What they found that so intrigued me was that, when looking at photographs of emotionally expressive faces, autistic children with better social and communication skills spent more time looking at the mouth rather than the eyes:
Analyzing whether there was a negative relationship between the amount of social and communication impairment and the factors differentiating between the two groups [i.e., the autistic and control groups], we analyzed the correlation between z-transformed ADI-R scores and the looking time data for the group with ASD. As there was a high correlation between the social and communicative impairment scales of the ADI-R (r = .83, p < .01), we combined these into a single averaged z-score. This score correlated negatively with overall total looking time in the whole screen (r = -.76, p < .01, one-sided). Furthermore, there was a negative correlation between social/communication ADI-R score and percentage of looking time directed at the mouth (r = -.62, p < .05, one-sided).
In other words, the higher the children scored on measures of impairment in verbal communication and reciprocal social behavior, the less time they spent looking at the mouth.
This might seem counterintuitive, if you're used to thinking of autism in terms of poor eye contact and avoidance of strong social stimuli --- the children who were most strongly "autistic" according to their degree of social and communicative impairment were the ones who displayed the most "normal" pattern of face scanning.
I think that's interesting, and points toward the conclusion I came to in this earlier post: that there's not a straight line connecting an obviously "autistic" behavioral phenotype with any particular impairment.
The study authors hypothesize that the presence of verbal ability far in advance of social skills might bias an autistic person to look more at the mouth for social cues, because the mouth is the source of language, and a verbal autistic person might try to harness their language-processing skills to help them decode these social signals:
Contrary to what Klin et al. (2002) found for adolescents with ASD when watching complex social scenes involving conversation, we found less scanning of the mouth in the group with ASD. However, in agreement with Klin et al. (2002), we found that looking at the mouth is negatively related to the severity of social and communicative impairments (notably, Klin et al. (2002) only included social scales). Specifically, we found that the higher the social and communicative disability, the shorter the time spent looking at the mouth. Why did Klin et al. find more mouth looking in ASD while we find an opposite trend? Given that Klin et al. studied adolescents with autism that had severe socio-emotional problems but normal intelligence, one must assume they were quite verbal (the intelligence test used in Klin et al. study includes many verbal subtests). It would not be surprising if a sample of persons with ASD biased in this way looked preferentially at the mouth rather than the more emotional eyes in a complex social setting involving conversation (see also Adolphs et al., 2005). Notably, we included children at all symptom levels in our study, not only the high-functioning individuals. If our interpretation of the discrepancy between our findings and the findings of Klin et al. is correct, it predicts the direction of abnormality of mouth scanning. It is related to the balance of socio-emotive and verbal skills of the participants, where predominantly verbal individuals will be expected to look preferentially at the mouth. To test this hypothesis, the sample must be heterogeneous regarding these two symptom areas.I think I should mention, also, that the two experiments being compared in this passage used very different kinds of stimuli; Klin et al. had their study participants watch clips from a film adaptation of Edward Albee's play "Who's Afraid of Virginia Woolf?", which naturally involved spoken dialogue as well as expressive faces, while de Wit et al. just showed a series of four still photographs of emotionally expressive faces.
(The presence of spoken language, whether in a movie or other video I'm watching or a real-life conversation, often leads me to look preferentially at the mouth of the person speaking, since I have difficulty understanding spoken language and find that reading a person's lips while also listening to their speech helps me understand it better. Maybe Klin et al.'s highly verbal autistic subjects have noticed the same thing).
One last thing: in the passage I quote from de Wit et al.'s discussion of their results, they use functional-level terminology uncritically, despite having produced evidence that somewhat perturbs that notion of an unbroken, internally-consistent hierarchy of greater to lesser autism severity. I have mentioned before that I am skeptical of functional levels as anything more than somewhat-tautological descriptors of IQ and language ability, and it is with this understanding (and also a desire to let the authors speak for themselves, even when I might not agree with them) that I have reproduced those terms here.
DEWIT, T., FALCKYTTER, T., & VONHOFSTEN, C. (2008). Young children with Autism Spectrum Disorder look differently at positive versus negative emotional faces Research in Autism Spectrum Disorders, 2 (4), 651-659 DOI: 10.1016/j.rasd.2008.01.004