baroncohen

Another Advanced Test of Theory of Mind: Evidence from Very High Functioning Adults with Autism or Asperger Syndrome

Simon Baron-Cohen and Therese Joliffe
University of Cambridge, U.K.

Catherine Mortimer and Mary Robertson
University of London, U.K.

Journal of Child Psychology and Psychiatry, Vol. 38, No. 7, pp 813-822, 1997

Introduction

There is considerable evidence that the majority of children with autism have impairments in the development of a theory of mind. Such , a deficit may underlie the social, communicative, and imaginative abnormalities that are diagnostic of the condition, since a theory of mind is necessary for normal development in each of these three areas. The theory of mind deficit appears to be expressed very early, from at least the end of the first year of life, as joint attention deficits.
However, there is some evidence that appears to contradict the notion that a theory of mind deficit is a pure cognitive deficit in autism. First, Bowler (1992) found that adults with Asperger Syndrome - who share the social and communicative symptoms of autism but who have no history of language delay - pass second-order theory of mind tests. Second-order theory of mind test involve the participant reasoning about what one person thinks about another person's thoughts. Ozonoff et al.(1991) also found some adults with "high functioning autism" or Asperger Syndrome who passed second-order theory of mind tests. Both studies contradict an earlier finding that there is an impairment in theory of mind ability in autism. In the latter study, no participants with autism passed the second-order test of theory of mind.
However, these studies cannot be taken as conclusive evidence for an intact theory of mind in such individuals with autism or Asperger Syndrome, because such second-order tests can easily produce ceiling effects if used with participants with a mental age above 6 years. This is because children with normal intelligence pass second-order theory of mind tasks at about 6 years. It is unfortunate that many workers in this field have thought of second-order tests as "complex" or high-level tests of theory of mind. Certainly, they are more complex than first-order tests (in which the participant simply has to infer the thoughts of another person)-but recall that normal 4-year-old children pass first-order tests.
In short, neither first- or second-order tests are complex tests of theory of mind. They are simply probes for 4- or 6-year-old level skills, respectively, in this domain. Perhaps if they had been labeled as such, no-one would ever thought of them as suitable tests of whether an adult (with autism, Asperger Syndrome, or any other condition) has a fully functional theory of mind. Finding a 30-year-old individual with autism, of normal intelligence, who can pass a theory of mind test at the level of a normal 6-year-old does not lead to the conclusion that they are necessarily normal in this domain. All we can conclude is that they have intact theory of mind skills at the level of a 6-year-old.
Happe's (1994) study is the only one to take this issue seriously. She tested adults with autism or Asperger Syndrome on an "advanced" theory of mind task. This involved story comprehension, where the key question in the task either concerned a character's mental state (the experimental condition) or physical events (the control condition). Happe's task was pitched at the level of a normal 8-9-year-old, and in this respect, it is certainly more advanced than previous tests of theory of mind. She found that adults with autism or Asperger Syndrome had more difficulty with the mental state stories than did matched controls, and that they used fewer appropriate mental state terms in their justifications for why characters behave as they did.

Method

In the original study the authors label the discussion of how the study was laid out as an "Experiment". I point this out, not because it is incorrect, because it is correct. To be an experiment, it must have an Independent Variable (I.V.) that varies. Since the procedure for all participants is the same, with a randomised order of testing, the thing that varies is the type of participant. This does not vary because of experimenter manipulation, but it varies naturally. So, what we have here is a Natural Experiment.

Sample

Three groups of participants were tested.
Group 1 This comprised of 16 participants with high-functioning autism (HFA, N = 4) or Asperger Syndrome (AS, N = 12), the sex ratio was 13:3 (m:f). The HFA Group all showed a history of "classical" autism (i.e. autism accompanied by language delay) and fulfilled established diagnostic criteria (American Psychiatric Association, DSM-IV, 1994). Note that because they were high-functioning adults, they would be considered "residual" cases. The AS group met all the same criteria for autism, but without any clinically significant language delay (i.e. they had single words by age 2 or phrase speech by age 3, as reported by their parents). They thus met criteria for Asperger Syndrome as defined in ICD-10 (World Health Organisation, 1994). They were all of normal intelligence. As such, the are relatively rare. They can be considered as cases of "pure" autism or Asperger Syndrome, unconfounded by mental handicap. They were recruited from a variety of clinical sources, as well as an advert in the National Autistic Society magazine Communication.
This group were selected for being of at least normal intelligence (i.e. scoring >85) on the Wechsler Adult Intelligence Test Revised Edition (WAIS-R) (Wechsler, 1981; full scale performance and verbal IQ). The WAIS-R was used because of previous work showing discrepancies between performance and verbal IQ in these groups. We therefore ensured that these participants had an IQ>85 on both verbal and performance scales.
Group 2 This comprised 50 normal age-matched adults (25 male and 25 female) drawn from the general population of Cambridge (excluding members of the University), all of whom had no history of any psychiatric condition (as established by self-report). They were selected randomly from the subject panel held in the University Department. IQ information was not collected on this group, but they were all assumed to have intelligence in the normal range.
Group 3 This comprised 10 adults with Tourette Syndrome (TS), also age-matched with Groups 1 and 2. The sex ratio was 8:2 (m:f) thus mirroring the ratio in Group 1. They were attending a tertiary referral centre in London, and had all been diagnosed by a leading expert in the field of Tourette Syndrome (Dr. Mary Robertson), on the basis of meeting DSM-IV criteria for Tourette Syndrome.
This group were also selected for being in the normal IQ range, prorated from four subtests of the WAIS-R (Block Design, Object Assembly, Vocabulary, and Comprehension). This short version of WAIS-R was administered to the participants with Tourette Syndrome because of its brevity, since these participants were only available for limited testing in this study. It correlates (0.91) with full-scale WAIS-R IQ.

Finally, participants in the two clinical groups were selected for being able to pass two first-order belief tasks, and a second-order false belief task. This was so that if any deficits were found in the Eyes Task, they could not be attributed to mindreading problems beyond a 6-year-old level. In fact, this did not exclude anyone from this study.

Chronological Age (CA) and IQ of the participants in each of the groups.

	CA	WAIS-R IQ
Autism/Asperger Syndrome (N=16)
Mean	28.6	105.31
S.D.	9.7	13.0
Range	18-49	86-133
Normal (N=50)
Mean	28.6
S.D	9.7
Range	18-48
Tourette Syndrome (N=10)
Mean	27.7	103,5
S.D.	7,81	10.0
Range	18-47	93-115

Procedure

The Eyes Test
The test comprises photographs of the eye region of 25 different faces (male and female). The faces were taken from magazine photos. All faces were standardised to one size (15 x 10 cm.), and all were black and white, with the same region of the face selected for each photo - from midway along the nose to just above the eyebrow. Each picture was shown for 3 seconds with a forced choice between two mental state terms printed under each picture. The experimenter says to the participant "Which word best describes what this person is feeling or thinking?" The maximum score on this test is 25.

Construction of the Eyes Test
The target word to describe the mental state behind each pair of eyes was generated by four judges (two males and two females) in open discussion. A foil word was selected that was the semantic opposite of the target word, in all cases. These were then tested on a panel of eight judges (four males and four females) who were all independent raters, blind to the hypothesis of the study. On the target words, there was unanimous agreement by all eight independent raters. The full set of words used is shown below. Please note that both basic and complex mental terms were used.

Target Mental State Terms, and Their Foils.

Number	Target Item	Target Term
1	Concerned	Unconcerned
2	Noticing you	Ignoring you
3	Attraction	Repulsion
4	Relaxed	Worried
5	Serious Message	Playful Message
6	Interested	Disinterested
7	Friendly	Hostile
8	Sad Reflection	Happy reflection
9	Sad thought	Happy thought
10	Certain	Uncertain
11	Far away focus	Near focus
12	Reflective	Unreflective
13	Reflective	Unreflective
14	Cautious about something over there	Relaxed about something over there
15	Noticing someone else	Noticing you
16	Calm	Anxious
17	Dominant	Submissive
18	Fantasising	Noticing
19	Observing	Daydreaming
20	Desire for you	Desire for someone else
21	Ignoring you	Noticing you
22	Nervous about you	Interested in you
23	Flirtatious	Disinterested
24	Sympathetic	Unsympathetic
25	Decisive	Indecisive

Validity of the Eyes Task
The "Eyes Task" was designed to be a "pure" theory of mind test, at and advanced level. This is because, as far as we can see, the test itself itself has no planning or context component. As mentioned earlier, in order to validate the Eyes Task as a theory of mind task, participants in the two clinical groups were also tested on Happe's Strange Stories. In the case of the subject with autism and Asperger Syndrome, this was part of a separate study. If performance on the Eyes Task was indeed tapping theory of mind, then performance on the Eyes Task should correlate with performance on Happe's Strange Stories.

Control Tasks
Finally, in order to check whether deficits on the Eyes Task were due to other factors, we administered two control tasks in Group 1.
Gender Recognition Task This involved looking at the same sets of eyes in the experimental task, but this time identifying the gender of the person in each photograph. This is a social judgement without involving mind reading, and allowed us to check if any deficits on the Eyes Task could be attributed to general deficits in face perception, perceptual discrimination or social perception. This had a maximum score of 25.
Basic Emotion Recognition (Emotion Task) This involved judging photographs of whole faces displaying "basic" emotions. This was to check whether any deficits on the Eyes Task could be attributed to a deficit in basic emotion expression recognition. Six faces were used, testing the following basic emotions: happy, sad, angry, afraid, disgusted and surprised. This test differs from the Eyes Task in two ways:

the Emotion Task shows information from the whole face, not just the eyes.
the Emotion Task only checks perception of basic emotions, the Eyes Task tests a fuller range of emotions.

Results

The results of the Eyes Task for all participants is shown below:

	Eyes Task	Gender Control
Autism/Asperger Syndrome (N=16)
Mean	16.3	24.1
S.D.	2.9	0.7
Range	13-23	23-25
Normal (N=50)
Mean	20.3	23.3
S.D.	2.63	1.1
Range	16-25	22-25
Tourette Syndrome
Mean	20.4	23.7
S.D.	2,63	0,9
Range	16-25	23-25

Autism/AS x Normal, t=-5.16, 64d.f., p=0.0001

Autism/AS x TS, t=3.98, 24 d.f., p=0.001

Inspection of the results with reference to the difference between Austism and Asperger Syndrome participants did not reveal any differences, but as the sample sizes were so small, a statistical test would have not been worthwhile.

Performance by Male and Females in the Normal Group on the Eyes Test

	Eyes Task	Gender Control
Males (N=25)
Mean	18.8	24.0
S.D.	2.53	0.6
Range	16-22	23-25
Females (N=25)
Mean	21.8	23.8
S.D.	1,78	0.6
Range	20-25	23-25

Males x Females, t=4.8, 48 d.f. p=0.0001

As predicted, female participants perfromed much better on the Eyes Test than male participants. This supports the folk psychology that females show more social awareness and empathy than males.

Conclusion

The authors predicted that despite the HFA/AS group having an IQ within the normal range, they would show a poorer performance on the Eyes Task, which shows that it is a subtle theory of mind test.
The authors also predicted that females would do better on the Eyes Task than males, that was confirmed also.
The authors said that a poor performance on the Eyes Task could not be explained by a poor IQ as all the participants in this study had IQ's that were within the nomal range. The poor performance was not due to neurological problems either as the Tourette group suffered from a neurologiccal problem and did much better than the HFS/AS group.
The pattern of scores on the Eyes Task mirrored the scores found on Happe's Strange Stories Test.
Some in the HFA/AS group had University degrees, but still performed poorly on the Eyes Task.
Even though the Eyes Task is a subtle theory of mind test it is much simpler than real life. Some in the HFA/AS group mentioned in passing that going to the cinema is a frustrating experience as they don't understand why someone did or said something, who knows what and who doesn't, and why people laugh at certain times.
Impairment in the Eyes Task mirrors the difficulties in autism e.g. they sometimes fail to interpret gaze direction as a cue to a person's state of concentration
The differences between the males and females in the normal group may be down to genetics or socialisation. It was pointed out that while females excelled at the Eyes Task there are situations where males tend to out perform females such as visuo-spatial problems.