Do autistic people have a deficit in reading faces?

Earlier this week in the post Neurological "Personhood," I made a comment about individuals with autism. My comment was as follows:

1) Some individuals do not show normal development in the system of identifying personhood described. For example, individuals with autism sometimes show deficits in this area. What does the fact that this system is not universal say about ethical behavior? Clearly, many autistic people are still getting there, but they must be getting to ethical behavior by some other route.

I realized later on that I was not being particularly clear about where I was going with that comment. The purpose of my comment was to question the core argument of the posted article -- that ethical behavior requires the neurological system for detecting personhood -- because in some individuals who possess detectable defects in this system -- people with autism -- there is no problem developing ethical behavior.

I realized the confusion after reading a post on Autism Vox to which I responded the following:

My question from reading this is: What might be this "other route" by which it is suggested autistic people are "still getting" to personhood---what are "other routes" to ethical behavior? And do those statements I just typed open up some misleading questions about autistic persons and ethics?

I don't think I was being very precise in what I meant in that post, so let me clarify.

We know from imaging experiments that some individuals with autism do not activation in the area of the brain associated with faces. This part of the brain is called the fusiform face gyrus. We know that faces -- as a special class of objects -- are represented different in the human brain than other things.

This observation in individuals with autism led some researchers to speculate that it might explain some of the social difficulties that they have. If you have trouble recognizing faces as a special class of objects -- that faces are fundamentally special -- it is very difficult to interact with others in the way that they expect.

Now the article I cited was arguing that the system of recognizing personhood -- of which the fusiform face gyrus is a part -- is critical to ethical behavior. In essence they are saying that representing persons as special in the brain and being ethical towards them are inextricably linked.

But that argument does not quite work for me. I know people with autism. I have never found them to be more or less ethical than everyone else, and if the argument presented in the article is correct we wouldn't expect that to be the case.

The point of my comment was to highlight a consequence of their argument. If we assume that ethical behavior is contingent on some brain system and autistic people show deficits in this system, then autistic people cannot be ethical in the same way as other people are.

As I think you and I would agree that this statement is patently false, then the argument in the article must be somehow fundamentally flawed.

I'm sorry I was so roundabout in my post, but I hope this clarifies where I was going with that.

I just wanted to clear that up because re-reading my earlier post I don't think that I made it clear that I wasn't questioning austistic people's ability to be ethical.

Anyway, I digress.

At the end of that post, there was another commenter who added that one of my core premises -- that people with autism have defects in face processing -- might also be incorrect. Here is Michelle Dawson commenting on that post:

Some fMRI face studies in which autistics display typical fusiform gyrus or fusiform face area activation: Pierce et al. (2004), Hadjikhani et al. (2004), Piggot et al. (2004), Wang et al. (2004), Dapretto et al. (2006), Bird et al. (2006). And if you count conference posters, you have to add Pelphrey et al. (2005). And there was an IMFAR presentation (no time to look it up) last year which found typical FG activation in autistics performing a 1-back task using face images.

The earlier fMRI face studies (most famously, Schultz et al., 2000) had important methodological issues. I owe my involvement in autism research at least in part to successfully predicting that controlling for attention in autistics in fMRI face studies would produce typical FG activation.

For a comprehensive (rather than selective) look at the autism face studies (behavioural, neurophysiological, and functional imaging studies), see Jemel et al. (2006).

Most of what is popularly said to be "known" about how autistics look at and process face images is based on cliché rather than on the published data.

The real problem is when some non-autistics deny the personhood and/or humanity of autistics. That is, non-autistics look at autistics and fail to detect that there is a person there, or anything human. This common phenomenon can be seen, among many other places, in the writing of a lot of famous scientists, including Thomas Insel, Ivar Lovaas, Michael Tomasello, Steven Pinker, V.S. Ramachandran, etc. (sorry, again, no time to go fetch all the sources, but they're out there...). (Emphasis mine.)

(Michelle blogs at Autism Crisis.)

This got me thinking. I mean I have been told in classes that this defect exists, but I don't think that I have ever checked what the research says. So I went and looked it up, and low and behold the situation is much more complicated than I knew.

First of all, it is not so much that people with autism have trouble detecting faces, but they do tend to examine them differently. In general, the evidence shows that people with autism tend to examine the parts of faces more than the whole face. However, this preference for parts had led some researchers to speculate that there was a deficit in so-called "holistic processing" of faces -- and this speculation has not been borne out by additional evidence.

Most of the original belief in the deficit was due to the results with what is called the Thatcher illusion. In the Thatcher illusion, two picture of Margaret Thatcher's face are inverted. One has (in the inverted form) the mouth and eyes turned right-side-up. The faces are then rotated such that now one of the faces is correct and the other is right-side-up with the mouth and eyes now inverted. The one with the inversions looks super weird, but it didn't look quite as weird when it was upside down. (You can see the Thatcher illusion here.)

This discrepancy -- the feeling of wrongness associated with the altered right-side-up face -- is due to a preference in the human visual system for correctly oriented faces. Originally, it was believed that people with autism lacked this discrepancy -- the difference in the wrong feeling between the inverted Thatcher face and the correctly oriented with upside down mouth and eyes Thatcher face -- suggesting a general deficit in face processing.

However, it is a bit more complicated. The review I read, Jemel et al., clarified the issue as follows:

The finding that individuals with ASD tend to process faces in terms of their component parts should not necessarily imply that configural/holistic processing strategies are impaired. This hypothesis Face Processing in Autism 95 has been addressed either directly or indirectly through the face inversion paradigm. Some studies concluded that individuals with ASD are less disturbed by face inversion compared to typically developing individuals. However, the actual results appear to be less conclusive. For example, Hobson et al. (1988) found that adolescents with ASD were superior to controls in both expression and identity sorting tasks when photographs were presented upside down, though the difference between the two groups of subjects did not reach significance. In addition, it was found that the performance scores of both the autistic and non-autistic groups was lower in the upside-down condition than in the upright condition, which clearly indicates that face inversion affected autistic individuals' performances. In Tantam et al. (1989), the absence of face inversion effect in children with ASD is obscured by an overall floor effect. In fact, while the control group's performance at labeling the face expressions was affected by turning the photos upside-down, the autistic group's performance in the inverted condition was as limited as in the upright condition. More recent studies did find a typical decline of performance due to face inversion in individuals with ASD (Joseph & Tanaka, 2003; Lahaie et al., 2006; Teunisse & de Gelder, 2003) as well as a Thatcher illusion effect (Rouse et al., 2004). Although the superior processing of local aspects of faces was previously thought to derive from a deficit in the perception of global and configural information, these latter findings suggest an enhanced low-level perceptual processing with absence of global perceptual impairments. (Emphasis mine.)

The second issue in my comment was about the Fusiform Face Area (FFA), an area in the brain dedicated to the detection of faces. The research that I was aware of suggested that individuals with autism show lower activation in this area in response to faces. However, this too is a more complicated issue.

The data on imaging and the FFA is a little to go over here, but essentially the problem is that results have been conflicting. It is unclear why this would so, but what is clear is that because it is so FFA hypoactivation is not an adequate marker for autism. (Again I would recommend reading Jemel et al because they go over this evidence in detail.)

What can we take away from the data related to autism and face processing?

1) It is dangerous to bandy about data that you aren't completely on top of. I have learned my lesson on that one.

2) The data suggest that autistics may process faces differently, but this is not suggestive of a global deficit -- and the deficit is therefore not an adequate explanation for problems with socialization. Scientists who have made a cottage industry toting this deficit should clarify that the results are much more mixed then their generalizations would suggest.

3) As we like to say at Pure Pedantry, reality is always more complicated than you think.

4) I am beginning to understand how charged the issue of autism can be at times. I get hostile comments from time to time, but I always thought it was strange how many of them were related to autism research. (Orac has a great post on mercury research and autism that goes a long way in explaining why.) Anyway, in the future I will try and be a bit more tactful.


More like this

Everybody knows the Margaret Thatcher Illusion. If you've forgotten about it, here's the best example I've found (from Schwaninger et al.1) Both the top and bottom pairs are the same photos, but they look very different depending on whether they're upright or inverted. In the top pair (the…
I was reminded of this illusion by the Seed Daily Zeitgeist yesterday. In order to get the full effect, I'll show you one set of photos here, and the rest of the post will be below the fold. The first are from Schwanginer et al. (2003)1: They look pretty normal, right? Now look at these: Gross,…
Do you recognize the faces in this picture? Sure you do -- you could recognize the authors of this blog anywhere, even upside-down. It might take you just a bit longer to realize that something isn't quite right with the picture. I'll show you what the problem is at the end of this post. We've…
Can you tell if these two faces are the same or different? How about these? Or these? If you're like most adults, it will be easier for you to identify different adult faces compared to the infant faces in the second example, and even the small children's faces in the third example. This…

Being autistic myself, I would like to say that I can read faces passably. The thing is, it takes a consious effort to look at a persons face. Its not that faces are as interesting as everything else, I find faces LESS interesting than everything else.

"Most of the original belief in the deficit was due to the results with what is called the Thatcher illusion."

This original belief came in part from early findings that some autistics fail to show a typical face inversion effect (not quite the same as a Thatcher illusion effect). In some conditions autistics were found to be better than non-autistics at recognizing the identity or facial emotion of inverted face images. This finding has not been consistently replicated. Researchers didn't get around to testing autistics on the Thatcher illusion until recently.

The first and only Thatcher illusion study in autism is Rouse et al. (2004). The authors report that autistic children aren't any less susceptible to the Thatcher illusion than non-autistic children. This is consistent with findings that autistics can make typical use of configural face information (Lahaie et al., 2006).

Otherwise, thanks for posting this! It's good to have someone else pointing out that in the area of autism, you have to slog through the data.

Thanks for linking to the site with the illusions [always love optical illusions] I plan to show them to the children when they come home from school - see what happens!

No problem, Michelle.

I really thought that research was much more clear cut, but it looks like in the last couple years the conclusions have become much grayer.

I'm an Asperger Geek, and I can add a couple of thoughts from my perspective; ethical behaviour can be ingrained by a sense of empathy for other people OR by the use of intelligence to comprehend the "Golden Rule". While autistics intrinsically ignore much of empathy (both Kanner and Asperger independently came up with the same Greek construction "autism" to describe people wrapped up in themselves, or the ultimate selfishness,)once they mature enough to embrace the logic of the Golden Rule, it becomes a satisfying constant in a world they have difficulty comprehending on an emotional level. Another interesting anecdote I can share about myself; I have trouble remembering faces, yet as I look out over a sea of office cubicles, which permit me to only see the very tops of heads bobbing by, I can identify people with near 100% certainty seeing nothing more than their hair or tonsure.

Thanks so much for posting this more extensive analysis and discussion, which is leading me to reflect on how my son has responded to various faces over the year. I recognize that this is very anecdotal, but he has always shown a marked interest in faces that are (as mine is) Asian. I do think (for my son) that there are certain details that stand out to him. Further, he does not look straight at objects especially when they are close to him (he is somewhat far-sighted).

"[T]he evidence shows that people with autism tend to examine the parts of faces more than the whole face": If I may extend this point to the way my son seems to process visual information generally, he is constantly seeing the part for the whole and (as it seems to me) scanning a visual field and then pausing over certain parts.

Thank you again and I certainly did not mean to imply anything "hostile"---I was very curious to your further thoughts and your post helps much.

I reacted to your initial post on this interesting point, that different processing individuals shows that some claims on moral behavior isn't as clear cut as some persons imply. (Which btw animals would show too, come to think of it.) And I didn't get impression that you meant that autistic persons had an impairment here, but the converse.

That said, I am glad that you cleared up any confusion here and on different observations.

The finding that individuals with ASD tend to process faces in terms of their component parts should not necessarily imply that configural/holistic processing strategies are impaired.

Btw, this reminds me of the post (at Cognitive Daily?) on artists. IIRC artists can be observed by monitoring eye movements to process scenes much more facilely and completely than non-artists, not concentrating as much on details and scanning a larger area. Still, non-artists can sometimes with a bit of effort make rather nice attempts at paintings.

So such differences can be a matter of familiarity and usage instead of an indication of a source to a problem.

By Torbjörn Larsson (not verified) on 01 May 2007 #permalink

I believe part of the problem here is that people tend to mistake empathy with goodness and sympathy, and not think of it as more of a social survival skill. Psychopaths may be better at reading faces and clawing up the corporate ladder, but you can hardly call them humanitarians.