April 24, 2011

Why Do Autistics Score Poorly On The Eyes Test?

"I think she wants me."  Wrong.

Asking what's lacking in an autistic kid is like asking what's lacking in a car.  Fuel efficiency?  Horns?  A duck?

There's no consistency in diagnosis, even though the diagnosis is immensely reliable.  That means that ten doctors will all agree a person has "ASD," but that person may look nothing like the other people all who have reliably been diagnosed.  This makes offering them treatment even more difficult.

So we have choices: try to refine the diagnostic criteria, or create separable categories, or dig backwards in time to find the "neurodevelopmental deficits" that existed in common. 

Trouble is, even an identical, genetically determined, structural pathology-- e.g. "larger cerebellum"-- may result in different phenotypes as each kid will learn different strategies to cope.  How, without the eye of God,  am I supposed to tell if someone has it?


"Reading the Mind in the Eyes" Test was devised by Simon Baron-Cohen, and revised in 2001 to improve sensitivity.  The test is widely employed and widely criticized, but it's useful to understand his logic first.

The test is 36 pictures of eyes like the one above.  The woman's eyes, above, have choices:

a) decisive
b) amused
c) aghast
d) bored 

The test is here-- but DON'T TAKE IT until you read this whole post first.  A similar version for children is here, using the same eyes but different words:

a) made up her mind
b) joking
c) surprised
d) bored

Here is a distribution of scores generated with normals:

eyes test distribution.JPGIn this particular study, the "high functioning autism" group scored 21.9 +/- 6.6.

What does Baron-Cohen think this test tests?  One starts from the assumption that ASD is a neurodevelopmental disorder-- i.e. a "static encephalopathy."  Something happened early, and though it may not be progressing, it is a physical limitation.

The point of this test is to be able to detect, even in a person who appears normal, whether they have this "static encephalopathy."   Do they have true deficits in social cognition that occurred early but have been masked by learning?

Hence the Eyes test.  The test is NOT testing the ability of the person to read faces or interpret their emotions, that's exactly the opposite of the point of it. People can have learned adaptive strategies to get at the right answer. This test is supposed to be immune to those tricks. 

This test was conceived of as a test of how well the participant can put themselves into the mind of the other person, and "tune in" to their mental state.

Importantly, this requires that the person understand their own emotions and have language to articulate it, which is what Baron-Cohen and others believe is the core deficiency of ASD: the "absent self."  So to do well on the Eyes test,  they must not be alexithymic.  You can't interpret the eyes as "judgmental" (lacking the "hints" that come from the mouth, forehead, context, etc) unless you understand that emotion in yourself.  The relationship between alexithymia and low Eyes Test scores has been directly measured.

For example: do psychopaths have difficulty "tuning into" the mental state of others, or can they do it just fine but don't care?  I've always felt the latter, and so I'd predict psychopaths do fine on the Eyes Test.  They do.  (Which, BTW, speaks to the legal question of sanity.)


Is the test flawed, i.e., does it really detect these pre-learning deficits?  Lots of ASD people do well on it, especially women, so the test may not be very sensitive after all.  Can we at least say that those who score poorly do have the deficits in social cognition?  That the test isn't particularly sensitive, but it is specific?

I have very little experience with ASD patients, but I had occasion over this holiday break to cover an ASD unit.  An occupational therapist was explaining this test to me, and we showed some of the eyes to a 6 year old boy with ASD who was hospitalized (his second) for behavioral dyscontrol.  The kid got several wrong, for example this one:

eyes test friendly.jpg
Baron-Cohen's theory of autism is predicated on a reduced ability to imagine the thoughts and feelings of the other person substantially because they cannot recognize them in themselves.

So I took the ones he got wrong, and three others that he had gotten right, and asked him to guess again, but this time I told him to imitate the eyes himself.  Doing it this way, he got them all right.  All of them.

Obviously, this surprised me.  Admittedly, it was a slow process, but of interest was why it was slow.  I watched him "get into character"-- it took five or so seconds to sculpt his eye muscles, individually, into the proper configuration, but once he had done this the answer came easily.

Which tells me that this kid has the ability not to "learn adaptive strategies" or "fake it till you make it," but truly access his own inner state and then apply it to others, i.e. to truly empathize.  What he seemed to be lacking is... practice?

So now I ask you to take the test yourself; when you click "Get Score" it will show you the ones you got wrong.  Cover the words, look at the numbers, and then go back and try to imitate the eyes for the ones you got wrong.  Did it help?


You might also enjoy:

Are You Good At Reading Faces?