To characterize the cognitive phenotype, the set of cognitive abilities that are impaired or spared in autism spectrum disorders.

Autism is a developmental disorder characterized by impaired social interaction and communication, and restricted and repetitive behavior. Because there are no biological markers for the disorder, it is defined and diagnosed purely behaviorally. But the precise cognitive phenotype is not well understood.

Understanding the cognitive phenotype of autism is crucial not only for discovering genetic and neural markers of autism, but also for understanding whether autism is truly a single disorder or rather a heterogeneous class of related disorders. Achieving these goals will help develop better diagnostic tools and treatments for autism.

The workshop focused on three fundamental questions: What are the cognitive functions that might be selectively impaired or spared in autism? What do we already know, and what do we still need to discover about these functions as they apply to autism? And finally, what are the best behavioral tools needed to better characterize the cognitive phenotype of autism?

After two days of vigorous discussion, researchers from a broad range of disciplines including psychologists, neuroscientists and clinicians reached consensus on three key issues.

First, the critical question in discovering the cognitive phenotype of autism is whether there are a number of distinct impairments that comprise the core phenotype of autism, or whether the many common impairments seen in autism are instead downstream effects of one basic core social or attentional impairment earlier in development.

Second, the current body of research describing impairments in autism is difficult to interpret due to contradictions and non-replications, possibly caused by limited sample sizes.

Third, rich, detailed modern scientific tools ― such as eye-tracking, psychophysical and neural measures ― are necessary to characterize cognitive abilities in individuals with autism. An example of a computational method used for a detailed characterization of infantsʼ facial expression is shown in Figure 2.

The discussion spanned four distinct cognitive domains that have been claimed to be affected in autism: social cognition; language; visual and social perception; and attention or executive control.

Social cognition encompasses a set of abilities that allow human beings to understand and interact with each other. George Gergely and Gergely Csibra described their work on the abilities of infants to learn through external social cues ― such as eye-contact, infant-directed speech, or even contingent interaction ― that signal adultsʼ intentions.

Young children with autism might lack the natural biases towards these cues that are found in typical development.

In older children and adults, Rebecca Saxe and James Blair suggested, the abilities that allow us to understand the thoughts and emotions of other people ― dubbed “theory of mind” and “cognitive empathy” ― are impaired in autism. Dan Messinger discussed the possibility that the impairment of emotional processing might also lead to further downstream impairments. Although there was a broad consensus that social impairments are a large part of the cognitive phenotype of autism, there was no agreement as to which, if any, of these deficits are at the core of the disorder.

Linguistic impairments present in autism range from mild abnormalities in prosody (vocal accent and intonation) and pragmatics (the ability to make social inferences beyond a sentenceʼs literal meaning) to severe or total impairment in the communicative use of language. Helen Tager-Flusberg suggested that this wide range might be the consequence of treating several distinct subgroups within the autism spectrum as a single research population.

Ted Gibson proposed that aspects of syntax ― the ability to learn and use the rules governing word order ― pragmatics and prosody in adult individuals with autism could be productively characterized using tools from psycholinguistics such as online language processing measures, including reaction times and event-related potentials, and spontaneous speech samples.

Based on research with children, Jesse Snedeker discussed the use of eye-tracking methods for evaluating pragmatic abilities (such as the ability to infer that “some of the candies” means “some ― and also not all ― of the candies”) and contrasted reports of pragmatic impairments in children with autism with other linguistic abilities that are relatively unimpaired.

These researchers concluded that in language, as in other domains affected in autism, it is unclear whether the observed impairments are part of the disorderʼs core phenotype or whether they are a downstream effect of social impairments earlier in development.

Individuals with autism display a complex array of impairments in visual and social perception. Based on results from eye-tracking experiments, Ami Klin demonstrated the tendency of adults and children with autism to look at mouths rather than eyes to interpret emotions. Liz Pellicano suggested that face processing in children with autism might also be impaired, but this impairment might stem from a primary deficit in social motivation.

However, individuals with autism can detect people and animals in complex, natural scenes just as well as controls can, suggesting a nuanced pattern of impairment, according to evidence presented by Brian Scholl. Reviewing findings on motion processing deficits, David Whitney suggested that those with autism do not have a deficit in motion processing per se, but that these impairments may be caused by deficits in attention.

Finally, Kevin Pelphrey reported that based on functional magnetic resonance imaging, individuals with autism show atypical activation in the region of the brain that is putatively responsible for the perception of social cues such as eye gaze and biological motion ― the characteristic motion patterns shown by living things ― suggesting that these abilities might also be affected in autism.

There was widespread consensus among the attendees that, in future studies, there is a need to use rich, detailed measures (such as eye-tracking, psychophysical tests, and functional neuroimaging). Such measures have the best chance of detecting reliable performance patterns in a single individual rather than at the group level. They may also allow researchers to identify distinct cognitive impairments rather than tapping broader issues in task understanding or motivation.

There is wide consensus that attention and executive control ― the complex of abilities necessary to focus on, plan and shift between different goals in the course of everyday activities ― are impaired in individuals with autism, although the specific nature of the impairments is still unknown. Deficits in control processes might even be the primary or core feature of autism early in development, causing later social, perceptual and linguistic deficits.

John Gabrieli suggested that brain regions that support the distinct functions comprising executive control ― action planning, working memory, inhibition, shifting, selection ― might be differentially impaired in autism. Claire Hughes described existing executive control and attention tasks and argued that the pattern of impairment is mixed ― suggesting that deficits in executive control might be specific to some subset of these abilities.

John Duncan explained that a major confound in current studies of executive control is that they do not control for general fluid intelligence ― the basic processing capacity that supports performance in nearly every cognitive task. Studies that investigate impairments in executive control should independently measure fluid intelligence.

Jeanne Townsend discussed impairments in visual attention, suggesting again that these deficits are important in understanding the developmental progression of autism spectrum disorders, and suggested that neuroimaging methods such as electroencephalograms could be productively used in characterizing attentional deficits in children with autism.

Yuhong Jiang also focused on visual attention, and hypothesized that individuals with autism might be impaired in their ability to shift attention. She then asked whether the problem with shifting attention might be related to an impairment in learning where to attend ― for example, to faces or to other potentially relevant stimuli.

Autism researchers face numerous methodological challenges. For example, Cathy Lord described how some individuals with autism perform within ― and sometimes above ― the normal range on experimental tasks in a laboratory setting, and yet fail at using these same abilities in real-life situations.

The cognitive profile of individuals with autism also changes considerably over the course of development, so studies of adults and children may reach very different conclusions. A study that finds a real pattern of impairment early in childhood may later find a complex pattern of compensatory strategies. Complicating these problems further is the fact that individuals with autism are a highly heterogeneous group, and so studies with small sample sizes often generalize very poorly to larger populations.

Finally, even with a representative sample, any study on cognitive functions in autism may include task demands that put individuals with autism at a disadvantage without measuring a true impairment in the domain of interest.

Nancy Kanwisher proposed that a large scale study of the cognitive phenotype of autism, incorporating many participants (N) and many tasks (T) might address some of these methodological challenges. This ‘large N, large Tʼ study would include rich measures such as eye-tracking, neuroimaging and psychophysical measurement, providing a detailed characterization of participants in the domains discussed during the meeting.

Working with a large group of participants who have already been assessed with standard diagnostic tools and for whom genetic information is already available ― such as, perhaps, the Simons Simplex Collection ― could allow researchers to gather a dataset that would be unprecedented in its scope and level of detail.

This dataset would open up a host of new research possibilities, including the ability to use powerful statistical techniques to discover subgroups, to search for neural correlates of behavior, and to discover shared genetic causes for specific behavioral impairments.

The meeting concluded on a high note: the combination of techniques from cognitive science and cognitive neuroscience with large-scale clinical and genetic studies might yield new results in the search to characterize and understand autism.