Is there really no Theory of Mind deficit in autism? Part V: Do Theory of Mind tests fail to predict understanding of goals and desires?

Cross-posted at 

In my last four posts, I critiqued the arguments in Gernsbacher and Yergeau’s 2019 article that ToM (Theory of Mind) tests lack empirical validity—in particular, that the original test results with autistic subjects have failed to be replicated, and that the tests themselves fail to converge on a meaningful psychological construct and fail to predict autism-related traits and empathy and emotional understanding.

Morton A. Gernsbacher, Professor of Psychology, University of Wisconsin

Gernsbacher and Yergeau’s final line of argument concerns the question of whether the ToM tests, as is generally claimed, tap into the ability of autistic people to infer other people’s goals and desires. Here they return to the argument made in Gernsbacher et al. (2008), and claim that “autistic people of all ages skillfully understand other persons’ intentions, goals, and desires.” In support of this claim, they cite several dozen studies. The problem here is that, as I discussed earlier, the kinds of intentions, goals, and desires are all basic, instrumental level goals, intentions, and desires—the kind represented by instrumental physical activities like reaching, pulling apart, and inserting. Thus, we are not talking about social goals and intentions like making a good impression, or complex psychological desires like romantic interest.

And, indeed, what some of the articles (e.g., Aldridge, 2008; Colombi et al., 2009; Carpenter et al., 2001; Falck-Ytter, 2010; Berger and Ingersoll, 2014; Fitzpatrick et al., 2013) show is that autistic individuals are no more impaired in their ability to detect intentions, goals, and desires vis-à-vis concrete, visible, physical tasks—reaching, pulling apart, inserting— than other people are, even when the actors performing these activities fail to achieve them.

On occasion, as Liebal et al. (2008) have found, the ability of individuals with autism to understand intentions vis-à-vis concrete tasks extends to an ability and willingness, in some cases, to help a person fulfill those intentions: whether in helping the person reach something, pick up something the person dropped accidentally, or cooperate with the person in taking apart a container.  However, compared to a group of developmentally delayed children, Liebal et al. found that fewer children with autism provided help and, when the person ceased performing a task, they made fewer attempts to reengage them and were poorer in their attempts to coordinate eye gaze with them. This suggests that, while the ability to infer these sorts of physical intentions may be intact in autism, the ability to share intentions is impaired. Colombi et al. (2009) agree, noting the difficulty that autistic individuals have with imitation and joint attention.

Berger and Ingersoll (2014) found that autistic children were able to use social-communicative indicators like facial expressions to infer whether an act (say, dropping an object) was deliberate or accidental. However, they also found that autistic children made these distinctions less often than typically developing children do—because they attended less frequently to faces. Vivante et al. (2011), meanwhile, found that high-functioning autistic participants could make inferences about intention based on people’s facial expressions, but not based on their head movements and gaze shifts. (We should note here that the facial expressions involved in this study are less challenging than the more subtle, eye-area facial expressions assessed by the Eyes test, which is where individuals with autism show difficulty). Finally, Fitzpatrick et al. (2013) found that autistic children, while able to infer and imitate intentional tasks, have more difficulty with tasks that involve synchronizing their movements with others.

As seen, much of the research on the ability of autistic children to infer intentions has focused on instrumental tasks. And as Berger and Ingersoll (2014) note, various researchers have pointed out that the sorts of tasks involved in these studies—reaching, pulling apart, inserting—“may not provide a true measure of the ability to infer the intentions of others.” Perhaps autistic children would perform equally well when watching robots perform the same actions. Or perhaps, as Vivanti et al. (2011) suggest, their inferences about intentions are based, in part, on the characteristics and/or standard use of the objects being acted upon (as when someone centers an object with a hole in it around a dowel). In support of this hypothesis, Vivanti et al. cite a study by Boria et al. (2009) that found that children with autism had difficulty inferring intent when the actor’s intentions did not match the conventional use of the object.

Colombi et al. (2009), furthermore, argue that understanding intentions may not be the same as sharing intentions and that it is sharing intentions (i.e., cooperating) that lies at the heart of the social deficits of autism. And Fitzpatrick et al. (2013) found that ToM tasks and intentionality tasks correlate with distinct psychological factors, which suggested that these tasks “may not be measuring the same underlying construct.” Fitzpatrick et al. propose, in particular, that sharing intentions correlates with lower-level perceptual and attentional dimensions of social competence (“social attention”), while ToM tasks correlate with higher-level cognitive socio-cognitive factors (“social knowledge”).

Some of the research cited by Gernsbacher and Yergeau, however, goes beyond findings about the ability of autistic participants to infer intentions. Kerr and Durkin (2004), for example, propose that eliminating the language confounds in explicit ToM tests eliminates the performance difference between verbal-age-matched autistic and non-autistic subjects. Notably, the autistic participants in Kerr and Durkin’s study were far below the higher verbal mental age that other studies have shown to be a prerequisite for passing the false-belief ToM tests in autism. The problem with Kerr and Durkin’s study, however, is that it’s not clear that the various measurements truly probed the ability to make inferences relating to beliefs. Children were taught that a thought bubble “shows what X is thinking”, such that, when asked which object a character was thinking about, they named the object depicted within the thought bubble rather than an object in the scene outside the thought bubble. But it’s unclear whether all the participants truly understand the bubble’s contents as a thought, as opposed to simply learning to look in the bubble when they hear the word “thinking.”

This calls into question Kerr and Durkin’s main finding. In their experiment, the participants were shown scenes in which the contents of a box were invisible, an object next to the box was visible, and a character with a picture of a different object in the thought bubble over his/her head was shown looking into the box. When participants were asked “What is in the box?”, both autistic and non-autistic participants named the object in the thought bubble, not the object next to the box. Kerr and Durkin take this as an indicator of ToM: an ability to attribute beliefs to other people. They do not rule out the alternative, non-ToM-based strategy of process of elimination: one object is clearly next to the box and not in it. As for the object inside the thought bubble, it is harder to see it as having a definite spatial relationship with the box (i.e., as being above the box): the graphic outlines of the thought bubble separate it from the scene, and if it were part of the scene, it would have to be somehow suspended in the air. These two factors make the object in the thought bubble a much better candidate for being the object in the box than the object next to the box is.

McAleer et al. (2011)’s study is similarly unconvincing. High-functioning adults and non-autistic controls were shown videos of actors enacting scenes of chasing, fighting, following, guarding, playing, and flirting. Both groups did equally well in identifying which interaction was being depicted. But these depictions involved only bodily movements, not facial expressions or gestural or verbal communication. Playing, for example, was enacted as a game of tag. Flirting was enacted by one actor moving to another actor, circling that person twice and moving back to their original position, then the other actor doing the same thing, and then, finally, the two walking off together. The authors propose that correctly identifying the flirting scenes involves ToM processing. But, as in Kerr and Durkin (2004), such identification could have been the product of non-ToM reasoning—for example, process of elimination. Given the other options, the double circling and then walking off together is not the best candidate for chasing, fighting, following, guarding, or playing. This means that even participants who were unfamiliar with the concept of flirtation may still have been able to infer which label was most likely.

Li et al. (2019) and Channon et al. (2011), meanwhile, looked at sensitivity to deliberate vs. accidental harm in making moral judgments. They find that autistics, like typically developing children, tend to judge deliberate acts more harshly. But the only ToM reasoning involved in this study is that of distinguishing intentional acts from unintentional ones—and this is not considered one of the core ToM impairments in autism. Channon et al. (2011) also found that the judgments of the autistic subjects were influenced more strongly by what was subjectively foreseeable by the perpetrator. The authors claim that calculating this subjective foreseeability requires an appreciation of the perpetrator’s internal beliefs. But, in fact, the participants are told explicitly what the perpetrators subjectively foresaw (e.g., “Nick himself thinks the virus is aggressive and is likely to cause serious damage.”). Furthermore, as the authors themselves note, autistic individuals may to some extent simply be applying “learned social rule”, and that

[o]ver-reliance on the rigid implementation of learned rules rather than more mentalistic or intuitive interpretations based on subtle nuances of the context could account for the more extreme blame judgments in those with AS in the present study.

Gernsbacher and Yergeau also cite Green et al. (2017), who examined the ability of autistic participants, compared with IQ-matched controls, to view a picture of a social interaction and then choose which of four pictures appearing below it shows the “same kind of interaction.” While Green et al. found no performance differences between the two groups, nor any correlations with autism trait severity, they note that they did not measure reaction time. They do not rule out, therefore, the possibility that the two groups used different processes, with the non-autistic groups making more intuitive judgments. Indeed, Green et al. propose that, while autistic individuals are “capable of using social analogic reasoning ‘effortfully’, it does not come spontaneously.”

Russell and Hill (2001)’s paper focuses on the ability to mentalize about oneself—which, again, is not the sort of ToM deficit generally attributed to autism. Looking at autistic children and IQ-matched controls whose mental ages ranged from 4 to 10 years, they found that both groups were equally able to distinguish outcomes that they desired or achieved from outcomes that they had originally intended. The two groups were also equally successful in a version of the Smarties experiment in which participants were only asked about their own beliefs (“Before I opened the lid of the box, what did you think was in the box?”). At the same time, the autistic group, as expected, performed worse than the comparison group on a Sally-Ann-like false-belief test. All this is consistent with current views of autism as involving perspective-taking deficits (along with other social deficits), as opposed to deficits in self-awareness.

Four additional studies only looked at high-functioning autism (HFA). Vivanti et al. (2011) found that children and adolescents with HFA increased their attention to a person’s face when he/she performed an action in an “irrational” way, as if seeking information about motivation. While this suggests some intact ToM skills, these may be specific to HFA.

Cole et al. (2018) found that HFA adults, as compared with typical controls, displayed typical eye fixation patterns when observing deliberate vs. intentional actions. On the other hand, they also report that the autistic adults had difficulty reporting on the actors’ intentions in the different scenarios. Their conclusion: “Adults with ASD displayed explicit but not implicit mentalizing deficits.”

Sebanz et al. (2005) focus on action representation (the representation of other people’s actions) as opposed to ToM (e.g., the mental representations of other people’s beliefs). Their participants were high functioning enough that they were able to solve first and second-order ToM tasks. Sebanz et al.’s findings—that these autistic participants behaved similarly to non-autistic controls depending on whether a co-participant’s prompt in a go/no-go task coincided with their own go/no-go prompt—do not relate to ToM skills in autism.

Finally, Ponnet et al. (2005)’s study found that high-functioning adults with PDD (a mild form of autism akin to Asperger’s Syndrome) were as able as their non-autistic counterparts to infer the thoughts and feelings of their interaction partner during an 8-minute first-time conversation. However, they also acknowledge that the PDD participants may, over the years, have learned scripts for “an initial conversation of the get-acquainted type,” and that such conversations may not present the same kinds of inferencing challenges as longer and non-initial conversations do. Despite this—but in line with the oft-observed deficit in social initiation seen in autism—Ponnet et al. found that two thirds of the conversations were initiated by the control (non-autistic) adult. Also in line with observations about individuals with HFA, they found that the PDD participants spoke for longer periods of time but made less eye contact, and that fewer of the thoughts and feelings they reported focused on their interaction partner. Finally, Ponnet et al. caution that their PDD participants had IQs far above normal and had been recruited “on the basis of their good performance on previous mind-reading tasks”, reminding us that:

Many researchers have stressed the role of both verbal and chronological age in the performance on theory of mind tasks and have suggested that intelligence might compensate for conceptual perspective taking strategies.

Many of these studies, thus, provide, at best, circumscribed support for Gernsbacher and Yergeau’s contention that “autistic people of all ages skillfully understand other persons’ intentions, goals, and desires.”

But one of the last experiments cited by Gernsbacher and Yergeau would appear to undermine their case as much as it supports it. Hubert et al. (2007) showed participants point-light displays of actions (lifting, hopping, pushing, clapping…), subjective states (itchy, bored, tired, cold), and basic emotional states (surprised, sad, frightened, angry, happy). They found that the autistic participants were just as able as the non-autistic controls to integrate the point-light displays into a holistic image and to identify the actions and subjective states. On the other hand, “they were significantly poorer at labeling emotional displays, suggesting that they are specifically impaired in attending to emotional states”:

For instance, participants with autism would describe the point-light display depicting afraid as someone who walks forwards and then goes back or a person who walks sideways, or the point-light display depicting anger as someone who jumps.

In short, the research on the ability of autistic individuals to infer other people’s goals and desires shows that this largely depends on whether the goals are instrumental or social in nature:

  • Individuals with autism are not generally impaired in their ability to detect instrumental goals and desires (reaching, pulling apart, and inserting), at least when objects are being used in conventional ways.
  • Some individuals with autism provide assistance to people engaged in instrumental tasks. But relatively fewer autistic individuals offer such assistance, and the quality of their assistance shows subtle social impairments.
  • Individuals with autism are able to use social-communicative indicators like facial expressions to infer whether an act (say, dropping an object) is deliberate or accidental, but show subtle social impairments in this as well.
  • HFA (high-functioning autistic) individuals are able to make moral judgments about other people’s actions, but this may be based on a “rigid implementation of learned rules” rather than more intuitive, psychologically nuanced reasoning.
  • ·Autistic individuals are better able to reflect on their own mental states than on those of others
  • In HFA, social reasoning skills may be intact, but they appear to be less spontaneous than in their non-autistic counterparts.
  • Even in the mildest forms of autism, like PDD, conversational interaction is still impaired.


Aldridge, M. A., Stone, K. R., Sweeney, M. H., & Bower, T. G. R. (2000). Preverbal children with autism understand the intentions of others. Developmental Science, 3, 294–301.

Berger, N. I., & Ingersoll, B. (2014). A further investigation of goal-directed intention understanding in young children with autism spectrum disorders. Journal of Autism and Developmental Disorders, 44, 3204–3214.

Boria, S., Fabbri-Destro, M., Cattaneo, L., Sparaci, L., Sinigaglia, C., Santelli, E., Cossu, G., & Rizzolatti, G. (2009). Intention understanding in autism. PloS one4(5), e5596.

Carpenter, M., Pennington, B. F., & Rogers, S. J. (2001). Understanding of others’ intentions in children with autism. Journal of Autism and Developmental Disorders, 31, 589–599.

Channon, S., Lagnado, D., Fitzpatrick, S., Drury, H., & Taylor I. (2011). Judgments of cause and blame: Sensitivity to intentionality in Asperger’s syndrome. Journal of Autism and Developmental Disorders, 41, 1534–1542.

Cole, E. J., Slocombe, K. E., & Barraclough, N. E. (2018). Abilities to explicitly and implicitly infer intentions from actions in adults with autism spectrum disorder. Journal of Autism and Developmental Disorders, 48, 1712–1726.

Colombi, C., Liebal, K., Tomasello, M., Young, G., Warneken, F., & Rogers, S. J. (2009). Examining correlates of cooperation in autism: Imitation, joint attention, and understanding intentions. Autism, 13, 143–163.

Falck-Ytter, T. (2010). Young children with autism spectrum disorder use predictive eye movements in action observation. Biology Letters, 6, 375–378.

Fitzpatrick, P., Diorio, R., Richardson, M. J., & Schmidt, R. C. (2013). Dynamical methods for evaluating the time-dependent unfolding of social coordination in children with autism. Frontiers in Integrative Neuroscience, 7, Article 21.

Gernsbacher, Sauer, Geye, Schweigert, and Goldsmith (2008), Why does joint attention look atypical in autism. Child Development Perspectives 2(1), 38-45. doi/full/10.1111/j.1750-8606.2008.00039.x

Gernsbacher, M. A., & Yergeau, M. (2019). Empirical Failures of the Claim That Autistic People Lack a Theory of Mind. Archives of scientific psychology7(1), 102–118.

Green, A. E., Kenworthy, L., Gallagher, N. M., Antezana, L., Mosner, M. G., Krieg, S., . . . Yerys, B. E. (2017). Social analogical reasoning in school-aged children with autism spectrum disorder and typically developing peers. Autism, 21, 403–411.

Hubert, B., Wicker, B., Moore, D. G., Monfardini, E., Duverger, H., Da Fonseca, D., & Deruelle, C. (2007). Recognition of emotional and non-emotional biological motion in individuals with autistic spectrum disorders. Journal of Autism and Developmental Disorders, 37, 1386–1392.

Kerr, S., & Durkin, K. (2004). Understanding of thought bubbles as mental representations in children with autism: Implications for theory of mind. Journal of Autism and Developmental Disorders, 34, 637–648.

Li, T., Decety, J., Hu, X., Li, J., Lin, J., & Yi, L. (2019). Third-party sociomoral evaluations in children with autism spectrum disorder. Child Development. Advance online publication.

Liebal, K., Colombi, C., Rogers, S. J., Warneken, F., & Tomasello, M. (2008). Helping and cooperation in children with autism. Journal of Autism and Developmental Disorders, 38, 224–238.

McCleery, A., Divilbiss, M., St-Hilaire, A., Aakre, J. M., Seghers, J. P., Bell, E. K., & Docherty, N. M. (2012). Predicting social functioning in schizotypy: An investigation of the relative contributions of theory of mind and mood. Journal of Nervous and Mental Disease, 200, 147–152.

Ponnet, K., Buysse, A., Roeyers, H., & De Corte, K. (2005). Empathic accuracy in adults with a pervasive developmental disorder during an unstructured conversation with a typically developing stranger. Journal of Autism and Developmental Disorders, 35, 585–600.

Russell, J., & Hill, E. L. (2001). Action-monitoring and intention reporting in children with autism. Journal of Child Psychology and Psychiatry, and Allied Disciplines, 42, 317–328.

Sebanz, N., Knoblich, G., Stumpf, L., & Prinz, W. (2005). Far from action-blind: Representation of others’ actions in individuals with autism. Cognitive Neuropsychology, 22, 433–454.

Vivanti, G., McCormick, C., Young, G. S., Abucayan, F., Hatt, N., Nadig, A., . . . Rogers, S. J. (2011). Intact and impaired mechanisms of action understanding in autism. Developmental Psychology, 47, 841–856.

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