C5 working memory

Question 1

What is working memory/definition

Answer 1

Working memory enables us to keep things in mind for short periods (2-15 seconds) as we think, e.g. while reading, making a list etc.

It’s related to but different to short-term memory (STM) and long-term memory (LTM).

The concept of ‘span’ means how many items from a briefly presented set can be remembered, e.g. ‘word span’ is the number of words that can be recalled if read a list of say 20 words. Digit span, operation span, reading span etc. are similar tests.

Question 2

Models evolving over time - Atkinson & Shiffrin (1971)

Answer 2

Modal: two-store; STM as a unitary store, the controller for LTM that encodes and moves information in and out.

Question 3

Models evolving over time - Baddeley & Hitch (1974)

Answer 3

STM is part of working memory, not an LTM processor, working memory is a tripartite, resource-sharing system (articulatory loop, visuo-spatial scratchpad, central executive), involved in complex cognition not just memory.

Question 4

Models evolving over time - Baddeley & Hitch (1984)

Answer 4

Articulatory loop concept fractionated into phonological loop (phonological store and subvocalisation).

Question 5

Models evolving over time - Baddeley (1986)

Answer 5

Adopted Norman & Shallice (1986) two-store Supervisory Attentional System (SAS) model as a model for the central executive (so an attentional instead of resource sharing system).

Question 6

Models evolving over time - Baddeley (1996)

Answer 6

Abandoned storage in the executive, instead proposed it’s a purely attention control system fractionated into modules for focusing attention, dividing and switching attention and using attention for LTM access.

Question 7

Models evolving over time - Baddeley (2000)

Answer 7

Added the multimodal episodic buffer that links up information about a single object spread across the memory subcomponents, to address the binding problem. Proposed conscious awareness is the mechanism by which this is retrieved.

Question 8

Memory is multi-faceted

Answer 8

STM: used to remember information for short periods, typically less than a minute e.g. keeping track of a conversation, remembering a phone number someone’s called out while dialling it;

Working memory: organises tasks into subtasks and tracks transient information as we do these e.g. doing mental arithmetic.
•Hitch (1978) suggests we can forget information while we sequence the subtasks.
• Errors due to overload show it’s a limited capacity system.

Question 9

Distinction between STM and LTM

Answer 9

Early studies showed STM and LTM were more likely to be different systems rather than different parts of one memory system:
• Effect of phonemic vs. semantic similarity: More errors were made when immediately recalling (implies use of STM) a list of similar sounding words than one of words with similar meaning. This effect is reversed if recall is tested after a longer delay (implies use of LTM) - Baddeley;
• Rate of forgetting: We remember information that is ‘learnt’ for longer than information we just ‘hear’ (committing to LTM vs. keeping briefly in STM) - Brown;
• Span of immediate memory: when presented with running lists we can only remember a small number of items at a time (about seven) - Miller.

Question 10

Distinction between STM and LTM - modal models

Answer 10

Modal models of the memory system (e.g. Atkinson & Shiffrin, 1971) agreed that:
• Memory is a two-store system;
• STM (or short term store) is the ‘gateway’ that manages transfer of information in and out of LTM, storing information briefly while processing (encoding/decoding) and moving it in and out of LTM;
• Subvocal rehearsal and other techniques are used to avoid forgetting it during this processing.
• They assumed that STM acted as a working memory that stored and processed information, and that it was essential for cognitive activities (reasoning, comprehension, calculation etc.).

They have been challenged because:
• Estimates of STM capacity differed widely depending on how it was measured, without any explanation;
• Neurological evidence undermines the claim that STM is necessary for LTM to function:
• Shallice and Warrington (1970) reported the case of K.F. who had only a two-digit auditory span due to brain damage from a road accident, hence impaired STM.
• However he tested normally for long-term learning and memory, and had no problems understanding normal speech, so spared LTM.
• This suggests STM and LTM are distinct and normal STM is not required for LTM to function normally.

Question 11

Garden path sentences

Answer 11

These show we do retain words in memory as we process them (building up and interpreting sentences) but there are competing theories:
• Just and Carpenter (1992) claimed individual ability to hold multiple interpretations depends on working memory capacity;
• Caplan and Waters (1999) argued working memory isn’t involved as comprehension is done by a separate system.

Question 12

Is WM more than STM? - Baddeley and Hitch

Answer 12

Baddeley and Hitch investigated whether STM acts as working memory:
• They used a dual-task paradigm (if two tasks interfere with each other they may be competing for the same limited resource):
• Participants simultaneously did an STM test, remembering and repeating a six-digit sequence, along with one of three cognitive tasks: reasoning, language comprehension or list learning;

They found:
• load in the STM task adversely affected cognitive performance, although a small number of items could be remembered without affecting the main task much, suggesting that information might be transiently stored and processed simultaneously by working memory, and that there may be two systems involved, one for storage and one for processing;
• even when STM became overloaded (participants couldn’t repeat more digits than they could remember at a time - memory span), cognitive performance didn’t fail, implying that working memory includes extra resource that is not used by STM.

They concluded that:
• STM tasks requiring information to be stored briefly compete with cognitive processing tasks for limited working memory resource;
• STM is more like a part of working memory than a controller in front of LTM (modal models).

Question 13

Is WM more than STM? - Daneman and Carpenter

Answer 13

Daneman and Carpenter (1980) investigated the relationship between STM and working memory by studying the relationship between participant performance on tasks that were thought to use the same psychological processes (if they do use the same underlying abilities performance should be similar on each):
• They argued that word and digit span testing does measure ability to store items (STM) but not how well individuals can balance storage and processing (working memory) because they don’t have to do both simultaneously;
• They proposed an alternative test of reading span (read sentences = processing, remember final words = storage);
• Three sets of cards were used at each test level, each card in the set had a sentence on it and each set was made up of 2-6 cards (for example three sets with two cards in each at the first test level, at the next level three sets with three cards in each etc.);
• Starting with the 2-card sets, participants were shown each card in the set and had remember the final word of the sentence on each; once they had been shown all the cards in a set they had to recall the final words for that set.
• When all three sets at a level were done, they moved up to the next level, e.g. 3-card sets.
• Reading span was measured as the highest level where they got two sets out of three correct, i.e. if they got more than one set correct on 2, 3 and 4-card sets but only one set correct on the 5-card level their reading span was 4.
• They found reading and listening span were much better predictors of comprehension than word span;
• They concluded that working memory (storing and processing information simultaneously) is an important factor in the individual difference of reading comprehension ability measured using verbal SATs etc.

• This supports Baddeley and Hitch’s conclusion that STM and working memory are different things.
• This study was criticised because their task measures (reading/listening span and comprehension) all involve language processing, but word span testing does not, so the test may not be a valid measure of working memory, just of language processing.

Question 14

Structure of WM - Baddeley and Hitch Tri-partite model

Answer 14

Baddeley and Hitch found:
• phonemic similarity slightly worsened performance in both the reasoning and comprehension dual-task tests (e.g. comprehension of sentences with rhyming words was worse than non-rhyming);

They explained this by proposing two working memory modules:
• an articulatory rehearsal loop that can store 2-3 speech-based items (matches the earlier concept of STM);
• a central executive that can process information (reasoning, learning etc.) and store information while it’s being processed, where processing and storage compete for limited resource - more storage implies less resource for processing and vice versa.
• interference in dual-task tests where both tasks were verbal or both were visuo-spatial, but not when there was one of each:
• suggests separate systems are responsible for verbal and visuo-spatial processing.

This is supported by neurological evidence:
• Corsi span is a visuo-spatial test where the tester points to a sequence of blocks and the participant has to repeat the sequence in the right order, where span is the number they can do correctly;
• De Renzi and Nichelli (1975) found different lesions were associated independently with Corsi span and auditory digit span implying different systems were involved in verbal and visuo-spatial processing.

This led to the original and influential tri-partite model containing an articulatory loop, central executive and visuo-spatial scratchpad.

Question 15

Structure of WM - Baddeley and Hitch Tri-partite model - challenged by

Answer 15

Hitch et al. showed visual features are used in some cases to form mental images:

Participants were better able to mentally combine line drawings when they shared figure/ground colours, showing that their mental representations contained visual features:

The visuo-spatial (non-verbal) store might be used to form mental images for verbal items so that they’re easier to remember (e.g. method of loci/memory palace).

Baddeley and Lieberman found evidence in dual-task tests that mental images are spatial rather than visual (i.e. contain information about location of objects, not just what they look like):
• memory of images was disrupted in dual-task tests where the secondary task was spatial (e.g. tracking a moving sound while blindfolded) but not where it was visual (e.g. report if a blank field changes brightness).

According to Logie we have both visual and spatial systems, with a spatial system that rehearses visual features in a similar way to how the articulatory loop rehearses verbal ones.Smyth and Waller proposed something similar for rehearsal of skilled movement (e.g. touch-typing)

Question 16

Structure of WM - Baddeley and Hitch Tri-partite model - challenged by

Answer 16

Baddeley (PAPER)

Baddeley (2003) notes that different representations may be used depending on the memory task, including motor or kinesthetic as well as visual or spatial;

Patients with Williams Syndrome showed normal test results for visual concepts but gross impairment on spatial ones (“above”, “below”) so may have visuo-spatial scratchpad issues;

The visuo-spatial sketchpad isn’t as important to language as the phonological loop but may also be involved in reading, for example tracking page layout so we can move from one line of text to the next.

Question 17

Structure of WM - Baddeley and Hitch Tri-partite model - competing accounts proposed

Answer 17

Baddeley and Hitch proposed that LTM and working memory are distinct systems, others such as Cowan proposed that working memory is a component of LTM because working memory performance is affected by what we already know (e.g. chess experts with prior knowledge in LTM are better at solving chess problems that use working memory)

Both agree that there’s a separate central executive;

Jones points out that assuming separate subsystems introduces a binding problem - if the features of an object are handled by all these separate systems, how do we re-assemble their outputs to create a unified representation of the object ? (see section 5.2.3.4 for more on this)

Question 18

Phonological working memory - evidence for

Answer 18

Experiments provide evidence for how the auditory loop might work and be affected by other factors:
• Phonemic similarity reduces performance of STM tasks within dual-task tests
• Baddeley et al. (1975) demonstrated that people who spoke faster could recall more words.
• This suggests word length also reduces performance, not number of items as was previously thought - it seems the limit might be the number of words the participant can rehearse in around two seconds;

Question 19

The auditory loop theory explains some of these effects:

Answer 19

The phonemic similarity effect: if an item in the list is forgotten due to the memory trace decaying, the fact that it sounds like all the others doesn’t help recall but if each one sounds different, remembering what it sounded like might improve recall;

Articulatory suppression: recall in a dual-task test is impacted if the secondary task is to repeat an irrelevant sound over and over - “the the the” - because this consumes some of the resources of the articulatory loop which are then unavailable to the primary recall task:
• memory span is reduced;
• word-length effect is eliminated in both auditory and visual stimuli (i.e. performance is equally bad with longer and shorter words);
• phonemic similarity effect is reduced with visual stimuli (e.g. words to be read).

Question 20

Phonological working memory - auditory loop/phonological loop

Answer 20

Baddeley et al. showed suppression continued into the recall part of auditory tests removed the word length but not phonemic similarity effect.

They proposed a model where the articulatory loop concept is replaced by the phonological loop, composed of:
• a phonological store that holds a small amount of what’s seen/heard and
• a continuous subvocalisation process that’s involved in rehearsal of both types of stimulus and that encodes visual stimuli like letters or words into auditory ones;

While auditory stimuli can be stored directly in the phonological store, visual ones need to be encoded by the subvocalisation process first;

Phonemic similarity affects the phonological store; word length affects how much can be rehearsed by the subvocalisation process;

Question 21

Phonological loop

Answer 21

The phonological loop model offers explanations for other aspects of working memory as well:
• Developmental and cross-linguistic differences;
• The irrelevant speech effect;
• Evidence from neurological studies;
• Theoretical issues.

Question 22

Phonological loop - Developmental differences

Answer 22

Faster rehearsal means better recall (more can be rehearsed in the 2-second interval):

Developmental differences:
• As children’s rate of speech increases their recall span also increases (Nicholson, 1981);
• Word-length effect decreases as the time children take to say different length words decreases;

Question 23

Phonological loop - Cross-linguistic differences

Answer 23

Mean digit span of speakers of different languages is related to how long the words for digits in those languages take to say

Note: these are not causal relations, they’re correlations.

Question 24

Auditory stimuli access the phonological store directly, visual ones need to be encoded through subvocalisation:

Answer 24

Younger children have poorer recall of spoken stimuli with long names than the same stimuli presented as pictures; older (seven-year olds) show a word-length effect for both modalities (Hitch et al.);

Younger children show poorer recall performance on visually similar stimuli; older children show poorer performance on visual stimuli with phonemically similar names (Hitch et al.).

Question 25

The irrelevant speech effect (background speech disrupts recall of visually presented words)

Answer 25

Salame and Baddeley observed that irrelevant speech is more disruptive/harder to ignore than random noise in a task to remember visually presented words:
• Suggests unattended speech enters the phonological store (blocking the subvocalised word stimuli) but other noises don’t;
• Suppression eliminates the irrelevant speech effect, maybe by blocking the word encoding.

Macken and Jones found contradictory evidence that unattended non-speech (tones) can cause the same disruptive effect, and suggested it might be due to broader memory mechanisms, not just the phonological loop.

It’s possible that both theories are correct, but in different situations.

Question 26

Phonological loop - Neural basis - Vallar and Baddeley

Answer 26

Vallar and Baddeley studied P.V. who showed a selective verbal memory deficit:
• a left parietal lesion due to stroke had resulted in her only having a two-digit auditory span, but fluent speech and normal rate of speech;
• she showed poorer recall with phonemically similar words (suggesting phonological store damage) but no word-length effect (also supports a phonological damage diagnosis as subvocal rehearsal wouldn’t help recall of shorter words over longer ones);
• she also showed better memory span for visually presented words than for spoken words, and this showed no phonemic similarity or word length effect, suggesting she may have compensated for phonological look damage by using her visuo-spatial scratchpad more to memorise visual stimuli;
• This is an example of single dissociation that suggests that storage and rehearsal are separate.

Question 27

Phonological loop - Neural basis - Paulesu et al.

Answer 27

Paulesu et al. used PET scans taken during three different tasks to look for the physical location of the phonological loop:

Task 1
• Say if the final letter of a sequence had occurred before
• Requires phonological loop (phonological store and subvocal rehearsal)
• Activated Broca’s area, the supramarginal gyrus of parietal cortex and “other areas”

Task 2
• Say if the final letter of a sequence of Korean letters had occurred before
•Does not require the phonological loop
• Activated the same “other areas” as Task 1 but NOT Broca’s area OR supramarginal gyrus

Task 3
• Say which letters in a sequence rhymed with ‘B’
• Requires subvocal rehearsal but not storage
• Activated Broca’s area but NOT supramarginal gyrus

Results of Task 1 and Task 2 imply the phonological loop may be associated with Broca’s area and the supramarginal gyrus;

Results of Task 1 and Task 3 imply rehearsal may be associated with Broca’s area, and the phonological store with the supramarginal gyrus.

However these conclusions can only be correct if the “Task assumptions” are also correct !

Question 28

Phonological loop - Theoretical issues - The Baddeley and Hitch (1974) phonological loop model has some limitations:

Answer 28

• it doesn’t explain irrelevant speech effect;
• it’s claim that word-length is related to the number of words that can be rehearsed has been challenged - Lovatt et al. (2000) claimed that word complexity was responsible, and that word-length effect could be eliminated by ensuring complexity was the same;
• Hulme et al. (1984) challenged the claim that rehearsal is affected by word-length by showing that four-year old children who apparently have not developed rehearsal ability at that age still showed word-length effect;
• Cowan et al. (1992) showed word-length effect being caused by output delays (which don’t affect rehearsal).

Question 29

Phonological loop - Theoretical issues - Baddeley (2000) postulated the existence of a fourth component - the episodic buffer:

Answer 29

• Articulatory suppression that should ‘destroy’ recall by occupying all the rehearsal/subvocalisation resource thereby preventing visual stimuli registering in the phonological loop, in fact only has a weak effect. This suggests some other store exists;
• Evidence from tests show that visual similarity of word stimuli affects span, implying that visual and phonological information are bound somehow, suggesting some system must exist to link them;
• Very amnesic patients (damaged episodic LTM) can keep current information in mind that is not in LTM (learned): e.g. Tulving reported a ‘densely amnesic’ patient who could keep track of the objective within an individual game of Bridge, and also of the cards that had been played and the scores across a series of games. If STM and LTM are not involved, some other store must exist.

Question 30

Phonological loop - Theoretical issues - ways to address limitations to the model

Answer 30

• Creating a completely new model from scratch;
• Revising the model and address its deficiencies;
• Using computational modelling to create alternative models (section 5.4)

Question 31

Central executive/Executive process overview

Answer 31

The central executive coordinates working memory processes but it’s not clear exactly which ones or how:

Baddeley & Hitch proposed it controls ‘slave stores’ (phonological loop, visuo-spatial scratchpad) and interacts with LTM but it’s hard to identify what exactly it does do;

One difficulty is explaining what controls the central executive (if it controls everything else) ?

DD303 looks at three aspects of the central executive studied by researchers: the central workspace, attention and fractionation.

Question 32

Central executive/Executive process - central workspace

Answer 32

Resource-sharing theories (Baddeley and Hitch, Daneman and Carpenter) proposed that the central executive combined storage and processing, and that if more of its resources were taken up by one function, the other function would have less to work with - e.g. reading complexity (process) affects word span (storage);

Towse et al. tested this hypothesis in children:
• found no clear evidence to support it;
• devised a task-switching hypothesis in which word span was limited by the words participants forgot as they switched between reading and storing:
• Tested this by presenting items of varying complexity in different orders (same list => same total processing time);
• Found having to remember items for longer resulted in lower reading, counting and operation spans.

Others have found that task complexity affects span, and deBeni et al. (1998) found ability to inhibit interfering information affects working memory span.

These results suggest that working memory span is affected by much more than just central workspace capacity, and may be fractionated (made up of separate modules).

Question 33

Central executive/Executive process - attention - Norman and Shallice

Answer 33

Norman and Shallice proposed a two-level explanation of working memory that explained utilisation behaviour (patients with certain frontal lesions were unable to inhibit pre-potent stimuli, e.g. when a glass of water was placed in front of them they couldn’t prevent themselves from picking it up and drinking it):
• a low-level set of schemata or learned behaviours, any of which could be triggered by appropriate stimulus;
• a high-level Supervisory Attentional System (SAS) that normally intervenes to inhibit schema firing that would lead to inappropriate behaviour;

Utilisation behaviour occurs when this system is damaged; everyday action slips can happen if attention being distracted means the SAS fails to intervene (e.g. going into a room to do something and without thinking carrying out a completely different habitual action).

Question 34

Central executive/Executive process - attention - Baddeley

Answer 34

Baddeley (1986) proposed a new theory that the central executive might act by controlling attention following the SAS model, rather than as a resource-sharing system:
• Tested using a random generation task (randomly select one item per second from a set, e.g. 0-9);
• Observed participants generated runs and stereotypical combinations quite quickly;
• Found that task complexity and rate of generation resulted in less randomness (harder to inhibit stereotypical (pre-potent) sequences like “1,2” leading to “3”);
• Concluded that this was due to competition for limited resource.

Question 35

Central executive/Executive process - attention - Towse

Answer 35

Towse suggests random generation is a complex task that may involve more cognitive functions than just executive function.

Question 36

Central executive/Executive process - Fractionation - Baddeley

Answer 36

Baddeley proposed fractionating the executive into three cognitive modules, responsible for
• focusing attention (e.g. on a single task and ignoring interference);
• switching and dividing attention (e.g. between two tasks);
• using attention to interact with LTM.

Question 37

Central executive/Executive process - Fractionation - Double dissociation

Answer 37

Double dissociation suggests these are conceptually separable:

Alzheimer’s patients find concurrent tasks hard (switching/dividing attention), normal people can find difficulty focusing attention as they age. (Older people with Alzheimer’s do not provide useful evidence in this way …)

Question 38

Central executive/Executive process - Fractionation - Mikaye et al

Answer 38

Mikaye et al. tested attentional control in a study of individual differences:
• Three factors were tested: shifting attention, monitoring/updating information, inhibiting pre-potent response;
• A three-factor statistical model matched empirical results better than simpler models, supporting the idea that these executive functions can be fractionated, even though they’re different to Baddeley’s list;
• Follow-on studies gave unexpected results - suggesting we don’t really know what’s going on ! More research needed …

Question 39

Coherence and the binding problem

Answer 39

Fractionation raises a binding problem - if executive control is carried out by an array of separate systems, how are the items of information about an individual object that are spread across these linked together?

Question 40

Coherence and the binding problem - Baddeley and Hitch

Answer 40

Simple example from Baddeley and Hitch (1974) - how is visuo-spatial and verbal information about a specific object tracked across the visuo-spatial scratchpad and phonological loop ? (Remember Jones (1993) criticised this model because it fails to address the binding problem it raises);

Question 41

Coherence and the binding problem - Baddeley

Answer 41

Baddeley (2000) acknowledged the binding problem in his 1986 model and added a multimodal episodic buffer that in this model lnks up all the information about individual objects that is spread across the fractionated subcomponents (how convenient !).
• This limited capacity system stores multi-modal information from different working memory subcomponents and LTM;
• This relieves the central executive of having to store information, so in this model it just provides attentional control;
• The information stored in the episodic buffer is retrieved through conscious awareness.

Question 42

Coherence and the binding problem - Jones

Answer 42

Jones et al. argued that similar interference effects across modalities are evidence for a unitary memory system (information about an object is stored in a single ‘episodic record’ rather than being spread across a number of fractionated systems):
• irrelevant speech should compete with words to be remembered in the phonological loop but tones shouldn’t yet the irrelevant speech effect (section 5.2.2.2) suggests this happens;
• Irrelevant speech disrupts spatial memory span (Jones et al., 1995) but this shouldn’t happen if the visuo-spatial scratchpad and phonological loop are fractionated;
• In both cases, performance worsened with more interference.

Question 43

Coherence and the binding problem - contrasting explanations

Answer 43

A third system may affect both of the disrupted systems - e.g. Beaman and Jones (1997) found that irrelevant stimuli disrupt serial ordering ability. A fractionation-based theory might explain the cross-modal interference as due to the the serial ordering system affecting both the systems responsible for spatial attention and for verbal attention.

Also the unitary account doesn’t explain dissociations (Alzheimers/old age);

So the debate continues …

Question 44

Vocabulary acquisition

Answer 44

A word to be learnt must be broken into sounds (phonemes) and these must be retained in memory for long enough … a working memory task.

Vocabulary acquisition is studied using:
• neuropsychological evidence,
• individual differences (vocabulary size) and
• experiments.

Question 45

Neuropsychological evidence

Answer 45

Baddeley et al. showed that P.V. had unusual difficulty learning novel words (e.g. in foreign languages)
• She could not learn Italian/Russian word pairs but could learn Italian/Italian pairs;
• As her native language was Italian, this suggests her long-term learning was intact but her short-term memory was not;
• This dissociation also supports the theory that STM and LTM are separate, but that they are linked for phonological learning - they could be parts of the same functional or even physical system for example;
• However the phonological STM/LTM might be distinct from other STM/LTM pairings such as semantic memory.

Question 46

Individual differences

Answer 46

Baddeley et al. (1998) showed children’s auditory span and vocabulary size are correlated;

Children’s ability to repeat “non-words” (not actual words but have valid form) has been found to correlate better with vocabulary size than digit span (could be that bigger vocabulary gives better ability to repeat “non-words” or the other way around);

Other studies (e.g. Service, 1992) found measures such as auditory digit span correlated with ability to learn a second language.

Question 47

Experiments - Papagno and Vallar

Answer 47

Papagno and Vallar studied adults’ abilities to learn pairs of words and word/non-word pairs and found:
• poorer learning when phonemic similarity or syllable number of nonwords was increased, but no effect when the same changes were made to real words;
• articulatory suppression impeded learning of word/nonword pairs but not pairs of real words;

They concluded that this shows the phonological loop is needed to learn novel words beacuse only nonwords are affected (I assume this means that by “words” they really mean words the participant knows already as opposed to “non-words” which could be any phoneme sequences the participant hasn’t learnt before, even actual words);

However this could be specific to adults, children might use other systems to learn new words.

Question 48

Experiments - Gathercole

Answer 48

Gathercole found that nonwords were easier to learn if they contained phonemes that were common in participants’ native language (e.g. blonterstaping is easier for English speakers than perplisteronk), irrespective of word length;
• Participants’ vocabulary size was correlated with performance on learning the difficult unfamiliar words but not on learning the others;
• Baddeley (2003) suggests this may be due to the nonwords needing both phonological store and the articulatory subvocalisation systems, but words may only need rehearsal but not storage;
• This has the advantage that when a mix of familiar and unfamiliar word-types are encountered the limited capacity phonological store can be largely devoted to learning the unfamiliar words rather than being overwhelmed by the familiar ones.

Gathercole et al. (2001) used recognition to study children who had difficulty articulating words:
• A sequence of words (or nonwords) is presented, then repeated either identically or with two words swapped around, child has to say if the order has changed;
• Words were recalled more successfully than nonwords but recognition performance was similar for both;
• Bilingual participants also showed better recall of words in their primary language over their secondary but no difference in recognition or re-ordering where a sequence of words must be placed in correct order.
• Again, Baddeley suggests this shows that the rehearsal system rather than the phonological store is used for acquiring vocabulary based on existing language.

Baddeley (2003) cites further evidence from study of language deficits that can be explained by the phonological loop model and suggests that it is a good model for explaining aspects of language learning.

Question 49

Modelling the phonological loop

Answer 49

The phonological store + subvocalisation model is simple and has been shown to be useful but has a number of limitations, including:
• it only tries to explain immediate recall;
• it doesn’t explain long-term phonological memory, e.g. how we remember words we’ve learnt;
• it doesn’t explain why people make order errors (e.g. transpose items) when recalling items in a list, or how the order of the items might be encoded.

Baddeley (2000) suggests the aim of a model is to capture what’s known about a topic and to prompt further questions, leading to further elaboration of the model or the creation of new ones.

Computer models have been built to study these and other limitations rather than over-complicating the simple two-component model by adding to it;

The quality of a model can be measured by how closely it predicts the outcomes of experiments with people;

A better test is whether it predicts the outcome of situations that its creator didn’t anticipate (and hence couldn’t have just programmed it to respond to specific cases);

Computer models may need to make assumptions about how the human mind works that can be difficult to justify due to lack of information - for example a model claiming to demonstrate how executive function works could be criticised because we just we don’t know enough about how executive functions work or what systems in the brain are involved to do more than speculate.

Question 50

Serial recall tests

Answer 50

Serial recall tests require the items in a list to be recalled in order, compared to free recall tests where the order is unimportant.

Different mechanisms have been proposed to explain the phonological loop involvement in immediate recall of serial items:

Question 51

Serial recall tests - The chaining hypothesis (Ebbinghaus)

Answer 51

The chaining hypothesis (recall by following associations from item to the next) (Ebbinghaus, 1885) suggests we form “associations” between consecutive items, and that each item then acts as a cue for its successor, but it has been criticised:
• It doesn’t explain how this works when items are repeated in a list (e.g. 194728370 would need two associations to be formed with “7”):
• Errors should happen on the items following the repeated ones because ambiguity of these multiple associations would affect the cueing mechanism but Jahnke (1969) found they tend to happen on the repeated items, not after them;
• Baddeley (1968) studied recall of lists with alternating phonemically similar and dissimilar items (“BXDJTQV”) and found more errors are made on phonemically similar items but the chaining hypothesis would predict more errors should be made on the dissimilar ones (due to ambiguous cueing as in the repeated items case);
• Phonological similarity also affects recall of lists of phonemically similar items such as “B, D, P” because the similar phonemes effectively create a similar/dissimilar sequence - “B-ee, D-ee, P-ee” or “B,E,D,E,P,E”.

Question 52

Serial recall tests - The positional hypothesis

Answer 52

The positional hypothesis (recall by moving from position to position) also suggests items in lists are linked by “associations”, but from another list that indicates position rather than to other items in the list:
• It doesn’t explain transposition errors where items are recalled in reversed order, e.g. 846573 instead of 845673 but some mathematical models have offered explanations of how these might happen;
• A mathematical model based on positional mechanisms by Burgess and Hitch (1999) showed a phonological similarity effect with the alternating similar/dissimilar lists studied by Baddeley (1968).
• The researchers proposed that recalling involved two steps - working out the position of the next item to be recalled, and then retrieving it. They suggested that phonemic similarity disrupts the retrieval step.

Question 53

Serial recall tests - Non-associative models

Answer 53

Non-associative models assume that no associations are formed:
• In the primacy model proposed by Page and Norris (1998) the order of items is encoded according to the level of activation of items stored in memory, with the level decreasing over the items in the list;
• Recall is a two-step process where the first step is to use this primacy gradient to work out which item is next, and the second step is to retrieve the item phonemes.
• Models like this can explain the effects found in alternating lists.

Question 54

Serial recall tests - comments

Answer 54

Chaining models don’t explain the alternating list effects, positional and non-associative models do;

Positional models can explain the temporal grouping effect, the primacy model doesn’t:
• temporal grouping effect is the fact that sequences are easier to remember if presented in rhythmic groups, e.g. 3-2-5-8-7-6-7-9-3 is harder to recall than 325-876-793;
• The most common error is to transpose items in the same position in different subgroups, e.g. 326-875-793 rather than immediately adjacent items;
• This suggests two levels of positional coding - one for the groups and a lower level for the items within groups - which positional models can accommodate while the primacy model only describes encoding on one level (but could be extended)

Question 55

Case studies - P.V.

Answer 55

Vallar and Baddeley (1984)

Findings - Selective verbal memory deficit (left parietal lesion due to stroke), only had a two-digit auditory span, but fluent speech and normal rate of speech;

Poorer recall with phonemically similar words (suggesting phonological store damage) but no word-length effect (also supports a phonological damage diagnosis as subvocal rehearsal wouldn’t help recall of shorter words over longer ones);

Better memory span for visually presented words than for spoken words, and this showed no phonemic similarity or word length effect, suggesting she may have compensated for phonological look damage by using her visuo-spatial scratchpad more to memorise visual stimuli;

This is an example of single dissociation that suggests that storage and rehearsal are separate.

Question 56

Case studies - K.F. - Baddeley

Answer 56

Baddeley et al. (1988)

Findings - Could not learn Italian/Russian word pairs but could learn Italian/Italian pairs;

As her native language was Italian, this suggests her long-term learning was intact but her short-term memory was not;

This dissociation also supports the theory that STM and LTM are separate, but that they are linked for phonological learning - they could be parts of the same functional or even physical system for example;

However the phonological STM/LTM might be distinct from other STM/LTM pairings such as semantic memory.

Question 57

Case studies - K.F. - Shallice and Warrington

Answer 57

Shallice and Warrington (1970)

Findings - Had only a two-digit auditory span due to brain damage from a road accident, hence impaired STM but tested normally for long-term learning and memory and had no problems understanding normal speech, so spared LTM.

This suggests STM and LTM are distinct and normal STM is not required for LTM to function normally.