The experiments I did on perception and memory were some of the first applications of information processing theory to autism. Hermelin and O’Connor had the brilliant insight to compare and contrast autistic children with mentally-retarded children who did not have a specific diagnosis, and with younger normally-developing children who were matched in terms of mental age on a variety of cognitive functions. This approach is now standard in the study of developmental delays.

In 1968 Neil O’Connor founded the Medical Research Council (MRC) Developmental Psychology Unit, and hired me as a research fellow, my dream job. In fact I am lucky enough to be still an MRC scientist to this day.

There have been some big changes. In the 1960s, psychoanalysts held sway and the general view was that autism is an emotional disorder with a psychological basis and the burden of blame was put on the mother. This was terribly wrong. Instead we believed that there must be a very specific fault in the brain that was responsible for derailing mental development. We believed that the mind is not a big bowl of spaghetti tangles, but more like a building with different floors and rooms. In other words, we began to think of the mind as developing in different ways for different cognitive capacities. This brought us closer to thinking about the brain basis of cognitive functions. It was the phenomenon of autism that encouraged us to think in this way. We found that autistic children had both stark deficits in some cognitive components, and outstanding talents in others. For instance, like normally developing children, they had the ability to form associations and to learn by rote. At the same time they were extremely poor at finding meaning and at reciprocating social overtures. But what was the cause of this failure? It took us a long time to find some preliminary answers.

In the mid-80s I developed a theory with Alan Leslie and Simon Baron-Cohen about a particular mental mechanism that was failing in autism. Our big idea was that autistic individuals lack a theory of mind. Once we had tested this idea through different experiments, with the help of Chris Frith and Francesca Happé I developed the ambition to look at the brain basis of this capacity.

It may be present in chimpanzees and bonobos, and maybe even in other species, but it’s not there in monkeys. In humans this network is active all the time; reputation management and political spin are only possible because of this feature of the human brain. The interaction between a teacher and a student is dependent on this as well, because in order to teach, you must make assumptions about what the student already knows and what they might find hard to understand. The ability to mentalize is hugely important for social interaction and in communication, and we believe it is impaired in autistic people.

I would characterize my own approach as a cognitive psychologist who is interested in developmental disorders from the point of view that there are certain innate mechanisms in the brain which we can make visible through brain imaging. The neural basis of the mechanisms may be faulty in some children, possibly as a result of faulty genetic programming. Many people now accept that the predisposition to autism and to dyslexia and other developmental disorders is largely genetic. How do we imagine these mechanisms to work, and how does learning and experience play its part? I think very much in terms of start-up kits that are already working in the brain from soon after birth. These start-up kits allow fast-track learning. From day one some mechanisms are already fully functional, such as those needed for recognizing faces and listening to voices and speech. Babies learn very fast what people around them look like and what language they speak.

The stories were based on a previous study by Francesca Happé, and Paul Fletcher was the young psychiatrist who was daring enough to take on the task of using positron emission tomography to study mentalizing. We examined which parts of the brain are active when the volunteer is asked to read a short story and then answer questions about the content of the story. Brain imaging studies require subtraction designs—you cannot just say, "Let’s see what happens in the brain when you read a story," because everything is active. We compared brain function for two types of story: one type required the volunteer to think about what the people in the story are thinking, and the other did not. For example, we compared brain activity for a theory of mind story, where you have to understand that the protagonist used a double bluff, with a physical story where you have to understand that it was more economical to buy in bulk. We found a specific pattern of activation associated with theory of mind. This study showed that the localization of brain regions involved in the attribution of mental states was feasible and that this should have implications for the neural basis of autism.

Since this first brain imaging study on mentalizing there have been many others with different materials and also different techniques, such as magnetic resonance imaging. Our 2000 Neuropsychologia paper⁴ reports impressively consistent results with stories and pictures. The emerging results from this and other studies prompted Chris Frith and myself to write a review paper², which speculated about the function of the different components in the brain that make up the mentalizing network.

Basically, the same regions are active as in normal volunteers, but much less so. Recently we did a brain imaging study where we showed animated triangles, either moving in a random fashion or interacting with each other. In the latter case, most people feel compelled to attribute mental states to these triangles—after all, that is why we enjoy looking at animated cartoons. We found that autistic individuals are much less inclined to do this.

We next compared the scans of normal and very able adults with autism or Asperger’s syndrome when viewing these animations. Again we found the brain’s mentalizing network. This was less active in the people with autism and, furthermore, it showed reduced interaction between the components of the network.

In autism there are not only social problems. From the very earliest description of autism in the 1940s, it was observed that these children often have narrow obsessive interests and may get captured by a minor visual detail in their environment. For example, they might zoom in on some very small part of the pattern on a decorated china teacup. Or they may notice that the cup handle has a small tea stain. First with Amitta Shah and subsequently with Francesca Happé, I have developed a theory to explain this tendency to focus on detail^3,5.

My hypothesis is that autistic children really do perceive details better than normal people, to the extent that they cannot see the wood for the trees. The nickname of the theory is "central coherence." What it means is that cognitive processing is normally geared towards extracting overall meaning and overall Gestalt at the cost of surface detail. So when your memory is overloaded (e.g., by the details of a long speech) what you will later recall is just the main message the speaker wanted to convey, not necessarily the speaker’s words. My hunch is that in autism the balance of processing both deep meaning and surface detail is tipped in favor of detail.

For instance, when their memory is full to capacity individuals with autism still do well giving verbatim accounts, but what they cannot get is the gist. They do not easily get to that level of processing information where you extract what it all means. We believe this is an interesting style of mental processing which isn’t always a disadvantage. For example if you can recall word for word what someone has said, or can really see visual details that others miss, this might have useful applications. This can lead to superior performance for certain tasks and shows that autism is not just a catalog of deficits. This idea made us think that perhaps weak central coherence is a feature of the broader autism phenotype and that biological relatives of autistic individuals may sometimes show this feature too. Indeed one of our recent studies shows that this assumption may be correct. The number of studies guided by the theory of central coherence has increased substantially in the last few years. It took a while for this rather outlandish idea to take off. But it also took a while for the idea of a failure of theory of mind to be widely accepted. The situation now is that these concepts need to be related to each other.

We have been trying to develop tasks that are sensitive enough to detect problems in theory of mind even in people who are so intelligent that they have already worked out logical solutions to the problems. But we do not think that autism means that every aspect of social knowledge is impaired. Right now we are interested in aspects of good social knowledge and social competence in autism. Our hope is that a brain imaging study currently in progress will tell us why people with autism may have problems in attending to social, but not to non-social stimuli.

Papers cited

1. Fletcher, P., Happé, F., Frith, U., Baker, S., Dolan, R., Frackowiak, R., and Frith, C. D., "Other minds in the brain: A functional imaging study of theory of mind in story comprehension," Cognition 57(2): 109-28, 1995.
2. Frith, C. D., and Frith, U., "Cognitive psychology – interacting minds – A biological basis," Science 286(5445): 1692-5, 1999.
3. Frith, U., and Happé, F. "Autism: beyond ‘theory of mind,’" Cognition 50: 115-32, 1994.
4. Gallagher, H. L, Happé F, Brunswick N, Fletcher PC, Frith U, Frith CD, "Reading the mind in cartoons and stories: an fMRI study of ‘theory of mind’ in verbal and nonverbal tasks," Neuropsychologia 38(1): 11–21, 2000.
5. Shah, A. & Frith, U., "Why do autistic individuals show superior performance on the Block Design task?" Journal of Child Psychology and Psychiatry 34: 1351-1364, 1993