Lecture 4 - WM capacity and variation

general intelligence

-performance on all cognitive tasks stems from a single factor → g

-but now idea of more than one general factor for intelligence:

• crystalised intelligence

• fluid intelligence

crystalised intelligence

-putting learned knowledge to use correctly

-vocabulary

-applying previous knowledge and experience practiced before

fluid intelligence

-ability to reason through and solve novel problems

-requires understanding of rules but adaptive understanding to create solutions to novel issues

linking fluid intelligence to WM

-all reasoning tasks require the construction of new structural representations which is limited by the capacity of WM

-storage capacity enables people to maintain distinct chunks of information and flexibly construct task-relevant bindings among them

Miller’s magical mnumber 7

-participants memorised series of letters in sequence

-one average get 7 ± 2 correct

boosting STM capacity

-active rehearsal

-chunking

Cowan’s magical number 4

-only representations in the FOA are available to conscious awareness and report

-on average the capacity limit of adults’ FOA is 4 ± 1 information elements

set size effect

-in simple span task the longer the set the harder it is to recall items, no manipulation so easy to remember 7

-in complex span task give memoranda and ask to do processing between them → rapid decline in holding onto information, 2 items perform very well and then after 5 drop off considerably

WM capacity hypotheses

1. decay

2. interference

3. limited

decay (working memory capacity)

-hold onto information and as a function of time will lose information

-longer we try to hold onto items in WM they will disappear

-but the passage of time causes nothing itself rather is is correlated with processes that cause forgetting

restoration mechanisms (decay)

• rehearsal → subvocally repeat memoranda to maintain them

• refreshing → think of memoranda to keep memory traces active

time-based decay

-representations in WM get weaker over time → if true would expect memoranda 1 in list (M1) to have worse rehearsal

-but instead have primary and recency effects

-so time alone doesn’t cause memory to decay and there must be other impacts that happen during time

interference (working memory capacity)

-WM is limited by mutual interference between representations

transience (interference)

• proactive interference → older memories impair the retrieval of new memories

• retroactive interference → new memories impair retrieval of older memories

types of interference

1. confusion

2. superposition

3. overwriting

confusion (type of interference)

-ask participants to encode position of 4 items

-masked then ask to recognise if item is in the correct position

-whether get it right or wrong is due to interference as confused item for another item

-more items the more confused you get

superposition (type of interference)

-binding together representations from interference by two different concepts

-presented with orientation of shape to encode

-presented with masks and distraction image

-then have to recall original position

-participants often recall combination/average of the distractors and memoranda

-so binding together and averaging out

-less similar items the worse the performance tends to be

capacity and resource (working memory capacity)

-WM capacity is determined by a limited quantity of resources that enables holding representations available

discrete allocation of resources (working memory capacity)

-allocation to a limited number of items

-no information stored about additional items

continuous allocation of resources (working memory capacity)

-equal spread of resource among all items

-fewer resource per item for larger arrays

resource models (capacity and resource)

-resource is a limited quantity that enables a cognitive function or process

-probability of success increases the larger the amount of resource assigned to it

slot models (resource models)

-resources are distributed in discrete units → defining the number of items one can store

-quality of the retained representations is not perfect but is sufficiently high

flexible-resource models (resource models)

-resources are distributed flexibly, allowing for either:

• a small number of high quality objects

• a high number of low quality objects

evaluation of resource models

-no theory gives comprehensive explanation and support conflicts with each hypothesis

Findings	Decay	Resource	Interference
Set-size effect  + complexity of items	-	+	0
Set-size effect when delay is 0 s	-	++	++
Domain specificity (lists from different domains better remembered)	+	+	++
Cross-domain set-size effect	+	++	0
Heterogeneity benefit	-	-	++

variation in WM

-general distribution have a mean of 4 items

-capacity in greater in:

• older children than younger children

• younger adults than older adults

• healthy people than people with frontal-lobe damage

WM and complex cognitive activities

-WM correlates with complex cognitive activities:

• reading comprehension

• reasoning

• problem solving

-predicts:

• cognitive development

• individual differences in intellectual abilities

measuring variation

-measuring individual differences in cognitive abilities

-letter updating task → always remember the last 3 letters → so have to shift and change list

task impurity problem (measuring variation)

-any task that assesses a cognitive ability most likely demands other abilities that are needed to process the structure and materials of the task

latent variable modelling (solution to task-impurity problem)

1. select multiple tasks that seem different on the surface but capture the same target ability

2. statistically extract what is common among those tasks - draws correlations between different variables and identify one variable that links them together

3. use the resulting variable as a measure of target ability

correlations with variation in WM

-reasoning

-attention

-reading

-storytelling

-vocab learning

differences in WM

-proposed explanation:

• executive attention hypothesis

• binding hypothesis

executive attention hypothesis (differences in WM)

-single top-down executive attention system underlies both WM and reasoning task performance

-responsible for maintenance and disengagement

-two systems:

1. quick, easy access to all info that you know

2. controlled, effortful processing of information → attention control system

maintenance (executive attention hypothesis)

• WM task → access to relevant information and append new information to the list

• reasoning task → disengage from outdated hypotheses and prevent returning to them

disengagement (executive attention hypothesis)

• WM task → disengage from and suppress outdated information from previous trials

• reasoning task → maintain problem and allow systematic hypothesis testing

limitations of executive attention hypothesis

-executive attention tasks do not correlate well - so it is difficult to directly test this hypothesis

binding hypothesis (differences in WM)

-a system for rapid formation of temporary bindings underlies both WM and reasoning task performance

-series of bindings that construct and manipulate representations of novel structures

-bindings are temporary links of content representations to places in a mental coordinate system

-WM capacity limit is the number of bindings maintained, arises from interference between bindings

-people who suffer less interference can build more complex structural representations so perform better on cognitive tasks

binding in a WM task (binding hypothesis)

-bind representations to a position

-flexible and can be manipulated/shift letter position, remember letter position or just remember letter

-so WM capacity is number of bindings we hold and manipulate at any one time

binding in a reasoning task (binding hypothesis)

-bind understanding to a semantic space

-binding predictors and outcomes

-limited by number of WM bindings we have

-binding temporarily

limitations of binding hypothesis

-bindings may be constructed and maintained with the help of executive attention

-makes it difficult to directly test this hypothesis against the executive attention hypothesis