Face Perception in Autism
Face Perception in Autism
Interview with Dr Owen Churches by Lorna Rouse
Your PhD investigated face perception and Autism Spectrum Conditions, when did you first become interested in this topic?
Like a large number of cognitive neuroscientists about my age, I became very interested in face perception after reading Oliver Sack's book The Man Who Mistook His Wife For A Hat. In the eighties and nineties there was a burgeoning of research on face perception and in particular the neurology of face perception that created remarkable insights. It felt like this was a field that was not only new and exciting but that would perhaps hold great promise to patients with some incredibly profound deficits.
What drew you to focus on face perception in autism?
A combination of reading Simon Baron-Cohen's works on 'theory of mind' and a few coincidences in my undergraduate education which gave me the opportunity to work with people with autism in a research programme examining the efficacy of an intervention programme for young children with autism. From talking to the researchers, the families and sometimes to the patients, I realised this was not only an incredibly important field where answers were needed very quickly, but it's quite disconcertingly incredibly interesting simply in its own right. A disorder that affects social ability is unlike a disorder that affects low level vision or hearing, this is something that seems to go right to the core of the psychology of the individual.
Was there any evidence that face processing might be different for people with autism?
Yes there was a huge wealth of data on face perception in autism to the extent that when I applied for my first grant to fund my PhD research the funding body came back and said 'please don't send us any more grant applications for face perception studies in autism. We know that people with autism have trouble with face perception.' In a superficial sense they were quite right, we have known for a long time that face perception is a problem in autism, but the mechanisms underlying this are still very poorly understood. Autism has always been defined, at least in part, by deficits in social ability and of course it's plausible that as faces are probably the most salient social stimuli in the environment, if one has a problem with social functioning one would also have problems with face perception. The first time that this was tested experimentally was by Landell in 1978 in what is now perhaps one of the most highly cited papers in autism research and it was really this paper that ushered in a whole wealth of studies of face perceptual ability in autism.
What were the main aims of your project?
Like a large number of PhD students, my aims were vast, grand and also entirely delusional! I originally planned to undertake an intervention study. Ofer Golan, a former student of Simon Baron-Cohen's, did a great intervention study looking at the Mind Reading self-paced intervention programme, which showed good promise for helping people with autism better understand emotion from faces. I planned to do an intervention study to investigate the neural mechanisms wherein this improvement in face perception was gained. That is to say, is the normalised face perception behaviourally created by a normalised neurological activity, or does this very systematic process by which face emotion is learnt actually create a whole new way of evoking the same behaviours used by the typically functioning population? After about six weeks I abandoned that entirely on logistical grounds. I realised that there simply weren't the resources available so I scaled back completely and decided to focus on a series of very specific experimental manipulations of faces and to record EEG (Electroencephalogram) as a measure of neural activity in people with autism and matched controls.
What made you choose EEG?
Although I've developed a great respect and I might even say fondness for EEG and in particular for the ERP (event-related potentials) technique, the simple fact is that on a per study basis it is about a hundred times cheaper to collect EEG than fMRI data. At the time I was doing my PhD neither of my supervisors had the money to do fMRI studies. The downsides of using EEG are evident; on one hand there's the very poor spatial localisation of the pathology that one might find. The well publicised upside amongst EEG advocates is the very good temporal resolution giving great certainty about when things are happening, but there is another upside and that is that EEG is the DIY of neuroimaging. I really did just do my PhD on my own, it's technology that someone with a basic knowledge of physics could understand and that lets you do it yourself, this of course is one of the reasons why it's so cheap, but there is a great sense of ownership and pride in using EEG. Having worked on fMRI studies since I always feel surrounded by physicists, radiographers , radiologists, neurologists and multiple Principle Investigators. These are hugely collaborative endeavours and of course one learns a great deal from working with so many people from such diverse backgrounds, but there is something quite delightful about literally putting together the equipment yourself, recording, analysing and knowing that your thesis is your very own.
Could you briefly explain how EEG works?
In the very outer layer of the cortex are arranged a type of cell called parametal cells and one of the lovely things about parametal cells from a neuroscientist's perspective is that they are aligned parallel to each other, not necessarily pointing straight out because of the bumps and folds of the brain, but at the very least they are always parallel to their neighbours. What that means is that when an area of the cortex that's functionally specified to a particular cognitive event, like seeing a face, then that population of cells is summated not only in time but also in electrical polarity, so the dipole that's created around every single one of those cells is magnified by the number of cells activated and in some cases one is incredibly lucky enough to find that the potential difference created by that local activation of cells is so large and pointed in such a direction that it can be recorded on the surface of the scalp. At its core what one is doing is recording the instantaneous potential difference between one point on the scalp and another point on the head. For a series of methodological and historical reasons most face perception people use the nose as the reference to which all the scalp sites are measured. After measuring that potential difference as an analogue signal, and hence every infinite instance of the signal, it is then digitised and viewed in real time as a series of very squiggly lines and this simply reflects the ongoing oscillation of different populations of cells in the cortex that are activated by different events and whose polarity competes with each other to create positive and negative going oscillations in an EEG. The ERP is created when, at the same time as recording and digitising that analogue signal of brain activity, one also records the event of interest like the presentation of a face and then cuts up and averages all the squiggly lines associated with that one particular event together. Once this is done the activation that was common to all those events, that is activation of that population of cells related to, in my case face perception, are evident.
Right, so for example, when participants are viewing either faces or objects you see different kinds of squiggly lines.
Exactly! In my case I was interested only in the region of the cortex that serves face perception; those components that are generated in the posterior superior temporal sulcus and fusiform face area. What I was looking at was simply the difference between faces and other objects in those components.
Were there any challenges to using the EEG equipment with people with autism?
Yes, one characteristic of people with autism is hypersensitivity to touch and in some cases a greatly decreased pain threshold. Recording EEG from members of the typical population is not something that is ever described as painful, it is like wearing a very ill fitting swimming cap, but for some people with autism this was a big challenge. In some cases of course, owing to the heterogeneity of the condition, it was no challenge at all. I found that it helped to describe in great detail to my potential participants exactly what would happen a long time before the experiment took place. Unlike the typical population who probably would do better if something that might be mildly discomforting was a bit of a surprise without a long build up, discussing the process in great detail seemed to allay my participants' fears and prepare them well.
Before you began the experimental stage of your research did you have any expectations about what you might find?
I ran a series of different experiments and had a series of hypotheses to match each. The first experiment looked at the effect of attention on face perception in autism. A lack of attentional modulation of face perception in autism was already well documented and there was a well documented effect of attention on face selective ERP components in a typical population so I quite simply put the two together in the one experiment. The results were consistent with the surrounding literature; showing decreased, in fact nonexistent modulation due to attention in face selective components in autism. In the second experiment I wanted to look at the parameters of the 'faceness' of face selectivity of these ERP components, I wanted to see how far they could be pushed - just what a face is. At first I thought of comparing simply faces and non-face objects but it occurred to me that it would be interesting to have not just those two ends of the face spectrum but to include something in between real faces and objects. These were objects that aren't faces but look a bit like faces and many people who are typically functioning will be familiar with the phenomena of glancing at a house, or the moon and thinking, that looks a bit like a face. I presented these three categories and recorded the face selective components following them.
How did you select objects for the face-like category?
I started off becoming a little obsessed with things that look like faces before I actually found a book of bizarrely face-like objects that a French photographer had previously collected and he and his publisher agreed that I could use his work. I first conducted a study in a separate typical population where I presented a very large number of stimuli that I thought may be face-like and asked them to rate how face-like they were and from that group I found the objects most frequently rated as face-like stimuli. What I found was that people with autism had a core activation of face selective components to actual faces, but that this activation tapered off rather quickly. It didn't generalise across the non-face and the face-like objects as it did in the typical sample, suggesting that the face processing was more conscribed in autism; it's limited to only the most stereotypical examples of what a face can be. This of course is a problem because outside the laboratory when you see a face it isn't perfectly stereotypical, it is partly occluded by a shadow or turned to the side, seen up close or at a great distance, so having a highly generalised face detection make us actually really well suited to be highly social.
The last experiment looked at how people become familiar with faces. People with autism have a decreased activation of new face memory representations compared to typical controls, suggesting that it takes longer for people with autism to get to know others, which of course is quite consistent with our clinical understanding.
Was there much variation in the ways different people with autism perceived faces?
Yes, and this is a frequent finding in autism research. Typically every graph of every study to do with autism sees the mean as being decreased in autism relative to controls, but the variance associated with that mean in autism is much bigger than the variance for controls. Having now met a large number of people with autism, that is quite consistent with this incredibly heterogeneous condition, even if one is only talking about autistic disorder as currently classified let alone Asperger's Syndrome and the broader autism phenotype. The condition is incredibly diverse, which is one of the reasons why it's so interesting.
As it is such a heterogeneous group, how did you set your recruitment criteria?
I included people who had been diagnosed by one of the clinicians here in the Department of Psychiatry at the University of Cambridge and who, during or after that diagnosis, had agreed to participate in research at the Autism Research Centre. The diagnosis of people with autism is one of the most keenly contested subjects in autism research. For a long time I looked into going through the procedure of diagnosing people with what is known as the gold standard diagnostic schedules - the ADOS (Autism Diagnostic Observation Schedule) and ADI (Autism Diagnostic Interview), but decided against doing so for a couple of reasons. Firstly it would have taken a great deal of time, secondly for me personally, and indeed for other people looking at social ability in autism, perhaps the most important and informative paper is Simon Baron Cohen's great paper Does the Autistic Child Have A 'Theory of Mind'? which did not use the ADOS and ADI. This gold standard has been rather recently construed and is the subject of debate. Of particular interest are the blogs of two great autism researchers - Dorothy Bishop at Oxford and Jon Brock, formerly of Oxford and most recently of Macquarie University in Australia who both write staunchly about the need to dispense with the necessity for using the 'gold standard' diagnostic schedules.
What do you think the implications of your findings are?
I think the first implication is that, unlike the funding body I mentioned earlier, I think that much more work is needed on face perception and autism because, yes we know that people with autism have problems with face perception but the mechanisms by which these problems are created are very poorly understood despite the best efforts of a very large number of people. Most importantly I think that the problems with face perception in autism don't seem to be specific; they cannot be explained as simply a problem with attention or a problem with generalisation. Instead perhaps a myriad of different processes all collude to create this problem with face perception.
Do you think your findings have implications for interventions for people with ASCs?
I think that there need to be more intervention studies, however, these are incredibly costly and time consuming and typically fall outside the remit of a three year PhD student's loan and so tend not to be done.
You mentioned intervention programmes for children with autism, are there ways of helping children with autism learn to read faces?
There are a large number of programmes, some of which have some empirical basis and others which have no scientific basis for their work at all. It's this burgeoning field of pseudo psychology for autism that I think most compels us to do the studies which enable us to say quite clearly what is helpful to people with autism and in what ways and to what degree it is helpful. I think there are strong reasons to want to do intervention programmes for people with autism. The hesitation I have about some intervention programmes comes most of all from a slight sort of awkwardness from a methodological perspective relating to what could be called the potential for the late bloomer. In a lot of cases with early interventions in childhood it is unclear several years after the programme has ended to what degree the intervention programme creates a difference between individuals who have gone through the programme and those who haven't. Children who go through the programme may see an initial gain in social function over peers who haven't gone through the programme, but as the years go by there is some evidence that in some of these cases the non-intervention children simply catch up to peers who went through rigorous and in many cases costly intervention programmes. This is not to say all intervention programmes are bunk. In fact, the first research programme I worked on was an early intervention programme and although we found little evidence for the efficacy of the programme broadly speaking across all domains, I did come away from it thinking that perhaps the parents, as a result of the intervention programme, had gained a great deal and that's actually not something to be dismissed lightly. Parents had not only gained a great knowledge of autism, they had spent time with their children in a way that they never would have felt they could have otherwise and that's potentially something very powerful, something very special for the parents of children with autism.
Are there any specific questions that you think still need to be answered about face perception and autism?
I think there are very, very important questions at every possible level of analysis; questions about the developmental trajectory of these face processing impairments are incredibly important to define, at what stage are they behaviourally evident, at what stages are the other neurological substrates of that impairment evident? Something that has only recently begun to be investigated is what happens in late adulthood. One of the reasons why this is an important question from a practical perspective is that children who were diagnosed with autism, particularly those diagnosed under contemporary criteria from the eighties onwards, are now in mid-adulthood and will soon be entering a stage in life that in the general population can be associated with pathologies that include social disability. In particular some of the dementias, such as fronto-temporal dementia have well documented evidence of theory of mind impairment. I think looking at people who have been diagnosed with a developmental disorder in early childhood and are now in late adulthood will be one of the most important and interesting agendas for cognitive neuroscience in the next decade.
Okay, that's great Owen. Thank you very much.
This interview originally appeared in News & Views February 2Ol2.