Autism and Theory of Mind: a critique of Gernsbacher & Co

(Cross-posted at

In a series of occasional posts that start today, I’ll be going through a half dozen articles co-authored by Morton Gernsbacher, a psycholinguist at the University of Wisconsin. Collectively these articles attempt to present evidence for the redefinition of autism upon which the plausibility of FC depends: namely, the notion that autism is not (despite eight decades of research to the contrary) a socio-cognitive disorder, but rather a motor disorder. More specifically, autism is, purportedly, a disorder in which intentional motor movements, including speaking and pointing, are difficult or impossible to perform.

The notion that autism is a motor disorder is three decades old: it originated with Douglas Biklen, the then-education professor from Syracuse University who exported FC to the US from Australia. But Biklen merely asserted this notion—as do most of his followers and other FC proponents. What’s unusual about Gernsbacher is that she actually attempts to prove it—i.e., to dispute the long-prevailing views of autism and the research that supports them. Indeed, no one has devoted more research and argumentation to combatting the dominant paradigm of autism. Vikram Jaswal doesn’t even come close.

Gernsbacher’s first published foray into the established research is her 2005 paper (Gernsbacher & Frymiare, 2005), Does the Autistic Brain Lack Core Modules? In this article, she and Frymiare dispute both:

(1) The strong claim that there is such a thing as a discrete Theory of Mind brain module that autistic individuals lack


(2) The less problematic claim that autistic individuals have difficulties with Theory of Mind (ToM) tasks.

Gernsbacher & Frymiare consider two types of ToM tasks: reasoning about false beliefs (what might be called cognitive perspective taking), and processing people’s faces.

The classic false-belief test is the Sally-Anne test. The child observes two puppets, Sally and Anne. Anne watches as Sally puts her marble in the basket. Sally leaves the room. Anne takes the marble out of the basket and puts it in the box. Sally comes back. Where will Sally look for her marble? To deduce the correct answer, that Sally will look first in the basket, the child must shift from her own perspective (which includes the knowledge that the marble is now in the box) to Sally’s perspective (which, lacking such knowledge, includes a “false belief” about where the marble is).

Gernsbacher & Frymiare’s case against an autism-specific deficit in false-belief reasoning is based on two non-controversial observations:

1. The standard versions of false-belief tests require sophisticated linguistic comprehension


2. All populations, with and without autism, fail these tests until they acquire a certain level of language.

As Gernsbacher & Frymiare point out in reference to the false-contents version of the false-belief test (in which, say, a candy box contains a pencil):

[P]erformance on… [the] participant’s ability to answer two critical questions: “What did you think was inside the box before I opened it?” and ‘What do you think [another person] will think is inside the box before I open it?”. (p. 5)

As they go on to say:

The syntactic form of these two questions is one of the most complex in the English language. These sentences exhibit sentential complement constructions, in which a complement clause [“is inside the box”] is embedded in the matrix clause [“What do you think another person will think…”]. Indeed, all mentalizing statements require sentence complement constructions, which are some of the most complex syntactic structures in the English language. (p. 6)

And, indeed, several studies (for example, Hale & Tager-Flusberg, 2003) show that mastering sentence complement structures is the biggest causal contributor to mastering false belief tasks.

Accordingly, autistic individuals, as Gernsbacher & Frymiare note, are able to pass these tests once they pass a certain linguistic threshold. As specific support, they cite Tager-Flusberg Sullivan (1994), who, they say

have demonstrated that when autistic children are compared with non-autistic children who are matched to the autistic children’s language skills, the difference between autistic and non-autistic children in their success rate of passing first and second-order false belief tasks disappears. (p. 6)

This characterization of Tager-Flusberg & Sullivan, however, is slightly misleading, masking a crucial distinction between the autistic individuals who pass ToM tests and their non-autistic counterparts. And it is at this point that Gernsbacher & Frymiare’s case starts falling apart.

As it turns out, it’s not enough to simply match autistic and non-autistic on their language skills to capture what’s going on. In Tager-Flusberg & Sullivan’s study, the autistic and non-autistic subjects who were matched for language skills had an average verbal mental age of 9 years, 8 months. This age is considerably above the 5-year verbal mental age threshold for when most non-autistic children pass Theory of Mind tests. But it is only slightly above the verbal mental age threshold for when most autistic children pass ToM tests.

Indeed, two of the key findings about the links between language and ToM tests—which Gernsbacher & Frymiare notably do not cite—is that autistic individuals who pass such tests, compared with their non-autistic counterparts:

1. Must reach a significantly higher verbal mental age

2. Appear to be especially dependent on language

In her pooling of data from a variety of ToM studies, Happe (1995) finds that a majority of autistic participants do not pass ToM tests until they reach a verbal mental age of 9 years, 2 months. As Tager-Flusberg & Joseph (2005), subsequently report, there is an additional difference: autistic individuals

3. Appear to be especially dependent on sentential complements that use verbs of communication (e.g., “Sally said that the marble is in the basket”) as opposed to verbs of belief (e.g., “Sally thinks that the marble is in the basket.”).

As Tager-Flusberg & Joseph (2005) propose, autistic individuals who pass ToM tests appear to be doing so using different, more linguistically based mechanisms than their non-autistic counterparts. These individuals

use language to reason logically through false-belief tasks, or to interpret what others know or believe on the basis of their experience with specific events. Instead of depending on a conceptual understanding of mental states that is grounded in the social-perceptual component of theory of mind, in autism, they rely on language as the sole route to understanding propositional attitudes; there is no independent language-of-thought in the domain of theory of mind. (pp. 311-312)

What all this suggests it that language delays aren’t the full story where autism-related difficulties with false-belief tests are concerned. Language aside, there appears in autism to be a deficit in automatic, intuitive perspective-taking.

Gernsbacher & Frymiare claim, however, that in experiments where the linguistic aspect of the testing is eliminated, children with autism and deafness outperform children with normal hearing. As they put it:

if one creates a false drawing task that tests theory of mind without reliance on language, one finds that children with autism and children with deafness actually outperform children with normal hearing. (pp. 6-7)

Their citation is a study done by Peterson (2002).

But a closer look at Peterson does not support this claim. First of all, the children with normal hearing in Peterson’s study were 4 years old while the deaf and autistic subjects were between ages 6 and 13. Second, Peterson’s tests, including the one in which the former outperformed that latter, in fact did rely on language. Students were shown an apple of one color (either red or green), given a marker of a different color (either red or green), and asked to draw a picture of what they saw. A second apple was added to the scene, and then the child was told “X is coming next. He/she hasn’t done any drawings for me yet. I am going to show him/her this picture of yours. I am going to ask him/her to guess which of these apples you were drawing. What will he/she say?” While this verbal prompt doesn’t contain any embedded complements, it nonetheless contains a fair amount of sophisticated language, including other sorts of embedded clauses [“you were drawing”; “which of these apples”]. As to why it was that the autistic and deaf 6-13 year olds outperformed the non-deaf, non-autistic 4 year olds, Peterson does not have a ready explanation—instead offering several hypotheses, none of them noting the possible effects of the age difference.

Nor do those tests that truly eliminate the linguistic factor in false-belief tests support Gernsbacher & Frymiare’s claim. In a study involving individuals with Asperger’s, Senju et al. (2009) examined false-belief understanding by using measuring (via an eye-tracker) where participants looked in anticipation of where someone with a false belief about an object’s location would look. Does that person, for example, look towards the basket when Sally returns, correctly anticipating that that’s where Sally will look? Senju et al.’s “showed that, like infants, neurotypical adults’ … eye movements anticipated an actor’s behavior on the basis of her false belief.” However, “[t]his was not the case for individuals with Asperger syndrome…” Their conclusion: individuals with Asperger’s syndrome “do not attribute mental states spontaneously, but they may be able to do so in explicit tasks through compensatory learning”—i.e., through language. (Emphasis mine).

Less definitive, but still suggestive, is Schuwerk et. al (2016)’s similar test with 7-8 year olds. Schuwerk et. al found a big difference between autistic and non-autistic children, with the former looking longer in the wrong location. (The catch was that, while the non-autistic children looked longer at the correct location, this difference was not significantly above chance).

What all this shows is that, even when you eliminate language, individuals with autism are distinguished by deficits in automatic, intuitive perspective taking. When they succeed in false-belief tests, they appear to depend on deliberate reasoning via language.

In the second half of their paper, Gernsbacher & Frymiare address the claim that a “facial expression processing module” is impaired in autism. Here, they correctly point out that many autistics find eye contact aversive and therefore tend not to look at faces as much as non-autistics do. They propose, furthermore, that it is only for this reason that autistic individuals don’t activate regions of the brain involved in face processing as much as non-autistic individuals do. So far, so good: none of this is controversial.

But Gernsbacher & Frymiare go further, claiming the aversion to eye contact is not a social issue; it’s just that, for this population, eye contact is uncomfortable or even painful. But Gernsbacher & Frymiare don’t explain why this would be—and, since eyes are highly social stimuli, it is hard to rule out that the underlying issue isn’t in some sense social. Furthermore, whether or not it’s social, its consequences surely are. Even if autistic individuals aren’t born with face processing difficulties—and, in particular, with their well-documented difficulties reading facial expressions—years of not looking at faces will lead to:

1. Diminished opportunities to learn/maintain skills in interpreting facial expressions


2. Diminished opportunities for social connection

The authors conclude by claiming that, for ToM skills to be considered a core deficit of autism, everyone on the autism spectrum would have to lack them. But, as we’ve seen, some deficits can be overcome, or at least bypassed, by compensatory strategies like language.

That fact, however, doesn’t rule them out as defining traits of autism.


Gernsbacher, M. A., & Frymiare, J. L. (2005). Does the Autistic Brain Lack Core Modules?The journal of developmental and learning disorders9, 3–16.

Hale, C. M., & Tager-Flusberg, H. (2003). The influence of language on theory of mind: a training study. Developmental science, 6(3), 346–359.

Happé F. G. (1995). The role of age and verbal ability in the theory of mind task performance of subjects with autismChild development66(3), 843–855).

Peterson C. C. (2002). Drawing insight from pictures: The development of concepts of false drawing and false belief in children with deafness, normal hearing, and autism. Child Development, 73, 1442–1459.

Schuwerk, T., Jarvers, I., Vuori, M., & Sodian, B. (2016). Implicit Mentalizing Persists beyond Early Childhood and Is Profoundly Impaired in Children with Autism Spectrum ConditionFrontiers in psychology7, 1696.

Senju, A., Southgate, V., White, S., & Frith, U. (2009). Mindblind eyes: an absence of spontaneous theory of mind in Asperger syndrome. Science (New York, N.Y.)325(5942), 883–885.

Tager-Flusberg H. & Joseph, R. M. (2005). How language facilitates the acquisition of false-belief understanding in children with autism. In J. Astington, J. W., & J. Baird (Eds.), Why language matters for Theory of Mind (pp. 298-318). Oxford University Press.

Tager-Flusberg, H., & Sullivan, K. (1994). A second look at second-order belief attribution in autism. Journal of Autism and Developmental Disorders, 24(5), 577–586.

3 thoughts on “Autism and Theory of Mind: a critique of Gernsbacher & Co

  1. What about children who put themselves in Sally’s place and reason thusly after returning to the room?

    “They sent me out of the room for a few minutes and upon my return asked me where the marble is. That indicates to me that this is some sort of test. Perhaps they have moved the marble. [Looking around the room] I’m going to say it’s NOT in the basket, but in some other place.”

    And then you have the ones that get that far and continue that line of thinking:

    “But that’s what they’re expecting me to say, so maybe this is all a trick and they didn’t move the marble after all. I have no idea where the marble is, and the only place I KNOW it has been is in the basket. Maybe I should say ‘basket.'”

    And then the child, with zero confidence, gives some sort of answer, basket or not-basket, but regardless he is certain he will be wrong.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s