Cognitive approach

• The "Reading the Mind in the Eyes" Test Revised Version: A Study with Normal Adults, and Adults with Asperger Syndrome or High-functioning Autism
• Theory of mind


  The ability to attribute mental states to oneself or another person
• Mentalising


  Another term for theory of mind
• Mind reading


  Another term for theory of mind
• Social intelligence


  Another term for theory of mind
• Empathy


  Overlaps with theory of mind
• Task analysis of the Eyes Test

  1. Subject needs mental state lexicon and know semantics of terms

  2. Map terms to fragments of facial expressions of mental states around the eyes

  3. Match eyes in each picture to examples in memory to judge mental state
• The Eyes Test only involves the first stage of attribution of theory of mind: attribution of the relevant mental state
• The original Eyes Test showed adult males in the general population scored a mean of 18.8 out of 25, while women scored slightly higher at 21.8 out of 25
• Adults with high functioning autism (HFA) or Asperger Syndrome (AS) performed significantly worse than sex-matched normal controls or adults with Tourette's syndrome, scoring on average 16.3 out of 25
• Problems with the original version of the Eyes Test

  • Narrow range of scores above chance (only 9 points)
  • Parents of children with AS scored similarly to those with HFA/AS, unable to distinguish 'broader phenotype'
  • Potential ceiling effects with normal performance close to test ceiling
  • Included basic and complex mental states, some items too easy
  • Some items could be solved by checking gaze direction alone
  • More female than male faces, potential bias
  • Target word and foil were always semantic opposites, too easy
• Modifications made in the revised Eyes Test

  1. Increased number of items from 25 to 36

  2. Increased number of response options per item from 2 to 4

  3. Focused only on complex mental states, excluding basic emotions

  4. Excluded items where gaze direction could provide the answer

  5. Balanced number of male and female faces

  6. Avoided using semantic opposites for word choices
• The original version of the test had more female faces than male faces, and it was unclear if this may have biased the test in some way
• In the revised version of the test, the number of male and female faces was carefully controlled to be equal
• In the original version of the test, the target word and its foil were always semantic opposites, making the test too easy
• In the revised version of the test, the foil words have the same emotional valence as the target word, making the test more challenging
• The revised version of the test includes a glossary of all the mental state terms, which subjects were encouraged to consult if they were unsure of a word
• The study reports data from the revised version of the test, with several additional aims
• Aims of the study

  • To test a group of adults with AS or HFA on the revised version of the test
  • To test if in a sample of normal adults, an inverse correlation would be found between performance on the Eyes Test (Revised) and the Autism Spectrum Quotient (AQ)
  • To test if the sex difference (female superiority) found on the first version of the test replicated
• Subjects


  • Group 1: Adults with AS or HFA (N=15, all male)
  • Group 2: Normal adults (N=122)
  • Group 3: Normal adult students (N=103, 53 male, 50 female)
  • Group 4: Randomly selected individuals in the general population (N=14), IQ matched with Group 1
• Procedure


  1. Subjects in all four groups were tested on the revised adult Eyes Test

  2. Subjects in the AS/HFA group were also asked to judge the gender of each person in each photo, as a control task

  3. Subjects in Groups 1, 3, and 4 completed the AQ
• The data from Groups 2 and 3 did not differ, so the results were combined, creating a sample of N=225
• New criteria were applied to the combined data: at least 50% of subjects had to select the target word and no more than 25% could select any one of the foils
• Items 1, 2, 12 and 40 failed to meet these criteria and were dropped, leaving 36 items for subsequent analyses
• no within-group differences in Group 3 (students) according to subject studied, F(1, 99)139, p 24
• On the AQ, as expected, Group 1 scored significantly higher than Groups 3 and 4: one-way ANOVA of group, F(2, 103)234, p 00001; Scheffé's tests indicated Group 1 scored significantly higher at the .05 level than Groups 3 and 4, for which there was no difference
• The predicted sex difference on the AQ (males scoring higher than females) in Group 3 was also found (t197, p 03 for one-tailed significance)
• The modifications from the original version led to normal performance being significantly below ceiling
• This study replicated the earlier finding that adults with AS or HFA are significantly impaired on such tests, whereas they are not impaired on the gender recognition control test
• Among the general population controls and student group, there was a trend towards a sex difference (female superiority) (p 07)
• There was no significant correlation between IQ and the Eyes Test, suggesting this is independent of general (nonsocial) intelligence
• Performance on the Revised Eyes Test was inversely correlated with performance on the Autism Spectrum Quotient (AQ), suggesting that both measure degrees of autistic traits across the notional spectrum
• The AQ is not diagnostic but may serve as a useful instrument for quantifying the extent of an individual's "caseness" in terms of ASHFA, measuring personality traits
• The present results confirm our earlier finding that adults with HFA or AS score significantly higher on the AQ than do general population controls
• Even with the new modifications, the stimuli are static, whereas the real world never is
• Future studies might usefully employ dynamic stimuli of eye expressions

  
• Static stimuli, however, make the test quick and easy to use, since it can be administered as a "pencil and paper" test
• In the laboratory, they are also experimenting with computer presentation of the Eyes stimuli so as to record response time in subjects' judgements of the most appropriate mental state term to match each picture
• Such speed of processing approaches may be a fruitful way to explore individual differences on this task
• However, it is clear that even a nonautomated format is sufficient to reveal group differences