WM Flashcards
1954: HM - bilateral removal of the temporal lobe
Severe amnesia – inability to form new LT memories for events and facts
Preserved STM
Preserved procedural memory
In 50s, researchers realised there was a dedicated memory system in medial temporal lobe
see notes
1969: KF - lesion to perisylvian cortex
Reduced digit span (STM – approx. 2 items)
Preserved LTM
Double dissociation between LTM and STM
- Where one patient impaired on process X but has preserved process Y and another is impaired on Y but preserved on X
o Strongest form of neuropsych ev for 2 functions being dissociable – idea that DD can only occur if 2 functions depend on separate brain regions
see notes
double dissociation logic
If 2 functions depend on overlapping brain regions, a lesion to one region will affect both functions and vice versa
see notes
STM
Atkinson and Shiffrin – Modal Model (1968)
At each level, info can be lost by decay, interference/combo of both
Serial model – info has to pass through STM to get into LTM
STM is unitary store
In model STM info comes from env, processed by sensory memory then, if is focus of attention, stored in STM – if rehearsed can be passed to LTM
see notes
ev against a unitary STM store
Patients with STM deficits didn’t show generalised cog deficits
Baddeley and Hitch (1974) examined effects of concurrent digit load on reasoning, comprehension and learning tasks
o E.g. “A not preceded by B —– AB”
o Subjects able to do all things with concurrent memory load (albeit w/ reduced cap)
o Argues against idea of unitary ST store
ev for separable phonological and visuospatial stores - Baddeley and Hitch
Articulatory suppression effect
o Subvocal rehearsal impairs memory for words but has no effect on perf of visuospatial task, e.g. chess
o In contrast, chess perf impaired by perf of concurrent visuospatial task
o Suggest existence of 2 separate systems for temp storage of info, one phonological and one visuospatial
ev for multi-component model
Evidence for the existence of a phonological store:
o Phonological similarity effect
Phonologically similar words (e.g. ‘fan’ and ‘ban’ harder to remember than dissimilar words (e.g. ‘fan’ and ‘cot’)
Effect doesn’t occur for semantically similar words
Suggests phonological code not semantic code used for temp storage of words
o Word length effect
Immediate memory span for short words greater than long words
Again suggests phonological coding system used for ST storage of words
Baddeley and Hitch (1974): WM
CE = ‘working’ component of WM
“The most imp but least understood component of WM” (Baddeley, 2003)
Coord of resources, attentional control, processing and manip of stored info
Basis of human intelligence
see notes
general int (Spearman’s ‘g’ factor)
1904, published paper examining correlations in children between diff disparate measures – academic ability (ratings from teachers, perf in exams etc.) and sensory discrim
Found correlations all pos
Correlation between sensory ability and academic ability almost perfect
FA revealed underlying factor common to perf of many diff kinds task (‘g’ factor)
Later adapted to form 2 underlying factors
o Lots of research on Gf as seems to form basis for human intelligent behav
o Tested using problem solving tasks – e.g. which of 4 shapes should go in empty square
Relevance to WM?
see notes
Kyllonen and Christal (1990)
Asked Q: “What underlies our ability to perf reasoning tasks of kind measured by Gf?”
Gave subjects reasoning tasks (see notes)
And also given WM tasks (see notes)
indv diffs in WM - Kyllonen and Christal (1990)
Close correspondence between WM and intelligence – suggested intelligence no more than WM
Other functions such as general knowledge didn’t correlate highly at all
Fluid intelligence basically problem solving
Crystallised basically acquired knowledge
WM correlates with fluid but not crystallised
Correlation between 2 types tests (logical reasoning and WM) = .76
Why so high?
WM tasks used require not only maintenance of info in WM but also online manip of info
o May be ‘exec’ component that underlies Gf?
see notes
ev that the ‘working’ component of WM correlated w/ fluid int - Kane and Engle (2002)
2 tasks differentially load onto ‘working’ component of WM
Complex span required retention of info in WM and in addition requires active processing, manip and updating of info
Simple span only requires retention of info in WM
see notes
Raven’s matrices - Kane and Engle (2002)
Test of general intelligence/fluid intelligence/’g’
see notes
structural equation model
see notes
Can measure r’ships between variables
On left = indv tasks given to Ps 0- complex span tasks at top, simple in middle and other at bottom
Structural equation model identified 3 factors underlying perf of tasks – 3 independent cog processes differentially engaged by tasks – WM, STM and processing speed
Examined extent to which factors correlated with measure of fluid (/general) intelligence (Gf)
Correlation between WM and Gf almost perfect
Correlation between STM and Gf neg
Correlation between speed and Gf low
Even when perf matched across diff types task, found same pattern of results
‘Working’ component of WM predicts fluid intelligence
Simple STM doesn’t
Fluid/general int involves ‘exec attention’ component of WM
o “Memory representations maintained in highly active state in presence of interference, and representations may reflect action plans, goal states/task-relevant stim in env” (Kane and Engle, 2002)
what is it about WM that makes it imp to int?
Manip of info in memory
/ construction and use of set of task rules?
rule WM task - Duncan et al. (2012)
see notes
Also gave Ps complex WM span tasks: solve each maths problem and say word aloud – then recall all 3 words at end
culture fair IQ test
see notes
correlation between rule WM and fluid int
see notes
Found strongest correlation between rule WM and int
Other types WM, such as complex span (digit, spatial/operation) correlated w/ fluid int but not as strongly
While process of manip/processing info in WM might be key component of fluid intelligence, it’s the construction and use of set of task rules that underlies indv diffs in fluid intelligence
Duncan (2013): effective fluid int involves construction of ‘mental program’ for task perf – subdivide goals into sub-goals to break down complex problems into manageable chunks
Problem seems complex at first but simpler if divided into sub-goals
3 features that define items in display – colour, shape and size
Effective problem solving involves not only maintenance of info in WM and manip of info
Also requires construction of effective ‘program’ specifying series of steps (‘rules) to get from state A to B
Effective WM requires not only maintenance of rules but also use
Failure to use rules results in ‘goal neglect’
goal neglect - Duncan et al. (2008)
see notes
Ps required to watch letters and numbers appearing in centre of screen and report all letters from one side
Then, when saw arrow, should switch sides and report letters from other side
Errors = keep reporting from same side even when cued to switch
Goal neglect rare in some and more common in others
Those when common had IQ scores at lower end of scale (80-100)
Subjects all reported knowing what supposed to do
Rules never explicitly forgotten – Ps could always report at the end of the task that they were supposed to switch sides when they saw a cue
Ps capable of following task instructions – error feedback generally corrected perf
Following of rule, translation of rule into behav, that was lacking
In follow-up exp, Ps perf 2 blocks trials w/ either letters or no’s as targets
One group given full instructions (have to detect either letter/no. targets)
Other group given partial instructions – only told about no. targets after had completed first block of letter targets
Both groups perf exactly same task but differed according to complexity of overall set of task rules
results
see notes
Goal neglect higher in subjects given full instruction
Increasing task complexity had effect of reducing perf in high IQ subjects to level of low IQ
Suggests overall complexity of mental program required to complete task that is imp when considering link between WM and fluid intelligence
