Working Memory: The phonological loop

1
Working MemoryThe phonological loop

• COGS 551 Human Memory, Spring term 2007, METU
• Annette Hohenberger

2
The phonological loop in Baddeley's WM model

• Old tripartite model
• While the central executive is a-modal, the two
  slave-systems, the phonological loop and the
  visuo-spatial sketchpad, are modality-specific.

Baddeley (2003830)
3
Evidence for the phonological loop

• 1. The phonological similarity effect
• Errors that subjects made in remembering letter
  sequences tended to be phonologically, not
  graphematically related
• B
• V R
• (Conrad 1964)
• --gt the visually presented verbal material must
  have been phonologically encoded

B interferes with V, not with R, although
B-R are more similar, visually.
4
Phonological similarity effect cont.

• Sequences of phonologically related letters were
  harder to remember than phonologically unrelated
  ones
• B G V T P gt Y H W K R
• (Conrad Hull 1964)
• Serial position is hard to recall for
  phonologically similar stimuli.

5
Phonological similarity effect cont.

• Immediate recall of phonologically similar items
  is poorer than recall of dissimilar items
• mad man cad mat cap 9.6 correct
• pit day cow sup bar 81.2 correct
• (Baddeley 1966a,b)
• By that time Baddeley thought the phonological
  loop was articulatory-based
• Congenitally deaf children also showed the
  phonological similarity effect, esp. those who
  articulated well
• Articulatory suppression leads to memory decline

6
Phonological similarity and dissimilarity

• The standard claim is that similarity is
  detrimental but dissimilarity is beneficial for
  (serial) recall.
• However, similarity in itself is not confusing.
  It can be highly beneficial for memory as well.
• That is the Paradox of similarity and
  difference (Nairne 2005)
• Items from categorized lists are recalled better
• Rhyme helps memory hat, rat, cat, mat,
• How can the processing of similarity and
  dissimilarity both benefit retention?

7
Similarity vs. distinctiveness

• Similarity only interferes with (serial) recall
  if it leads to confusion between the items, as
  in
• mad man cad mat cap
• But this confusion arises via the
  indistinctiveness of each item with respect to
  the other in the overall list.
• Consider the list
• mat, hat, bat, cat, fat
• The items in this list are highly similar too,
  however, they are distinctive enough to ease
  memory.

8
Defining Distinctiveness

• Mnemonic distinctiveness (...) is a property of
  a cue in context. It refers to the ability of a
  cue to access a particular target in a particular
  context. (Nairne 2005, 41)
• A cue is a feature that characterizes an item. If
  this cue has a value that distinguishes it from
  other items, then this cue or feature adds
  towards the distinctiveness of the item.
• (Nairne 1990, Neath 2000)

9
The similarity effect in terms of distinctiveness
(Neath 2000)

• In the feature model of Neath (2000), items in
  lists have features which are set to 1 or -1.
• Each item is represented in Short-term and in
  Long-term memory. The memory traces of the items
  in LTM are intact, however, memory traces in the
  to-be-recalled lists can be corrupted due to
  phonological similarity/indistinctiveness.
• If another item in the list has the same feature
  value, then this value looses its
  distinctiveness, it is set from 1 or -1 to 0,
  that is, it becomes useless for later recall.
• At recall, the item is retrieved by accessing the
  most similar trace in LTM, i.e. the one with the
  highest feature overlap. If, however, most
  features have become useless, correct serial
  recall will be poor.
• (in Hanley and Bakopoulou 2003)

10
2. Articulatory suppression

• If the subject has to produce meaningless speech
  ('the the the') during the input/retention phase
  of a recall task with visually presented
  stimuli,
• 1. memory declined
• 2. the phonological similarity effect vanished
• --gt Articulatory suppression prevents the
  visually presented verbal material from becoming
  encoded phonologically
• (Murray 1968)
• --gt next experiment?

11
3. The word length effect

• Subjects remember lists of short words better
  than lists of long words
• Chad Burma Greece ... ??4.17w
• Czechoslovakia Somaliland Nicaragua...??2.8w

Next experiment?
12
4. Irrelevant speech effect

• Irrelevant speech is speech played to the subject
  in either the presentation or the retention phase
  of a memory experiment. Although the subject is
  told to ignore the irrelevant speech and although
  irrelevant may not be parsable at all
  (meaningless, foreign language), memory
  performance decreases.
• --gt auditorily presented stimuli have automatic
  and mandatory access to a phonological system

Next experiment?
13
Combinations of effects(after Hanley and
Bakopoulou 2003)

• How do phonological similarity and irrelevant
  speech interact?
• In auditory presentation

Additive effect Auditorily presented items will
automatically enter the phonological loop, i.e.,
the similarity effect will operate and
performance suffers. Irrelevant speech will also
have a detrimental effect since it uses the
phonological loop also.
14
Combinations of effects

• How do phonological similarity and irrelevant
  speech interact?
• In visual presentation

Single irrelevant speech effect/phon similarity
effect cancelled Visually presented items are
possibly prevented from entering the phonological
loop, since the loop is busy with irrelevant
speech. Therefore, the similarity effect will be
cancelled. However, irrelevant speech will exert
its usual detrimental effect so this effect will
affect recall negatively.
15
Combinations of effects

• How do phonological similarity and articulatory
  suppression interact?
• In visual presentation
• In auditory presentation?

The phonological similarity effect vanishes under
articulatory suppression since articulatory
suppression prevents the visually presented
verbal material from becoming encoded
phonologically
The phonological similarity effect will not
vanish altogether under articulatory suppression
since it cannot prevent auditory items from
entering the loop. So both effects may be
additive.
16
Limitation on slots or on time?

• Comparing same number of words but with different
  (vowel) length
• 1. bishop wicket mimics ...
• 2. harpoon Friday shampoo...
• --gt better retention of the short words in (1)
• Welsh vs. English numbers
• Digit span for Welsh numbers was consistently
  lower than for English digits
• --gt Welsh numbers had longer articulation time
• If controlled for articulation time or through
  articulatory suppression, the difference vanishes

17
Limitation on slots or on time?

• --gt word duration is the crucial variable, not
  number of slots
• (usually both are highly inter-related)
• Forgetting is the joint function of trace decay
  and rehearsal rate. Rehearsal is successful if
  it is faster than trace decay.

18
Memory span and reading rate

• Word length influences both reading rate and
  memory
• Memory span the longer the words, the fewer can
  be recalled
• Reading rate the longer the words, the lower the
  reading rate (number of words per time interval)
• Relation of memory span and reading rate is
  constant.

2 s R rate 6/23 M6/32 R rate 3/21.5
M3/1.52
Mumps stoat Greece Maine
zinc lip tubercoulosis
hippopotamus aluminium
19
Trace decay

• A simple decay hypothesis can explain the data
• You can remember everything that is in the
  phonological loop, no matter how much, short or
  long. If the sequence is longer, rehearsal cannot
  cover them and forgetting through trace decay
  sets in.

20
The 'articulatory loop'

• Simple 'tape loop' concept everything within the
  last ca. 2 s is recorded and gets refreshed by
  subvocal rehearsal. The rehearsal is based on
  articulatory programs which can
• Be confused by phonological similarity/indistinct
  ive or 0 features
• Be strengthened by clear articulation
• Run empty under articulatory suppression

21
Problems with the articulatory loop concept

• Articulatory suppression erases the phonological
  similarity effect and the word length effect only
  when the stimulus material is presented visually,
  not if presented auditorily.
• If articulatory suppression empties the
  articulatory loop, the phonological similarity
  and the word length effect should be erased no
  matter how the material was presented.

22
Effects of articulatory suppression with long and
short words

• Word length effect without articulatory
  suppression
• Articulatory suppression effect
• Word length effect almost disappears under
  articulatory suppression, but not completely

error
Long words suppression short words
suppression long words without
suppression short words without suppression
Serial position
23
Evidence against a merely articulatory loop

• Word length effect did not completely vanish
  under articulatory suppression
• Phonological similarity had an influence no
  matter if suppression was administered only
  during the input phase or also during the recall
  phase
• In visual presentation, the phonological
  similarity effect always vanishes under
  articulatory suppression
• Under articulatory suppression, subjects can
  still judge rhymes and homophony of words
• --gt there must be some other storage independent
  of articulatory rehearsal

24
A revised phonological loop model

• The word length effect and the phonological
  similarity effect rely on different aspects of
  the phonological loop
• Word length reflects a control process of
  articulatory rehearsal. Long words take longer to
  be rehearsed.
• Phonological similarity relies on a short term
  store that is accessible by auditory stimuli or
  by visual stimuli having been coded
  phonologically. Rehearsal is not yet guaranteed
  in this store. Articulatory suppression weakens
  memory in auditory presentation and prevents
  phonological recoding in visual presentation.

Phonological store
Articulatory loop
25
Two systemsThe 'inner ear' and the 'inner voice'

• Inner ear
• Sets up a phonological representation
• Acoustic image
• Too feeble for memory
• Phonological similarity effect
• Judging rhymes/homophones

• Inner voice
• Articulatory loop system
• More durable store for memory,
• Needs subvocal rehearsal
• Word length effect
• Affected by articulatory suppression

26
2nd fractionation
Visual
Cache
Inner
Scribe
The phonological slave system was
fractionated into a phonological store and Inner
speech
The visio-spatial slave system was
fractionated into a visual cache
(visual) and Inner scribe (spatial)
Inner
Speech
PERCEPTION
Psychology 4 OptionHuman Working Memory
Lecture 3Professor Robert Logie
27
Alternative one system, different input leaves
different traces

• Weak traces (inner ear)
• Through visual input
• Enough to judge rhyme/homophony. A phonological
  representation is set up indirectly through
  visual material converted into phonology through
  LTM

• Strong traces (inner voice)
• Through auditory input
• Fostered through rehearsal

28
Perception and production of speech

• Which relation holds between perception
  (acoustic) and production (articulation) of
  speech?
• --gt phonetic features are articulatory rather
  than auditorily defined (place, manner, voice)
• --gt Lieberman's Motor theory of speech
  perception Hearing in terms of speaking. We
  represent perceived speech in terms of the
  articulatory gestures with which the sounds are
  produced.
• --gt generally common coding approach (W. Prinz)
  perception and production/action share a common
  representational code

29
Irrelevant speech and cognitive performance

• Which effects do meaningful and meaningless
  speech in the background have on the performance
  of cognitive tasks?
• decrease
• Irrel. Speech, meaningless PID BOT... 40
• Irrel. Specch, meaningful BID POT... 40
• White noise 19
• Memory task recall sequence of digits
• Digits 8 5...

30
Irrelevant speech under articulatory suppression

• Hypothesis 1 (phonological loop)
• Language material has mandatory access to a
  phonological processor. Articulatory suppression
  should prevent rehearsal of to-be-remembered
  stimuli so the irrelevant speech effect should
  vanish
• Hypothesis 2 (general attention)
• Both factors, irrelevant speech and articulatory
  suppression decrease general attention. The
  irrelevant speech effect should remain.

31
Unattended speech under articulatory suppression

• Task serial recall under
• 1. quiet no suppression art. suppr effect
• 2. quiet suppression
• 3. irrelevant speech no suppression irrel.
  speech
• 4. irrelevant speech suppression effect
• No additional irrelevant speech effect on top of
  the articulatory suppression effect
• --gt phonological loop hypothesis

32
Combinations of effects

• How do articulatory suppression and irrelevant
  speech interact? Hanley and Bakopoulou (2003)
  predict instead
• In auditory presentation
• In visual presentation

Additive effect. Auditorily presented items will
automatically enter the phonological loop and
there both articulatory suppression and
irrelevant speech will interfere with them
because irrelevant speech also demands access to
the loop and articulatory suppression prevents
the items from being rehearsed.
One effect. Visually presented items will be
prevented from being recoded phonologically and
from entering the loop since the loop is already
busy with articulatory suppression. Irrelevant
speech will also use the loop. Unless subjects
find an alternative strategy (visual/semantic),
performance will be bad.
33
The nature of the phonological store

• Does the phonological store operate on lexical
  representations (words) or phonological
  representations (segments, syllables)?

34
Lexical vs. Phonological distractors

• Phon. Dissim. words jelly tipple... least
  disruptive phon. Similar words tun woo...
• Auditory digits one three ... almost
  identical
• Memory task Recall sequences of digits
• Digits 8 5...
• --gt the phonological store operates at the level
  of individual phonemes/syllables rather than on
  lexical representations

35
Obligatory access

• 2 ways in which auditory verbal material has
  automatic access to language processing
• 1. to the phonological loop --gt irrelevant speech
• 2. to the syntactic parser if meaningful --gt
  Fodor 'Parsing is a reflex' (Merrill Garrett)
• Can subjects be manipulated to consider the same
  stimulus as language or as non-language? Would
  this affect automatic access?

36
Trace decay and suffix effects

• Following presentation of digit sequence,
  subjects were given a short instruction ('left'
  or 'right') indication where on the response
  sheet they should write down their answers. The
  instruction acted like a 'suffix' to the list and
  impaired memory, the more, the longer the suffix
  was.
• --gt due to trace decay or
• --gt due to overall information overload?

37

• Suffix (the instruction) was given before the
  last item of the list so that recall could begin
  immediately
• Predictions by decay hypothesis Suffix-effect
  should vanish (no delay of recall)
• Prediction by overload hypothesis Suffix effect
  shoud remain (same information load)
• Result Suffix effect vanished --gt decay

38
Neuropsychological evidence for the phonological
loop

• STM patients are bad on STM tasks but have
  unimpaired LTM. The WM model explained this in
  terms of an impaired phonological loop but intact
  central executive.

39

• However, 3 critical observations for the old
  model ('articulatory loop')
• 1. STM patients have a better memory of visual
  material than for auditory. Some visual retention
  of position must have occurred. However, this was
  not the task of Baddeley Hitch's viso-spatial
  sketchpad
• 2. STM patients have a reduced recency effect.
  However, Baddeley Hitch had shown independence
  of memory span an recency
• 3. STM patients may have STM problems despite
  good articuation.

40

• Patients and unimpaired control subjects had to
• recall words vs. Environmental sounds under
  articulatory suppression
• Patients did worse on words than on sounds
• Unimpaired controls did worse on sounds than on
  words.
• Articulatory suppression should have equated both
  groups (unimpaired controls becoming impaired
  like the patients). Since the patients were still
  doing worse, the problem could not have been
  articulatory but must have been acoustic
• --gt there is an additional input-related acoustic
  store that is damaged in STM patients ('inner
  ear')

41
Level of processing in the phonological loop

• Does the phonological loop depend on peripheral
  processing (neurophysiological gestural code) or
  on internal processing (phonological code)?
• Language vs. speech?

42
Case study Anarthria/Dysarthria

• Anarthric patients such as G.B. suffer from
  subcortical motor aphasia. They are not
  linguistically impaired but peripherally they
  cannot speak anymore.
• If the phonological loop is language-based, G.B.
  Should perform phonological loop tasks normally
• If the phonological loop is speech-based, G.B.
  Should show deficits

43
Result of Patient GB

• Normal digit span 6
• Phonological similarity effect for auditory and
  visual presentation --gt phonological store and
  rehearsal
• word length effect --gt subvocal rehearsal
• Judging of phon. Similarity of homophones and
  nonwords (only slower) --gt phonological store
• --gt phonological encoding (in the adult) does NOT
  depend on the ability to pronounce words

44

• The phonological loop operates on a 'deep' level
  of a central speech control code, i.e., language,
  not on a 'peripheral' level of pronunciation.
• Does this control of inner speech develop in
  childhood through highly overlearned speaking
  routines or is it automatically set up?

45
Functions of the Phonological Loop(Baddeley
2003a)

• Aid language acquisition, vocabulary learning
• Control of behavior switching from one cognitive
  OP to another

46
The phonological loop and language modality

• Evidence from Sign language

47
References

• Alan Baddeley (1986) Working Memory. Oxford OUP
• Baddeley, Alan D. (2003a) Working memory and
  language an overview. Journal of Communication
  Disorders, 36, 189-208.
• Baddeley, Alan D. (2003b) Looking back and
  looking forward. Nature Reviews Neuroscience, 4,
  829-839.
• Haneley, J. Richard and Bakopoulou, Eirini
  (2003) Irrelevant speech, articulatory
  suppression, and phonological similarity a test
  of the phonological loop model and the feature
  model. Psychonomic Bulletin Review, 10 (2),
  435-444.
• Nairne, J.S. (2000) A feature model of immediate
  memory. Memory and Cognition 18, 251-269.
• Nairne, J. S. (2005). Modeling distinctiveness
  Implications for general memory theory. In R. R.
  Hunt J. Worthen (Eds.), Distinctiveness and
  memory. New York Oxford University Press, 27-46.