memory Memory Processes
Structural attempts to separate STM and LTM
forgetting by decay or interference
capacity differences
coding by sounds or by meanings?
The simple multi-store model fails, so
an alternative to structural distinctions?
a revised model, based upon “working memory”

4. Common distinctions in the description of memory
Short-term (1°) vs. Long-term memory (2°)
The phone number of your dentist (look it up in the phone book, then dial it) vs.
A very familiar phone number (e.g. Home)
Distinction based on temporal aspects of memory
Temporary and fleeting vs. more permanent and stable
Primary memory (STM) is what is in consciousness at the present time
Secondary (LTM) is what is more permanently etched in memory

5. Common distinctions in the description of memory
(LTM)
Procedural
[non-declarative]
(Skills, priming, conditioning)
Declarative
(Knowledge)
Open to intentional retrieval
Squire (1992) – components of LTM
Measured through performance rather than conscious recall/recognition
Episodic
Semantic
Larry Squire (1992)

6. Procedural vs. Declarative LTM
How to tie your shoelaces vs. the name of the current US president
Procedural Memory: ‘knowing how’, remembering how to perform skilled actions, implicit
Declarative Memory: ‘knowing that’, explicit
How?
That?

7. Episodic vs. Semantic LTM
What did you have for lunch today? vs. What kinds of animals are canaries?
Tulving (1972) distinguished between episodic & semantic memory
Tulving (2002) – “Episodic memory is about happenings in particular places at particular times or about “what”, “where” & “when”.
Episodic Memory: autobiographical, temporally dated, memories of life experiences
Semantic Memory: memory for words (lexical memory), world knowledge, not temporally dated, very well-organized (organization protects memory from interference)

8. Semantic Memory – Tulving (1972)
“It is a mental thesaurus, organised knowledge a person possesses about words and other verbal symbols, their meanings and referents, about relations among them, and about rules, formulas, and algorithms for the manipulation of these symbols, concepts, and relations.”

9. Memory Processes
Process - Activities taking place within memory
Encoding
Converting information into a form suitable for use in memory
Storage
Retaining information in memory
Retrieval
Bringing to mind information stored in memory

10. Memory Structure - Multi-store Model (Modal model)
Sensory registers
Short-term store
Long-term store
Structure refers to the organization of memory/ the architecture
One model of memory is that it is structural with a number of separate stores for information (multi-store)
Prior to this many assumed a single store (unitary) model of memory!! Structural distinction between STS and LTS
Developed from numerous findings at that time in the memory literature.
Sensory registers: modality specific, fast decay
STM: Limited capacity; information displaced
LTM: Unlimited capacity, holds info over long period of time, prone to interference
Animation: 3 processes in which information is lost from 3 stores (decay, displacement, interference)
Attention allows us to transfer information from sensory store to STM
Rehearsal allows us to transfer from STM (primary memory) to LTM (secondary memory- info left consciousness)
Maintenance rehearsal- consciously repeating info (STM to LTM)
Elaborative rehearsal- consciously focusing on the meaning of info (STM to LTM)- see next slide
Forgetting mechanisms
DECAY
DISPLACEMENT
INTERFERENCE
(Atkinson & Shiffrin,1968)

11. Elaborative Rehearsal
A strategy of associating a target stimulus with other information at the time of encoding
I.e. Creating meaningful associations
By creating meaning and associating meaning with other information in memory, information can be maintained in memory
Will come back to elaborative rehearsal later in ‘levels of processing’ section

12. Differences Between Memory Stores
Evidence for separate and distinct memory stores has been supported by looking at:
Forgetting mechanism(s)
Storage capacity
Temporal duration
Effects of brain damage
In order to justify the 3 qualitatively separate and distinct memory stores, we would expect to find major differences between them.
This has been shown to be the case.

13. Properties Of The Short Term Store
FORGETTING MECHANISMS
Loss of information through autonomous decay
Distraction results in loss of information (Peterson & Peterson, 1959)
LIMITED STORAGE CAPACITY
STM span limited to 3 items (pure STM capacity limit; not Miller’s 7+-2; see next slide)
Recency in serial recall curve is read-out from Primary Memory (Craik, 1964)
Delayed recall eliminates recency (Glanzer & Cunitz, 1966)
STM loses information if it is not rehearsed by processes of forgetting (decay/displacement)
Decay- the memory trace fades as a function of time
Distraction prevents rehearsal- discuss P & P study on next slide
Limited to 3 items (limit to STM)- Immediate memory span which includes LTM is 7+-2 (Miller Magic Number)- allows for chunking!!
Immediate memory span allows info to be transferred to LTM increasing the span. When rehearsal stopped a more accurate capacity of 3-4 items is obtained

14. Chunking Allows More Information To Be Stored
Miller (1956) suggested that memory capacity is not limited by number of items that can be stored but by number of chunks. Chunking requires support by LTM
Storing a long phone number may be helped along by chunking information into meaningful bits as in slide. This however requires support from LTM e.g can be stored as 1 chunk – Nottingham dialling code

15. Evidence for forgetting by decay in a STM task
Peterson & Peterson (1959) distracter task
Listen to a consonant trigram (e.g. SKB)
Now start counting backwards in threes from this number (e.g. 751)
Varied how long they did this for (variable delay to allow the counting to have an effect)
Now recall the trigram
Repeat 1-3 for more trials/trigrams
Q. What % of trigrams are correctly recalled, and how is this influenced by the delay between presentation and recall?
Distraction prevents rehearsal – P & P examined effects of stopping rehearsal on STM – stopping rehearsal should prevent information being passed into LTM (according to the modal model!)
A trigram contains less items than the ST immediate memory span. So should be easy to remember.
Presented consonant trigrams (3 letters)
Count back: etc varied delay btwn presentation of trigram and recall (i.e. length of distracter task)

16. Loss Of Information From STM By Decay
Results
Interpretation: We lose information from STM through autonomous decay when information is left unattended.
Around the 1960’s forgetting in LTM was thought to be by interference
See also Brown (1958)
Also known as Brown-Peterson distracter task
Ability declined to approx 50% after about 6 seconds of distracter task; after 15 sec recall close to zero
So P&P showed rapid loss of information through trace decay.
Same effect when words used instead of trigrams
Compare STM to the reverberations of a bell, bang bell and hear sound for a while but it dies down unless banged again
Delay between presentation of trigram & recall, during which participants are distracted by counting backwards, & during which forgetting is possible

17. Glanzer & Cunitz (1966) – Further evidence of 2 stores (STS & LTS)
LTM
STM
Usual recency effect (about 3 items!)
G & C presented 10 words to subjects and then tested free-recall immediately or after 10 & 30 seconds delay.
Primacy & Recency effect – in free recall (words recalled in any order) words at beginning & end recalled better
Serial Position Curve (U-shaped) – Serial position effect
The U shape was evidence for 2 memory stores – Primacy section items made it to LTM (more time for rehearsal) , Recency section items left in STM before decay
Delay effects the end part of the curve but not the middle or start:
The recency effect is eliminated by delay in a Brown-Peterson distracter task (see orange)-trace decay due to no rehearsal
So Recency effect is caused by read-out form Short Term Memory store
Rest of curve represents read-out from LTM
Recency effect is eliminated by delay
Primacy Effect

18. Other factors effecting the serial-position curve: Word Frequency & STM
high freq (common)
low freq (rare)
LTM
STM
Unlike delay, word frequency seems to effect the beginning and middle parts of the curve but NOT the end
More frequent/common words recalled better
Word frequency affects recall from LTM but not STM (Raymond, 1969; )
Common words have better established meanings which are stored in LTM
Raymond, 1969

19. Double Dissociation The primacy effect is altered by
Double Dissociation: two situations or theoretical entities are affected in opposite ways by one or more independent variables
The primacy effect is altered by
The word frequency of items used in the task
Low frequency words → lower recall of first few items in list
The recency effect is altered by
the use of a delay between learning & recall
distracter task → lower recall of last items in list
The previous 2 slides show an example of a Double Dissociation: two situations or theoretical entities are affected in opposite ways by one or more independent variables (e.g., distraction affects WM but not LTM; word frequency affects LTM but not WM)
if we can influence primacy without recency AND
we can influence recency without primacy
But NOT by a delay!
But NOT by word frequency!

20. Recall These 8 Letters In Order
V C B D G E P T
R F Q P L N Z K
Ask class to recall letters in order –click appear then click disappear!
Did they do better in List 1

21. Did you perform better for list 2?
Recall These 8 Letters In Order
V C B D G E P T
R F Q P L N Z K
Did you perform better for list 2?

22. Further Properties Of The Short Term Store
Coding
Coding in LTM is thought to be semantic (based on meaning)/associative.
What about coding in STM then?
Evidence for Acoustic coding
Confusions in recall between items which have similar sounds (Conrad, 1964; RLTKSJ →RLCKSA)
Categorical but non-associative storage
Meaningful relationships between words are not appreciated (Tulving & Patterson, 1968; )
Evidence that the code is acoustic/phonological. Conrad (1964) – found that when recalling letters in a 6 letter stimulus people made substitutions with similar sounding letters
Unlike Iconic memory, code is categorical- words can be assigned to semantic classes e.g. Letters( consonants vowel consonant), numbers.
T & P (1968) lists of words to free-recall, some lists contained 4 semantically related words (arm, hand, leg, foot)
C – Unrelated words, E – Related words form a group at the end of the list (STM), M – Group in middle (LTM), D – distributed
Findings: related words retrieved from LTM as a single unit from the M list but not the E list.
Unitization of related words much smaller in E(nd) lists. Semantic relationships between words NOT coded in STM!

23. Summary of Properties Of STM
The short-term store retains a limited amount of information for a brief interval, and what it retains is categorical but non-associative items using an acoustic code
Limited capacity
Short duration (information lost by decay if not rehearsed)
Phonologically coded
More common words recalled better ( common words have better established meanings – stored in LTM)

24. Problems With Earlier Interpretation of Peterson & Peterson’s (1959) exp’t!
Keppel & Underwood (1962)
An average taken for each subject across numerous trials
This assumes no performance change across trials
No forgetting on trial 1, some on 2, more on 3
After an 18 sec delay, PI from triplets presented on earlier trials builds up.
Pro-active interference/inhibition (interference from info presented before the to be remembered info)
So it;s looking like STM and LTM have different properties, suggesting that they are separate distinguishable stores. BUT!!!!!
K & U suggested that interference could explain the results!
Interference theory stemmed from work of SINGLE MODEL THEORISTS
Used the same technique as P & P but looked at trial by trial analysis.
NO forgetting apparent on trial 1, despite the large delay
At this stage there was evidence to support both single and multi-models of memory.
So as STM may use interference as a mechanism for forgetting it is now looking very similar to LTM!!!!!!

25. Problem: Evidence of Semantic Coding In a STM task
Warren & Warren (1976)
S hears PRINCE, KNIGHT, QUEEN (the “memory set”)
10 seconds of a distracter task
Subject recalls the “memory set” (“Prince…”)
Next trial sometimes contains a homophone-related set e.g. MORNING, NOON, EVENING
10 seconds of distraction, etc, etc
Question - Do subjects ever confuse KNIGHT and NIGHT in the recall of the homophone-related set? (Is night offered in error in the recall of MORNING, NOON, EVENING?)
Answer - YES, therefore there has been some semantic processing of these words even though they are in a STM task
By the early 1970’s the Atkinson & Shifrin 3 stage structural model was being questioned.
Patients with poor short term memory still could lay down detailed long term memories. This shouldn’t happen according to 3 stage model as rehearsal in STM should be necessary to transfer to LTM.
So for STM the code seems acoustic but is there evidence of coding by meaning (semantic coding)?
W&W seemed to show that there is evidence of semantic processing.
Suggests that the idea of a multi-modal model may be flawed and that we needed something else.

26. Separate Properties?
So, is STM a “pure” system, with its own forgetting/coding characteristics, or does it share its properties with LTM?
A qualitative difference from a quantitative change?
Oxidation of silver over time
After 10 minutes (insensitive method)
After 10 months (now more sensitive)
The same process is at work, but the effects appear to be different over different time intervals
Perhaps the methods of studying forgetting/coding are also insensitive and have also given misleading conclusions
We assume from the data / studies that quantitative differences mean qualitative difference between STM & LTM
We can compare this with looking at oxidation of silver- nothing seems to be going off over 10 minutes but after 10 months there are noticeable differences.

27. Separate Stores? So evidence for 2 separate stores is not conclusive!
Rather than focusing on structure, why not look at the processing of information instead?
Craik & Tulving (1975)

28. Levels Of Processing Approach (Craik & Tulving, 1975)
Structural model breaks down
Atkinson & Shiffrin (1968)
Rehearsal always enhances transfer to LTM
Not true!
Remembering is influenced by the “level of processing” during input (e.g. visual processing vs. semantic processing)
The emphasis is now upon process rather than structure
Such evidence was making it harder to argue for separate structural distinctions between STM & LTM.
A newer approach was to look primarily at the processes rather than the structure of memory
C&T argued that nature of processing affects how effectively info is transferred and stored in LTM

29. Craik & Lockhart (1972)- Levels of processing hypothesis
Depth of processing predicts durability of memory
Craig & Tulving (1975) - List of visually presented, unrelated words.
3 groups each with a different decision task
1. Is the word in CAPITALS?
2. Does it rhyme with “ate"?
3. Is it a type of fish? Or does the word fit into the following sentence?
Measured latency (decision time)
Then surprise memory test – recognition from list of targets & distracters
Distinguished between shallow( e.g. detecting specific letters in words, physical characteristics) and deep/semantic processing (meaning)
C&T did a number of exp’ts (10) to investigate their levels of processing theory
C &T had subjects make various decisions about unrelated words presented to them
Visual processing
Acoustic processing
Semantic processing

30. Craik & Tulving’s Results
As the depth of processing increases decisions take longer to make (latencies increase – red figures)
and
Unexpected recognition improves (blue bars)
0.746
0.689
Increased depth of processing improves recognition
Amount of processing is also important
Increased latencies infer increased depth of processing
They also looked at Recall as well as recognition and also found same affect
0.614
RED = latencies (secs)

31. Updating The Multi-store Model Of Memory
Baddeley & Hitch (1974)
Model emphasises both storage AND processing
Not unitary – verbal /visual-spatial subsystems
Original incarnation was tripartite (3 components)
The shift towards focus more on process rather than structure prompted Alan Baddeley to suggest a newer and more intricate model of STM which he named his “working memory” model
Shifted the emphasis from passive storage (short-term memory) to active processing (working memory)
Purpose of STM was to hold information so it could be worked on
WM not a unitary system- has separate verbal and visual-spatial subsystems

32. Working Memory
Central executive: directs and controls all WM functions; can be thought of as ‘attention’. Integrates info
Visuospatial scratchpad/sketchpad: a system for holding visual information
Articulatory/phonological loop: a system for holding and recycling auditory/acoustic information
Rehearsal/Articulatory loop : a process for recycling, using subvocalization
Phonological buffer/store: a structure for holding acoustic information
Each component is capacity limited
Articulatory loop & VSSP are modal, whereas CE is modality independent and whether CE is a unitary system or has multi-components.

33. Working Memory Updates
Later version (Baddeley, 2000) incorporate an “episodic buffer” which time-stamps memories
So we know when something happened and not just that it happened
Later versions split the articulatory loop and visuo-spatial sketch pad into two dual slave systems which are used by the CE
Still problems with this model though.
Little known about the CE and its precise functions