Cognitive psychology: Case study 3: Baron-Cohen's eyes test

Cognitive psychology

Case study 3: Baron-Cohen's eyes test

Author: Baron-Cohen et al. (2001)

Key term: Eyes test

Background/context: One symptom of autism is mind blindness. (the ability to know what someone is feeling). Baron-Cohen et al. devised a theory of mind test in 1975 for autistic adults called the 'eyes test', but there were many methodological problems with this test so a new version was revised in 2001.

Problems with original test

• Only a choice of two words for each picture, meaning the answer could be a 50/50 guess. Four words were added to the revised test. (Prevent ceiling effect)
• Revised test had 40 (reduced to 36) pictures instead of 25. (Prevent ceiling effect)
• Too many easy pictures as well as pictures that could be guessed by looking at the gaze direction, which did not assess the mental state. (Prevent ceiling effect)
• More female faces, Equal numbers of male and female faces were put in the revised test.
• In the original test the words were semantic opposites (eg. happy and sad). Words changed so they were not semantic in revised test. (Prevent ceiling effect).
• The words used were not known by some participants so in the new test a glossary of terms were given to participants.

Ceiling effect: Too many scores at the top end of the mark range (too many people getting good scores).

Hypothesis:

To see if revised version works on autistic adults

Predictions:

1. People with AS/HFA (aspergers/high functioning autism) will score lower on the eyes test than other groups
2. People with AS/HFA will score higher on the AQ test than other groups.
3. 'Normal' females will score higher than males on the eyes test.
4. 'Normal' males will score higher than females on the AQ test.
5. Scores on the eyes and AQ test will be inversely correlated.

Predictions 3 & 4 are because Baron-Cohen predicts that a surplus of testosterone causes autism, so males, who generally have more testosterone than women, should have more autistic tendencies.

Method: Natural experiment with questionnaires

Questionnaire design

Eyes test: 

36 black and white photos of different male and female eye regions taken from a magazine. Each photo has four words to describe the mental state of the person in the photo. Participants need to choose the word they think describes the person's mental state from the photo of their eyes. After completely all 36, the correct answers are rounded up to give an overall score.

The words for each set of eyes were chosen by two authors and judged by a four male four female team. At least 5 judges had to agree the word was correct for the mental state. Four photos did not achieve this consistency so they were dropped, leaving 36 photos.

AQ test:

5o statements of which the participants get to choose "slight agree, definitely agree, slightly disagree, definitely disagree". There was actually no difference between 'Slightly and definitely',  one point is given to either agree or disagree. Total score out of 50. This test is reliable and valid, and both the eyes and AQ test are psychometric tests. 

Groups

1. 15 adult males with AS/HFA. Mean IQ of 115
2. 122 normal adults (control group)
3. Normal adult students (cambridge undergrads) 103 (53 male 3 female). Higher than average IQ.
4. 14 IQ matched controls with group 1.

Apparatus

Eyes test, AQ test and quiet room in Cambridge.

Procedure

• All groups given eye test
• Groups 1, 3 and 4 given AQ test

Data- Quantitative

Results

1. Mean eyes test score was lowest for group 1 (AS/HFA) at 21.9.
2. Sex difference was examined in group 1 and 2 but the difference was not statistically significant enough for the AQ and eyes test.
3. Group 1 scored significantly higher on the AQ test (mean 34.4) than groups 3 and 4 (mean  18.3 and 18.9).
4. Significant inverse correlation of -0.53.

Conclusion

• This test is an improvement of the original as the old weaknesses were not observed.
• All initial aims met and predictions confirmed.
• Validates eyes test as useful for identifying impairments related AS/HFA.