terms and models - TERM 2 Flashcards
iconic and echoic memory
sensory memory
iconic = visual info
echoic = acoustic info
working memory
storage and manipulation of information
flexibility
arbitrary connections between items
limited capacity
multicomponent model of WM
CE as homunculus
visuospatial sketchpad, episodic buffer, phonological loop = subvocal rehearsal through articulatory loop
assumptions of multicomponent model
central executive = flexible allocation of attention
storage systems = domain specific STM
episodic buffer = binds information from different sources
problem = CE is homunculus (not explained)
word length effect exp
recall shorter words easier than longer words as refreshed quicker within 2 seconds
decay if not refreshed
phonological similarity effect
recall is worse when items sound similar
words that are semantically similar have no effect on WM - means that WM coding is phonological (only affected by sound of word not meaning)
articulatory suppression
asked to utter irrelevant word while presented with words to remember
stope subvocal rehearsal
word length effect doesn’t exist with visual presentation - only auditory (if someone reads the words aloud to you)
because words enter straight to phonological store
semantic relatedness
improves recall when related
interference can strengthen semantic link between items
deafness
have sign-based phonological store
use manual articulatory rehearsal mechanisms to refresh information in phonological store
visuospatial info
doesn’t integrate with phonological loop except in the episodic buffer
prediction that visual and spatial stores are separate supported
mental rotation task
presented with pairs of objects and asked to decide whether they are identical or mirror images of each other by mentally rotating one of the objects to align it with the other
blind participants generated spatial representations just as good
Klauer and Zhao
memorised dots on a grid (spatial) or Chinese characters (visual)
visual interference tasks affected visual task (dots)
spatial interference tasks affected spatial task (character)
= competition
domain specificity
complex span task
predicts lower recall for same-domain (overloading)
combination of verbal and visuospatial materials
Vergauwew - no effect
decay
info gets weaker over time = time-based decay
restoration mechanisms = rehearsal and refreshing
forgetting may be due to events rather than time
focus of attention
only representations in the focus of attention are consciously available
capacity = 4+/-1
cowans embedded process model
WM holds limited info - heightened state of availability
LTM has an activated portion holding relevant information for current cognitive task (small)
WM has narrow focus of attention - excludes irrelevant information
what limits working memory
decay
interference
limited resource
interference
types=
proactive = older impair new memory
retroactive = new impair old memory
confusion - similar info competes for retrieval
superposition - new information (that looks similar) encoded on top of existing info
overwriting -new info (that sounds similar) replaces stored info
limited resource
resourced flexibly allocated and in discrete (limited number of items)/continuous (equal spread of resource to all items) units
slot models
resources are distributed in discrete units
quality not perfect but high
flexible resource models
distributed flexibility
either SMALL number of `HIGH quality objects
or HIGH number of LOW quality objects
why does WM capacity vary?
executive attention hypothesis - differences in ability to control attention
binding hypothesis - encoding information simultaneously. Capacity relies on number of bindings maintained
more bindings = better WM
binding hypothesis - DETAIL
bindings are temporary links
WM capacity limit = number of bindings maintained, arises from interference
less interference = more complex structural representations
difficult to test against executive function hypothesis as bindings may be maintained by executive attention
executive attention hypothesis - DETAIL
2 systems in brain in which info us used, engaged with:
system1 = quick easy access
system2 = controlled, effortful (ATTENTION CONTROL SYSTEM
we have one executive function that underlies WM and reasoning through maintenance (keep relevant info) and disengagement (getting rid of old info)
transfer effects
improvement on practice task lead to improvements on unpracticed task.
improvements due to strategy based training (task specific e.g mnemonics) or process-based training (transfer to other contexts, complex span tasks)
functional overlap
improvement expected if practice and no practice tasks share underlying processes
measuring training effects
performance at pre-test compared to performance at post-tests
put against active control group (help with placebo)
e.g
n back task
complex span task
Klingberg training study - WM
children with ADHD
training programme with WM tasks
used ravens progressive matrices - test reasoning
big pre-post difference in intensive training for reasoning compared to active control
-uncorrected differences in change are only small (small sample)
Klineberg study - repeated
larger sample
training was adaptive (changes as you improve) rather than high v low dose
larger benefit in adaptive training
limitation - experimental baseline was higher before exp
near transfer v far transfer
near = transfer of skills to a task closely related (same underlying process)
far = transfer of skills to a task not closely related
Redick et al
no significant near or far transfer effects in spatial and verbal reasoning tasks
limitations of WM training
insufficient evidence
lack of active controls
small sample sizes
lack of active controls
lack of theoretical frameworks
multiple sources of variances framework
training affected by intervention specific factors, individual differences
mechanisms of transfer:
enhanced capacity - training increase info held in WM (training leads to broad transfer effects)
enhanced efficiency - more efficient use of training = selective transfer effects
de simoni - mechanisms of transfer study
binding task
updating task
visual search task
no evidence for near or far transfer effects
differences in training benefits
magnification - people with higher ability gain more - larger improvement seen in younger adults in initial task
compensation - people with lower ability gain more - use different strategies for same outcome in older adults
cogmed
WM training programme
study:
larger improvements in verbal near transfer tasks
n-back yields larger for far transfer
lexical characteristic that affect speed of access
word length
frequency of words
neighbourhood density - lots of neighbours= slower processing
spreading activation
facilitates predictions of words next appearing via activation of items that are related to acoustic input
challenges for lexical access
accents
speech is a continuous stream
co-articulation
homonyms (same sound, different meaning)
ambiguous word boundaries (only fools and horses - four candles/fork handles)
categorical perception
ability to distinguish between sounds on a continuum based on voice onset times
Ericcson study
increase in memory span from 7 to 79 digits with 230 hours of brain training
bottom up processing
process by which speech sounds initially analysed and recognised based on acoustic features
top down processing
use of linguistic knowledge and contextual clues to facilitate recognition of speech sounds
mechanics of lexical access
- Gradual activation of the word that matched the sound
- Activate all words that match same start sound of a word and gradually de-activate words that no longer match sounds
- Gradually activate the matching word that relates more than other words
cohort model
bottom up processing
we access words in lexicon via activation of words sharing initial features and gradually de-activate words that stop matching = uniqueness point
= neighbourhood effects (similar words compete), frequency effects
gating experiments
shown fragment of words that gradually reveal whole word
asked to guess what the word is
- aligns with assumptions of cohort model
architecture of cohort model
speech input > lexical item
facilitatory signals are sent to words that match
inhibitory signals are sent to words that do not match
phoneme restoration effect
don’t need to hear all the phonemes to understand word
doesn’t align with bottom up processing
cohort model - 3 stages to word recognition
access - acoustic phonetic info mapped
selection - candidate words that mismatch are deselected
integration - semantic, syntactic properties of word are checked against the sentence
cross modal priming
prime word is auditory
target word is visual
shorter RT when words are related
then do the same but with fragments not full words
biasing the sentence had no difference in priming effect - only when given the full word not fragments
context in the cohort model
Sentence context doesn’t influence the process of lexical access – integration is affected by sentence context
items that match acoustic input but not sentence context are activated - context only relevant when reached uniqueness point
priming paradigm
what we did in RM
prime word then target word
unrelated/related
uses spreading activation
TRACE model vs cohort model
TRACE emphasises top down processing while cohort minimises its impact
cohort predicts lexical accès is bias to activation of words with shared onsets
TRACE accommodates activation of rhyming competitors
TRACE provides no account on context
uniqueness point
point at which other candidates have become deactivated
TRACE model
Features activate phonemes that activate words
more matching features = more activation = correct word
radical activation model
architecture of TRACE
hierarchical network of nodes (facilitatory connections): features, phonemes and words = dominant bottom up processing
opposite direction = top down
top down processing increases activation of phonemes and features
visual word paradigm
eye tracking study
showed words overlapping phonology that don’t start with same onset as speech input, are activated in speech perception
results:
rhyming competitor receives activation (looked at)
TRACE - top down
faster identification of letters in words rather than nonwords
other evidence:
could detect phonemes in nonword that were word like