Dr' Simon Davies

1
Attention
Dr. Simon Davies
Read Ch. 2 of Brasiby Gellatly
2
Lecture structure

• Brief overview of attention characteristics.
• Auditory attention and key questions
  (experimental).
• Visual attention and key questions
  (experimental).
• Attention dysfunction (neurological).
• Objects and attention.

3
Definitions

• James (1890) Everyone knows what attention is.
  It is the taking possession by the mind, in clear
  and vivid form, of one out of what seem several
  simultaneously possible objects or trains of
  thought. (p. 3 Cited in Scholl, 2001).
• Naish (2005) Attention is the process which
  gives rise to conscious awareness. (p. 67).

4
What defines attention?

• Selectivity We can process some stimuli more
  than others.
• Limited capacity We can only process a small
  amount of information at any one time.
• Effort Processing information that we are
  attending to requires us to make an effort.
• Attention can be directed (i.e.
  selective/top-down) or captured automatically
  (i.e. bottom-up).
• Similarly, attention can be focused or divided.

5
Auditory attention

• Ears cannot be directed away from competing
  stimuli, as eyes can.
• Spotlight of attention demonstrates that we can
  direct our attention to encompass only a small
  amount of information.
• Dichotic listening tasks developed to explore how
  effective this is and exactly what we screen out.

6
Read only bold words

• Somewhere Among hidden the in most the
  spectacular Rocky Mountains cognitive near
  abilities Central City is Colorado the an ability
  old to miner select hid one a message box from of
  another. gold. We Although do several this
  hundred by people focusing have our looked
  attention for on it, certain they cues have such
  not as found type it style.

7
Dichotic listening methods

• Different sounds or messages broadcast to each
  ear (headphones).
• Participant asked to monitor one ear.
• For short messages (approx. three words)
  participant can recall both ears (using echoic
  memory).
• Longer messages mean unattended material
  effectively lost to conscious awareness (semantic
  content).
• However, the quality (e.g. male, angry) of the
  unattended message could be recalled.

8
Early vs. Late selection key concepts

• Early selection information selected prior to
  semantics have been resolved.
• Late selection information selected after
  semantics have been resolved.
• Thus, we have the paradox of intelligent
  selection!

9
(No Transcript)
10
Broadbents (1958) filter theory Early
selection

• Information is processed at a sensory level in
  parallel.
• Some input is prioritised. A filter restricts
  input to selected information.
• Short-term memory has a limited capacity, and so
  the filter prevents STM from overloading.
• Broadbents theory however fails to account for
  findings that suggest that some non-attended
  information is processed below awareness.

11
Cocktail Party Effect study

• Cherry (1953)- showed that information other than
  that attended can grab our attention.
• In public situations we can become aware of
  information outside our spotlight of attention if
  we are sensitised to it (e.g. our name). We also
  code for characteristics of unattended material.

12
Serial vs. parallel processing key concepts

• Serial processing as with computers, some
  theorists thought the brain could only process
  one piece of information at a time. Some
  processes are serial.
• Parallel processing as with connectionist
  models, some theorists think that we can process
  all the information received at the same time
  (i.e. parallel).
• Psychologists now agree that perception is a two
  stage process (parallel then serial) but disagree
  on the exact details (e.g. how information is
  selected, what happen to unselected information,
  etc.).

13
Treismans Attenuator Theory early late
selection

• The Cocktail Party Effect suggests that
  unattended info isnt all lost.
• Some seems to be processed below awareness.
• Treisman (1960) argued against Broadbents gate
  theory, and instead saw input being attenuated
  through a filter.
• Input is still selected early in processing
  through gross features, but unattended material
  is processed at a reduced level.

14
Shadowing methods

• Shadowing requires the participant to repeat out
  loud one of two concurrent messages presented to
  different ears.
• When the messages have finished participants are
  unaware of the unattended message.
• Participants would, however, follow the meaning
  of the story even if it switched to the
  unattended ears if the story was familiar.
• Treisman suggests that when we expect something
  we become sensitised to it, thus the current
  information primes us to similar material.

15
Shadowing - example

• Left ear
• Little Red Riding Hood finally reached the
  cottage, the wicked wolf was in beds one was
  large, one medium and one small.
• Right ear
• When she had finished the porridge, Goldilocks
  went upstairs and found three bed, dressed in
  the grandmothers clothes.
• indicates point at which message switches ears.

16
Other evidence of sensitisation (priming)

• Lexical decision task methods- when asked to
  make a decision if a series of letters is a word,
  we make quicker responses if the preceding word
  is semantically related.
• Corteen Wood (1972) used electric shocks to
  associate some categories with unpleasant
  sensation. When these words were then played to
  the unattended ear, there was still a GSR,
  showing that the unattended material had been
  processed.

17
Deutsch Deutsch (1963, 1967)Late selection

• All input is fully analysed in parallel.
• Decisions about priority are taken later, at
  which stage capacity limitations come into
  operation.
• Treisman and Geffen (1967) and Treisman and Riley
  (1967) demonstrated that late selection theories
  of attention must be wrong concurrent tasks
  with presentation were not superior for the
  non-attended targets.

18
Late vs. Early resolved?

• Lavie (1995)argues that both early and late
  selection can occur, but depend on perceptual
  load.
• High load early selection due to cognitive
  demands.
• Low Load late selection due to additional
  capacity from lower cognitive demand.

19
Visual Attention

• Unlike auditory attention we can move our eyes to
  look at specific objects, or to screen other
  aspects of the environment.
• Similarly, vision doesnt contain mixed signals
  to the extent that hearing does.
• Similar questions abound in the study of visual
  attention.

20
Visual questions

• Does visual processing operate in serial or
  parallel?
• How long does it take to process a visual image?
• How much of what we see can be processed and in
  what way?
• Is selection of visual material early or late?

21
Serial vs. parallel selection

• Sperling (1960) examined this question by asked
  participants how much of a 3 x 4 grid of briefly
  presented letters they remembered.
• Whole report Typically only three or four
  letters reported.
• Partial report The whole display is available
  for report if a single line is cued after
  presentation.

22
Iconic vs. vSTM

• Sperling showed that we have a brief icon of the
  whole display (lt 200 ms). This iconic memory is
  parallel and unlimited.
• A second stage then occurs which is serial in
  nature and which does not fade a quickly, but can
  only hold four items. This stage is called vSTM.
• Sperling E-Prime script here.

23
Backward masking methods

• Masking an item means presenting another item
  soon after its offset.
• Masking can prevent conscious awareness of the
  first item, by
• Integration the two items become integrated so
  that the first cannot be recognised (short SOAs).
• Interruption the process of perceptual
  resolution in interrupted (longer SOAs).
• Coltheart (1980) argues that attention is
  necessary to fuse episodic and semantic features
  of an object.

24
Rapid Serial Visual Presentation methods

• Participant sees a stream of letters at fixation
  in rapid succession.
• Task to identify a target letter (T1) in new
  colour, as well as say whether an X was present
  (T2).
• If T2 is shortly after T1 then there is no memory
  of T2.
• If T2 is distant from T1 then there will be a
  memory of T2.
• Shapiro E-Prime expt here

25
Other RSVPs and the Attentional Blink

• Vogel et al. (1998)used words rather than
  letters. Showed conceptual link can allow memory
  of T2 (the reverse of masking).
• Raymond et al. (1992)- RSVP effect disappeared if
  letter after T1 removed.
• Attention blink it appears that it takes time
  to resolve T1, and in that time all new items are
  not processed.

26
Negative Priming Methods

• Tipper Cranston (1985) showed that non-attended
  material is processed below awareness.
• Participants attend to one of two overlapping
  shapes, the unattended shape then becomes the
  attended shape in the next trial.
• Participants are slower to respond when the new
  attended shape was previously attended.
• Tipper Cranston E-Prime VIAT expts here

27
Reflection

• It thus appears that attention is needed to give
  us a conscious perception.
• Attention fuses episodic and semantic
  information.
• Unbound semantic information can be revealed by
  using priming studies thus we do have a memory
  trace of which we are unaware.

28
How does selection operate?

• Selection seems to operate over two stages, one
  parallel, and one serial.
• Automatic (data-driven, bottom-up, pre-attentive,
  parallel) behaviour occurring without
  intention, involuntary, without conscious
  awareness, no interference with other cognitive
  processes. (Posner Snyder, 1980)
• Attentive (concept-driven, top-down, serial)
  requires conscious effort, interferes with other
  cognitive processes by taking up limited
  resources.

29
Feature Integration Theory

• Treisman Gelade (1980) showed that attention is
  required to integrate visual information.
• Asked participant to identify a target letter
  (e.g. X amidst field of O distractors).
• Varied set size and whether target could be
  detected by feature or conjunction of features.
• Treisman E-Prime and VIAT expts here

30
FIT results
31
(No Transcript)
32
Illusory Conjunctions

• Treisman Schmidt (1982) compared focused vs.
  divided attention.
• Task to recall coloured letters.
• When attention is divided participants
  misremember e.g. a red C might be remembered as
  a blue C.
• Thus, attention is needed to bind features
  together.

33
FIT criticisms

• The Flanker Effect (Shaffer LaBerge, 1979).
  Target word category task words flanked by
  non-task words.
• Treisman argued that serial focused attention is
  needed to join features together.
• The Flanker Effect shows that parallel processing
  influences what we are focused on.

34
Other attention paradigms

• Change blindness failure to notice changes in a
  scene.
• Inattentional blindness blindness to visible
  objects when attention is focused elsewhere.
• Stroop effects demonstrated features of an
  attended object are coded together and can
  interfere with response selection.
• Stroop VIAT expts here CB demos

35
Neurology of attention

• The most common neurological condition affecting
  attention is unilateral neglect.
• This usually occurs due to a damage of the right
  parietal lobe.
• It results in neglect to the contralateral visual
  field, resulting in blindness for objects or
  object features in that area.
• A common task to detect such a condition is to
  get patients to draw an object or to cancel out
  features of a pictre.

36
Visual neglect
37
Unilateral neglect

• Evidence that this is an attention disorder
  rather than sensory comes from several sources
• Driver Halligan (1991) patients cannot detect
  features on the left side of an object despite
  these being oriented to the right visual field.
• Bisiach Luzzati (1978) patients can imagine
  remembered scenes, but their recall is affected
  by their imagined orientation in space.

38
Balints syndrome

• Failure to see anything in the visual field
  except a single object (Balint, 1909). Balints
  sufferers suffer four main symptoms
• Ocular apraxia inability to change fixation.
• Simultagnosia inability to perceive more than
  one object.
• Spatial disorientation inability to orient and
  localise objects.
• Optic ataxia inability to reach out and touch
  an object in space.

39
Additional sources

• www.viperlib.com visual perception website.
• Palmer, S. (1999). Vision Science. Cambridge, MA
  MIT Press.
• Pashler, H. E. (1998). The psychology of
  attention. Cambridge, MA MIT Press.
• The change detection database via Google. Check
  out the demos.

40
Object-based or space-based?

• An initial distinction can be between early and
  late selection.
• Early space-based, thus locations are selected
  from a low-level sketch.
• Late object-based, thus location-independent.
• Look at material on Marrs work from two weeks
  ago.

41
Neurology again

• Another distinction between objects and locations
  can be raised
• Occipital (V1) areas use a retinotopic code
  (locations).
• Inferior Temporal (IT) areas represent whole
  objects, independent of their location.
• Where does selection occur?

42
Location-Based 1

• Evidence comes from spatial cueing tasks
• Eriksen Eriksen (1974) showed that responding
  to a target letter is affected by
• Whether that letter is in the same letter set.
• Distance from the target letter.
• See also Posner Cohen (1984)for an example of
  spatial cueing.

43
Location-Based 2

• We have already seen Treismans FIT.
• Here, for features to be bound together, the
  location in different feature maps needs to be
  attended.
• In this sense, perception is very much
  feature-based.

44
Object-Based 1

• Duncan (1984) used overlapping objects to show
  that two features of a single object (same-object
  advantage) could be judged more quickly than two
  features of two objects (see next slide).
• Thus, selection would appear to be object-based.

45
(No Transcript)
46
Resolution

• Duncan (1984), Vecera Farah (1994) and others,
  have shown that selection can be both object- or
  location-based.
• If we ignore the issue of a single bottleneck in
  terms of selection, we can see that attentional
  selection might be influenced by a number of
  factors one of which is the level of processing
  at which selection is achieved.