What goes wrong in the mind - in the brain - from before birth to cause developmental disorders? I have focused particularly on autism and dyslexia, and aim to discover the underlying cognitive causes of these disorders and to link them to behaviour patterns as well as to brain systems. I would like to make this research relevant to the education of people with developmental disorders and to improve their quality of life.

Concerning autism, my work has explored mentalising and central coherence to explain the social and non-social features of autism. Concerning dyslexia, my work has explored access to phonological and orthographic representations to explain the delay in reading and spelling development.

In 2012 wrote a review paper of autism research from a personal perspective: Why we need cognitive explanations of autism. It can be downloaded from file archive.

Why study autism and dyslexia?

Dyslexia and autism are disorders with a genetic cause and a basis in the brain. To understand them, it is necessary to consider many factors, those that are internal to the person and have a biological basis, and those that are external, and have a basis in the social and cultural environment. The challenge is to explain why sometimes learning of a particular skill can be so difficult, while at the same time, learning of other and just as complex skills can be easy. One answer that such a pattern suggests is that there are start-up kits in the brain and in the mind that make learning almost effortlessly easy, and that sometimes these start-up kits are faulty or missing.

In autism my work has explored a failure in intuitive mentalizing and how compensatory learning about mental states proceeds slowly and in an effortful fashion. This work indicates that adults with high-functioning autism who can understand mental states and can use a theory of mind to explain other people’s behaviour, still bear signs of the struggle to learn these skills which come effortlessly to normally developing individuals. Their intuitive mentalising ability remains poor. This is revealed in reduced activation in the brain’s mentalising network.

Concerning dyslexia, my work has explored difficulties in a start-up kit that leads to rapid speech processing. Here it can be shown that compensatory learning leads to perfectly adequate speech processing, and to adequate processing of written language. However, even well-compensated adults show that the skill they acquired is fragile and lacks the intuitive underpinning of phonology. In particular, it seems that dyslexics never gain the immediate access to the mind’s store of word sounds and word images that represent their letter-by-letter pattern (orthography). Brain imaging shows that well compensated adults have reduced activation in areas of the brain critically involved with rapid access to phonological and orthographic forms of words.

Examples of work on autism

Awareness of feelings and empathy 

The definition and diagnosis of autistic disorders at present has led to a very heterogeneous population. We tested the widely held belief that Asperger Syndrome may predispose to antisocial and callous acts of behaviour and showed that this is not the case (Rogers et al. 2007). Our results indicate that antisocial and callous behaviour was unrelated to severity of autism, and unrelated to core autistic cognitive deficits, specifically in 'mind-reading' or executive function. Instead the cognitive profile of boys with co-occurring autistic disorder and psychopathic tendencies resembled that of boys who have psychopathic tendencies alone. We therefore proposed that callous and psychopathic acts, which are regrettably observed in a small number of individuals with ASD, reflect an additional impairment of empathic responsiveness which is not part and parcel of autism itself. This finding has implications for forensic cases.
In an adult sample we found a wide range of self-reported empathic concern, from very low to high, which correlated with self-reported awareness of own emotional states, indicating a link between understanding one’s own and others’ emotions. In previous studies we found that a high proportion of the adults with autism spectrum disorder we tested showed low awareness of their own feelings and bodily sensations, called alexithymia (Hill et al. 2004). Work on alexithymia is being continued by Elisabeth Hill. We investigated the neurophysiological underpinnings of alexithymia in an fMRI study (Silani et al. in press). The results showed that when monitoring inner feelings evoked by pictures as compared to judging the colour content of the pictures, the brain’s mentalising system (Castelli et al. 2002) was activated. This activation was reduced in individuals with ASD. Quite independently, the degree of alexithymia and lack of empathy correlated negatively with degree of activation in the anterior insula (AI). This is a region of the brain that has been previously found to be involved in the processing of internal arousal and bodily states. Our results confirm that the AI may serve the awareness of own feeling states, and may also underlie the capacity to empathize, that is to share the feelings of others who are, for example, in pain or experiencing disgust (Singer et al., 2004). Further fMRI studies on empathy in autism are continued by Tania Singer.

Lack of top-down modulation

It is widely believed that individuals with autistic spectrum conditions see the world differently. Stimuli appear to be salient that are not very salient to other people, and vice versa. The lack of salience of social stimuli is particularly noticeable. We have pursued the hypothesis that these stimuli are normally enhanced. This hypothesis fits into a wider theory of lack of top-down control. Such a high-level failure would cause a lack of the enhancement of stimuli that are relevant to social communication. Likewise it would cause a lack of suppression of stimuli that are not currently relevant. Thus, lack of top-down modulation can explain perceptual overload.

As an example, one of our recent studies, led by Geoff Bird, attempted to find out to what extent selective attention to social and non-social stimuli modulates brain activity in autism. Normally, modulation occurs such that brain areas known to be active during face processing and object processing show enhanced activity when attention is directed to a location where these stimuli are going to appear. This modulation was almost absent in autistic adults when faces were compared with houses (Bird et al., 2006).

Text comprehension

An additional study on text comprehension (Saldana and Frith, 2006) yielded results that are consistent with this conclusion. We found that, surprisingly, autistic readers with comprehension problems were primed by implicit inferences while reading just like normal readers. Thus their comprehension problems cannot be attributed to an inability to make implicit inferences or an inability to draw on relevant world knowledge. Instead their poor comprehension must be due to problems at the discourse level of text processing. This has educational implications as it questions the widely-held belief that poor comprehension is caused by lack of low-level inferences. Instead, it appears that the poor comprehenders fail to make use of their perfectly adequate inferences. This work and work on hyperlexia is being continued by David Saldana at the University of Seville.


A project, led by Antonia Hamilton, investigated possible impairments in the perception-action mirror system, which is at the centre of one of the current theoretical debates about basic mechanisms of the social brain. She found no evidence for an impairment in the imitation of goal-directed actions, which would have been predicted from the mirror neuron hypothesis of autism (Hamilton et al. 2007).

Signalling that a communication is forthcoming

In one of our neuroimaging studies with normal adults (Kampe et al, 2003) we attempted to delineate brain regions that support the initiation of communication by making eye contact or by calling someone by name (see right). Increased activity in MPFC and temporal poles was found in a conjunction analysis for both modalities. A lack of orientation to direct gaze and to hearing one's own name called is one of the earliest signs of autism (Osterling et al 2002). Hence understanding the circuitry involved in this process should aid our understanding of possible neurological dysfunction in autism.

This study also resulted in a serendipitous finding: The effect of eye gaze is modified by facial attractiveness. Thus, brain regions associated with reward expectation are activated only in the condition where an attractive person is gazing at you (Kampe et al 2001).

Islets of social abilities

In a collaboration with Larry Hirschfeld we found an important islet of social ability in children with autism (Hirschfeld et al. 2007). While understanding others in terms of their mental states was impaired in these children, understanding others in terms of their group membership was not. We used the PRAM, which presents pictures and vignettes of males and females, brown and white, to assess social stereotypes in children. Surprisingly, given their lack of interest in social information and limited opportunities for social learning, autistic children showed good knowledge of gender and racial stereotypes.

The neuro-physiological basis of mentalising

We have used a variety of types of stimuli to examine brain activation associated with mentalising in healthy volunteers, including short stories (Fletcher et al 1995; Vogeley et al 2001), cartoons (Gallagher et al 2000), animated shapes (Castelli et al 2000). A meta-analysis of mentalising studies revealed remarkable convergence (Frith & Frith, 2003) in three key regions of theory of mind-associated activity: a) medial prefrontal cortex (MPFC), b) temporal poles and c) superior temporal sulcus (STS).

In a structural MRI study of high-functioning individuals with autistic disorder, whole brain voxel morphometry (VBM) revealed abnormalities in grey matter volume in a number of cortical and subcortical regions, including MPFC, temporal poles, and cerebellar regions (Abell et al 1999). The coincidence of two of these regions with two components of the mentalising system is striking. These are preliminary results, as this technique is still undergoing development. However, unlike other types of structural analysis, VBM is entirely objective and not confounded by brain size. We plan to apply the technique in future imaging studies.

Tasks to probe intuitive mentalising in high-functioning adults 

Surprisingly, despite the vast amount of research published on theory of mind (numbering more than 1,500 papers to date), most of it concerns young children, and it remains unclear whether stable individual differences in social insight can be measured in ordinary adults. We have been working on developing new tasks to probe mentalising abilities in adults. Some examples are:
  • Silent animations. 
    We designed a set of animations featuring two triangles, which convey different types of action/interaction purely through their kinetic patterns. You can download some example movies below. The files are approx 4MB, and the format is avi with Cinepak compression, so they should be visible in Windows Media Player or Quicktime

    If you are a researcher who would like to use these animations, please contact Sarah White.

Cognitive style in autism: Weak central coherence

We proposed that autism is characterised by 'weak central coherence' (review by Happé & Frith, 2006). To establish whether this cognitive style is a primary feature of autism we assessed central coherence in the first-degree relatives of children with autism, dyslexia and normal development (Briskman et al 2001; Happé et al 2001). Half of a group of fathers of children with autism showed consistent detail-focus, which is typical of weak central coherence. A review on visual perceptual characteristics in autism summarises the current debate on strengths and weaknesses in the cognitive profile of autism (Dakin & Frith, 2005). Sarah White is currently researching weak central coherence at the ICN and Francesca Happé and her team at the Institute of Psychiatry.

The theory of weak central coherence attempts to account for the amazing talents that can occur in people with autism. An example of an artist illustrating a detail-focussed processing style is Gilles Trehin, who creates imaginary vistas of the ever expanding city of Urville.

Examples of work on dyslexia

A comparison of current theories of dyslexia

Our studies allowed us to conclude that sensorimotor dysfunctions have only a limited prevalence in dyslexia (about one third for each of auditory, visual and motor deficits) and are not universal (White et al., 2006, Ramus et al., 2003). Thus, low level sensorimotor deficits cannot explain the phonological and literacy deficits in dyslexia.

In one study where reading was compared between children with autism and children with dyslexia, a double dissociation was found between phonological and sensorimotor abilities (White et al., 2006).

Studies on reading, phonology and dyslexia are continued by Franck Ramus.

Development of tests to assess phonological impairment

In collaboration with educational psychologists, we developed a standardised test of phonology (Frederickson, Frith & Reason, 1997) which is now widely used. Our experimental studies showed that the tests in this battery are powerful indicators of dyslexia, with phonological ability being a strong predictor of literacy skill. We have also developed phonological tests for adult dyslexics.

Longitudinal study of children at familial risk for dyslexia

In collaboration with Maggie Snowling at York University we carried out a study that has yielded new information about the course of dyslexia over the crucial period from preschool to school age (Snowling, Frith & Gallagher, 2000; Snowling, Gallagher & Frith, 2003). 60% of the 56 children at familial risk had significant literacy impairments at age 8.

Cross-cultural studies of skilled reading and dyslexia

This project, which Uta Frith coordinated with collaborators Eraldo Paulesu (Milan) and Jean-Francois Démonet (Toulouse), was partially funded by the European Biomed programme. 

In the first study, normal skilled Italian and English readers (using respectively a shallow and deep orthography) were compared during word reading or simple exposure to print (Paulesu et al 1999). PET scan results revealed an 

extensive reading system in the left hemisphere that was common to both groups. However, the different components of this system were weighted differently. Italian readers showed comparatively stronger activation of the planum temporale, which has been linked to grapheme- phoneme decoding. Conversely, readers of English relied more strongly on the inferior-posterior temporal gyrus (BA 37), a region linked to the orthographic lexicon.

The main aim of the project was to investigate the neurophysiological basis of dyslexia in speakers of different languages. We designed matched phonological and reading tasks in different languages and established that on these measures a striking deficit is found when dyslexics in all three languages were compared with normal controls. The results revealed a common pattern of reduced activation, which may represent a neural signature for dyslexia (Paulesu et al 2001). This work is being continued by Eraldo Paulesu.

Subpages (1): Experimental Stimuli