Working Memory

 

I ended up using the brilliant Cognitive Psychology: A Student’s Handbook by Eysenck & Keane (click image), which I recommend for its readability and theory evaluation.

 

But I made the mistake of starting out with Cognitive Psychology by Braisby & Gellatly. I found it difficult in places and ultimately frustrating.

 


 

 

DEFINITIONS

STM span: no of e.g. words or digits recalled

Reading span, operation span etc: read set of sentences/ do set of sums on separate cards; recall e.g. last word of each sentence seen

 

INTRO

LTM: system holding info for long periods of time

STM: holds info for much shorter periods (seconds)

Working memory

– function allowing processing and maintenance of temporary info

– related to STM, but supports operations on info, beyond just retention of temp info

– Before 1960’s was thought that memory was unitary

– but evidence pointed to separate systems: e.g. briefly shown stimuli was forgotten quickly

– Miller (1956): longest sequence recalled after a single showing is few letters/ nos/ words

– suggested STM is separate from LTM because of its very limited capacity

 


 

ATKINSON & SHIFFRIN (1971) – MODAL MODEL OF MEMORY

– STM acts as working memory performing many control processes

 

ATKINSON & SHIFFRIN (1971) - MODAL MODEL OF MEMORY

Model assumes:

  1. STM is limited capacity and subject to rapid forgetting
  2. control processes e.g. subvocal rehearsal can be used to keep info in STM
  3. info in STM is gradually transferred to LTM

Others suggested STS was crucial for cognitive functions e.g. comprehension & reasoning

BUT

– research estimates of STS capacity were different

 

Shallice & Warrington (1970): brain damaged KF

– had auditory digit span of 2 items, way below the norm but LTM learning was OK

– Broken STS with normal learning & reasoning etc meant STS wasn’t WM behind these

– KF diff. between STM for sound (affected) & vision (less affected) – can’t be just one STM

– LTM learning doesn’t have to go through STS (as shown in diagram)

 


 

see my working memory essay plan: here

 

BADDELEY & HITCH (1974) – WORKING MEMORY MODEL

– STM is a (sub) component of working memory

– WM overlaps with STM but they aren’t the same

– WM is a limited-capacity workspace that’s flexibly allocated to e.g. retention or reasoning

Central executive: limited-capacity processor coordinating two slave systems:

(b) Phonological Loop: for briefly holding speech-based information

(c) Visuospatial Scratchpad: for holding and manipulating visual-spatial images

Used technique dual-task paradigm: do cognitive task + STM task simultaneously

EK: (1) if 2 tasks use same component they cant be performed successfully together

      (2) if 2 tasks use different components they can be done successfully together or separately

– If STM task interferes with cog task then WM must be involved in performance cog task

– high STM loads caused interference but people kept 3 items without affecting primary task

– meant working memory had 2 components:

  1. a store holding small amounts of temp info
  2. a central executive that coordinates mental operations

– loop can hold about 3 digits before central executive attention is called for

– if 2 tasks use same component they should interfere; if different components not

 

BADDELEY & HITCH – ARTICULATORY/PHONOLOGICAL LOOP (AL)

BADDELEY AND HITCH - ARTICULATORY OR PHONOLOGICAL LOOP (AL)
BADDELEY AND HITCH – ARTICULATORY / PHONOLOGICAL LOOP (AL)

 

– B&H found that phonemic similarity affected reasoning & cognitive tasks

– STM is sensitive to phonemic similarity so cognitive tasks must share common factor

– concluded WM has a limited capacity speech based store, storing 2-3 items: AL

 

2 components:

  1. AL which can store small memory loads during cognitive tasks
  2. central executive (CEx) for control & coordination of mental operations (e.g. reasoning, learning or memory…)

– CEx is flexibly allocated to control or temp info storage depending on the task

– small memory load stored in AL handled without taxing CEx, but larger load would

  • holds speech-based information for ~2 s, about both sound and articulation
  • maintains information by rehearsal (is why hindering rehearsal reduces recall)
  • recall of digits is disrupted by speech sounds, but not non-speech noise

 

THE FOLLOWING FROM EYSENCK &KEANE TEXTBOOK:

PHONOLOGICAL SIMILARITY EFFECT

– confirms important verbal rehearsal part of the model

– show a short list of words and recall immediately in correct order

– Baddeley et al (2000) recall of phonemically similar words 25% worse

 

WORD LENGTH EFFECT

– expts show word length effect depends on PL

– number of items recalled immediately in the correct order

– Baddeley et al (1975) memory span bigger for short words than long words

– (means PL has a time duration)

 

WORD LENGTH EFFECT & DUAL TASK

– shows word length effect depends on PL

– Ss had to remember 5 visually presented words while repeating digits 1-8

– word length effect disappeared with visual words as they don’t enter the loop

– word length effect is there with auditory words as words have direct access to the loop

 


 

BADDELEY (1986) – VISUO-SPATIAL SKETCHPAD

– dual-task studies showed that combining 2 visuo-spatial tasks or 2 verbal tasks is more difficult than combining visuo-spatial & verbal tasks (e.g. Logie 1986)

– suggests separate specialised resources for verbal & visuo-spatial info

– interference between the two fits involvement of a common resource

– so CEx controls visual & verbal tasks and there is a storage for each

 

BADDELEY MODEL OF WORKING MEMORY 1986

 

Basso et al. (1982) – patient PV

– digit span of ~2 but normal recall of sequences on the Corsi blocks task (span=6)

Hanley et al (1991) – patient ELD

– opposite pattern: hard to explain without separating verbal and visual stores

 

SUPPORT FOR MODEL:

– some patients have impairments in visuo-spatial or verbal WM: evidence for separate verbal and visuo-spatial stores

– e.g. Renzi & Nichelli (1975) – corsi span & auditory digit span can be independently impaired in patients with diff types of lesion

Daneman & Carpenter (1980)

  • link between reading span (last word of sentences) and reading comprehension
  • due to fact they use the same underlying process: WM with info processing and storage

 

SUPPORT FOR PHONOLOGICAL WORKING MEMORY

Baddeley et al (1975) – faster readers

A – recall better for shorter words & faster readers

– people could recall as many words as they could say aloud in 2 seconds: rapid rehearsal

– other cross cult & developmental studies show same link with speech rate & STM span

– articulatory suppression (using the loop) kills the word length effect, making results similar

– consistent with rehearsal loop to stop decaying memory traces

 

B – articulatory suppression (recall plus additional task e.g. repeating “the the the”) disrupted recall, as this disrupted phonological loop

 

C – Baddeley et al (2000) recall of phonemically similar words 25% worse

 

 

SUPPORT IN THE FORM OF A NEURAL BASIS

Paulesu et al (1993) – PET (positron emission tomography scan)

– used PET during tasks activating separate components of phono loop

– Brocca’s area: sub-vocal rehearsal; left supramarginal gyrus: phonological store

Baddeley & Vallar (1984) – PV

– PV had 2 digit auditory span after stroke damaging left parietal cortex

– poor recall for phonemically similar words (but OK for word length)

– indicated phonological store deficit

 

BUT

Lovatt et al (2000) – word length/ spoken duration not important when phonological complexity of items controlled

Gathercole et al (1994) – STM word-length effect in kids before developing ability to use sub-vocal rehearsal

Gathercole & Hitch (1993) – output delays can cause word length effects: rehearsal as a reason isn’t required

Jones (1995) – Binding problem

  • integration & coordination about same objects between the separate sub-systems is not addressed by the model
  • speech & non-speech interference with verbal & visual memory tasks suggests unitary memory system (but dissociations evidence says not)

 


 

CENTRAL EXECUTIVE

CENTRAL WORKSPACE

Towse et al (1998) – switching hypothesis

  • loads increase in e.g. reading span tasks so less resource for processing tasks because of exec workspace trade-off
  • found no decline in processing speed: i.e. no support for trade-off in kids’ reading span, operation span etc
  • suggest kids switch attention between processing and storage
  • time was key: performance linked to forgetting during time interval leading up to recall

Camos & Barrouillet (2001) – interval and operation complexity affects span: hybrid model of attention switching & resource sharing

 

SUPPORT:

Support for prefrontal cortex involvement in the central executive

Baddeley & Della Salla (1996)

– prefrontal damage doesn’t produce a problem with basic maintenance of information

– but disproportionate problem in dual tasks when compared to controls

Passingham & Sakai (2004)

– Imaging also confirms the centrality of DLPFC in working memory

 


 

EPISODIC BUFFER

Cowan (2012) – working memory is boosted by expertise (e.g. chess formations), indicating link with LTM not in the model

Towse (1998) – reading span studies: can exceed the capacity of the phonological loop

Baddeley (2000) – added ‘episodic buffer’ (EB) to overcome these & binding issues

– limited capacity multidimensional store, backing up CEx as well as providing link with LTM

– binds info from the other short-term stores & LTM into chunks and hold roughly 4 of these chunks

– CEx processing power is needed for binding so research focus on loading CEx to prevent binding

 

Multicomponent Model of Working Memory - Baddeley, 2003

Multicomponent Model of Working Memory – Baddeley (2003)

 

SUPPORT:

Luck & Vogel (1997) – people can remember about 4 multi-featured objects, regardless of no. of features

Brener (1940) – memory span for sentences higher than word span (15 vs 7), due to binding into chunks using info from LTM

 

BUT:

– revised model doesn’t explain how e.g. AL link to LTM operates

Allen et al (2009) – loading CEx in visual dual task didn’t particularly disrupt binding more than disrupting memory for indiv features

Baddeley et al (2009) – an unexpected result!

– found similar in verbal domain: sentences vs scrambled word lists

– not particularly disrupted because binding is automatic

– binding happens outside episodic buffer and VSS & AL feed EB directly, not CEx