Isolating the impact of visual perception on dyslexics

1
Isolating the impact of visual perception on
dyslexics reading ability

• Mark M. Shovman Merav Ahissar

Vision Research 46 (2006) 3514-3525
2
Dyslexia

• Consensus neurological disorder with a genetic
  origin
• Diverse associated cognitive deficits
• Co-morbidity with other learning disabilities
• Several parsimonious theories of single cause
• Most broadly accepted cognitive theory
• core deficit at level of phonological
  representations
• impact on decoding written script relatively
  understood
• Impact of potentially impaired visual abilities
• open and debated question
• In general, process of reading taxing for visual
  system
• demands fine spatial discrimination and rapid
  processing (Vidyasagar, 2004)

3
Visual Deficits

• Magnocellular hypothesis low-level visual
  deficit
• Focus of recent studies
• dynamics of spatial visual attention (Geiger et
  al., 1994)
• deficit associated with higher levels of
  processing along dorsal stream (Vidyasagar
  Pammer, 1999)
• Attentional shifts sluggish (Hari Renvall,
  2001)
• resemble minor case of neglect
• Visual deficits only when comparing between
• spatial or temporal aspects of serially presented
  stimuli (Ben-Yehudah Ahissar, 2004)

4
Relevance for Dyslexics

• Implications of visual deficits on reading
  abilities scarcely addressed
• Designed experiments to assess adequacy of
  visual routines that play important role in
  single word reading
• Stimuli similar to single words in graphical
  characteristics
• no other aspects of natural reading (e.g.
  phonological, morphological)
• If dyslexics reading-related visual routines
  mildly impaired
• weakness revealed when relevant visual
  requirements increase and visual routines
  challenged

5
Visual Manipulations

• Reduced letter size
• affects spatial frequencies used to identify
  symbols (Majaj et al., 2002)
• impaired performance of dyslexic children with
  unstable binocular control (Cornelissen et al.,
  1991)
• Crowding (added distracting letters - flankers)
• crowding effects near fixation for dyslexic
  children (Atkinson, 1991)
• but may be confounded by verbal memory processes
• Visual noise
• examined contrast sensitivity for letter
  identification
• letters presented on uniform gray background or
  embedded in white noise
• Linear Additive Model
• discuss later

6
Methods

• SET-UP
• Participants seated 1 metre from screen
• Response time neither limited nor measured
• STIMULI
• Symbol set subset of Georgian alphabet
• letter-like symbols graphically similar to know
  alphabets
• not resemble English or Hebrew scripts (no
  phonological interference)
• but similar graphical complexity
• very similar to pseudowords (3-4 letters) in
  visual aspects and lack of semantic content

7
Trial Sequence
A Fixation bar suggested by Dr R Shillcock in
private communication B Single-symbol stimulus
on uniform gray background (47.4 cd/m²) C
Letter-triplet in noise (0-94.4 cd/m², 2' x 2'
square grain) D Masking screen (random scatter
of symbols presented for 500 ms) E Response
options
8
Test Conditions

• Eight conditions, comprising up to 100 trials each

Set Stimulus Size Background SOA Luminance
1 Single symbol 0.5? Uniform gray Adaptive 53.3 cd/m²
2 Single symbol 1? Uniform gray Adaptive 53.3 cd/m²
3 Single symbol 1? Uniform gray 200 ms Minimal (adaptive)
4 Single symbol 1? White noise 200 ms Minimal (adaptive)
5 Triplet 0.5? Uniform gray Adaptive 53.3 cd/m²
6 Triplet 1? Uniform gray Adaptive 53.3 cd/m²
7 Triplet 1? Uniform gray 200 ms Minimal (adaptive)
8 Triplet 1? White noise 200 ms Minimal (adaptive)

• Thresholds assessed with 2-up, 1-down staircase
  method
• SOA adaptive steps 30 ms decreasing to 10 ms
  after 5 reversals
• Contrast 2.4 (no-noise) / 3.7 cd/m² decreasing
  to 0.7 / 1.7 cd/m² after 5 reversals, and to 0.2
  / 0.4 cd/m² after additional 4 reversals
• Step size increase possible after 3 changes in
  consistent direction

9
Participants

• Mainly students (aged 21 27)
• 20 dyslexics (14 female, 6 male)
• self-selecting exclusion criteria
• well below average cognitive within average
  pseudoword reading scores
• some participants from previous studies
• 20 controls (13 female, 7 male)
• excluded two for well above average cognitive
  scores, one for slow task completion and one
  because he was male!
• Reading and Cognitive Tests included
• Hebrew pseudoword and paragraph reading
• Rapid Automatic Naming of digits
• Questionnaire on history of learning disabilities
• WAIS-III Block Design and Similarities
• Ravens Standard Progressive Matrices
• Digit Span (forward and backward)

10
Results Assessments

• Cognitive abilities similar, dyslexics digit
  span poorer
• Dyslexics reading related measures significantly
  lower

• Both groups para reading speed correlated with
  RAN-D
• Controls only pseudoword reading ---------
  ------------
• dyslexics performance limited by e.g.
  phonological processing

11
Results Symbol Identification

• Second symbol most accurately identified
• Dyslexics less accurate for fifth (not reported)

12
Results Contrast Duration

• Performance of two groups similar in all
  conditions

13
Results Second-order Effects

• Similar performance for SOA and contrast
  thresholds

1. Large effect of white noise on contrast
   thresholds
2. Intermediate effects of size and flankers on SOA
3. Negligible effects of flankers on contrast
   threshold

14
Results Visual Measures

• Dyslexics
• no performance difference for two visual
  subgroups
• visual measures and reading-related scores not
  significantly correlated

• Controls
• correlation between Contrast threshold and
  Pseudoword reading speed
• dyslexics scores not normally distributed

• Most visual thresholds highly correlated
  (particularly controls)
• suggests common hidden factor - overall grapheme
  processing
• primary factor almost-equally-weighted average
  of all thresholds z-scores

15
Discussion

• Manipulations reduced performance similarly both
  groups
• substantial differences in reading abilities
• lack of differentiation by relatively broad
  battery or visual ability tests
• suggests dissociation reading difficulties and
  visual skills
• Tasks tap common visual mechanisms
• presumably related to grapheme identification
• Pseudoword reading speed contrast threshold
• controls correlation implies visual measures tap
  reading related visual abilities
• no dyslexic correlation suggests visual abilities
  not limit reading ability
• Alternative explanation experimental population
• dyslexics tested not have substantial visual
  deficits
• university students
• specific reading difficulties
• above average general cognitive abilities

16
Discussion (contd)

• Complaints of visual discomfort
• greater in dyslexics, but not related to visual
  task performance
• interpreted as visual stress being consequence
  rather than cause of reading difficulties
• reading puts heavier load on visual attention for
  dyslexics
• Conclude
• visual problems may be prevalent
• could be used as markers for reading deficits
• probably not relevant for any amelioration
  program
• do not seem to pose any functional bottleneck
• Final message from Abstract
• difficulties with single word reading not visual
  processing deficit

17
Criticisms

• Task supposedly similar to single word reading,
  but
• stimuli not letters
• matching task (not letter identification)
• viewing distance not realistic (1 metre)
• Crowding condition not really crowded
• Contrast sensitivity controversial measure
• previous research shows not reliably different
  between groups
• cited articles (Dosher Lu, 2005 Gold et al,
  1999) relate to perceptual learning in
  non-clinical participants
• No explanation/predictions for SOA condition
• Experimental tasks not timed
• Tasks not sensitive enough to tap dyslexics
  visual deficits

18
Linear Additive Model

• Linear Additive Model irrelevant
• mentioned in Introduction, not referred to in
  Discussion, not implemented
• Model specified by Pelli (1999) comprises two
  factors
• Efficiency Rates the computation underlying
  our perceptual decisions on the absolute
  performance scale defined by the ideal observer
• Equivalent Noise Specifies how much noise the
  observers visual system adds to the
  display

19
Shovman Ahissars LAM

• Interpretation
• simple assumption only one source of inner noise
  (additive to signal)
• observers discriminatory ability (D) and
  equivalent inner noise (Neq) define overall
  efficiency of visual system for this type of
  letter identification
• so measured contrast thresholds for
  identification at two different levels of noise
  (one with, one without)
• Lack of implementation
• confusion re nature of noise white noise, SOA,
  contrast, crowding, size
• white noise, crowding and size not additive
• no measure of Equivalent Internal Noise
• baseline performance of control group (not
  ideal observer)