Autism at the cognitive level: Primary deficit models

Autism: A New Introduction to Psychological Theory and Current Debate - Tiffany Watt Smith 2019

Autism at the cognitive level: Primary deficit models

THE FIRST ATTEMPTS to understand autism within a modern psychological framework came in the 1980s. Autism was more narrowly defined at that time (in the DSM-III) and still considered a relatively rare condition. Research was framed by Wing’s “triad of impairments” framework — social interaction, communication and imagination — which was represented in the DSM-III as a triad of social interaction, communication and RRBIs. Nevertheless, psychologists aimed to determine a single, underlying difference that could explain all the diverse features used to diagnose autism. Three commonly accepted criteria for assessing the primacy of a specific feature are universality among people with the condition, specificity to that condition and causal precedence in development. In addition, any such theories of autism must steer a course between explaining too little — in effect merely describing features — and explaining too much. Autistic people show a strikingly uneven profile of abilities, with superior skills which require explanation just as much as do areas of difficulty. The challenge was to pinpoint pivotal psychological skills, significant enough to explain the differences between people with and without autism, but nuanced enough to allow for uneven profiles of ability.

Today, we do not characterise autism in terms of deficit or impairments, but as a pattern of differences which present advantages and disadvantages in relation to neurotypical social norms and expectations. However, these early, primary deficit models were formulated in a context where autism was almost universally seen as an impairment. Rather than whitewash this historical fact, we choose to use the term ’primary deficit’ to describe these explanatory frameworks (though their more modern iterations could be characterised as ’primary difference’ models). It should be noted that they represented a significant advance on earlier notions that autistic children were fundamentally uninterested in others and/or suffering from childhood psychosis. They were also formulated based on work mainly with children with autism and intellectual disability, often with additional language impairments — young people with a clear and significant need for support. Another key fact about this time in psychology is that we tended to adopt a modular view of brain- behaviour relationships when attempting to explain developmental phenomena. Many psychologists were looking for some discrete ability which was ’switched off’ in autism but switched on in everyone else. This notion now seems out-dated — psychology today is engrossed by the multiple interacting factors that contribute to every developmental milestone and group difference. Nevertheless, we can learn a great deal from these early attempts to characterise autism at the cognitive level.

1. The Theory of Mind model

The term ’Theory of Mind’ (ToM), referring to the ability to attribute independent mental states to oneself and others in order to explain behaviour, was coined by Premack and Woodruff (1978) in the context of exploring chimpanzee social cognition. The attributed mental states were ’independent’ both of the real world state of affairs (because people can believe things that are not true) and independent of the mental states other people have (because you and I can believe, want and pretend different things from one another). The philosopher Daniel Dennett (1978) pointed out that only understanding and predicting a character’s behaviour based on a false belief could demonstrate ToM conclusively, since otherwise the real state of affairs (or the participant’s own beliefs) would be sufficient to pass the task without the need to postulate mental states at all (Dennett, 1978). Thus the Sally-Anne, false-belief task (shown in Figure 5.3, previous chapter) was the litmus test used by Uta Frith, Alan Leslie and Simon Baron-Cohen to assess ToM in autism (Baron-Cohen et al., 1985). In this seminal paper, a significantly higher proportion of the autistic group answered incorrectly on the false-belief task, compared with typically developing and learning disabled groups. The ToM account of autism — which posited a meta-representational deficit as the defining feature of the condition — was born.

The original version of this book, written when the resulting landmark paper was still relatively new, devoted much of a chapter to describing how the ToM account connected Alan Leslie’s theory about the meta-representational underpinnings of pretend play, with the observation, by Lorna Wing and others, that pretend play was largely absent in autism. Now the ToM account of autism is over 30 years old, and the original 1985 paper has more than 8,000 citations on Google Scholar as we write! The theory has had enormous influence on research, but also permeated thinking about autism by teachers, parents and clinicians. Why? Perhaps because, in its original form at least, it made clear and specific predictions about what aspects of ’doing social’ would be hard for autistic people, and — just as importantly — which would be easy.

The ToM explanation of autism allowed researchers to make clear cuts between what appeared to be very similar behaviours — ’carving nature at the joints’ according to a precise theory about the underlying cognitive ’bone structure’. For example, Attwood et al. (1988) found that in autism the clinically noted reduction in use of gestures actually applied only to those gestures that normally influence mental states (e.g. expressions of consolation, embarrassment and goodwill), whereas children with autism showed as many gestures that manipulate behaviour (e.g. signals to come, be quiet, or go away) as did comparison participants with learning difficulties. Similarly, Baron-Cohen (1989) found that a group of autistic children did not use or seem to understand pointing for the sake of sharing attention (protodeclarative pointing) but were able to point in order to get a desired object (protoimperative pointing). Other fine cuts have been made between, for example, understanding seeing versus understanding knowing (Perner et al., 1989; Baron-Cohen & Goodhart, 1994) and recognising happiness versus recognising surprise (Baron-Cohen et al., 1993). Such distinctions in the smooth continuum of everyday behaviour would appear to be hard to derive from or explain by other psychological theories of autism (e.g. primary problems in social motivation — see Chapter 7 for more on this).

The other major appeal of the ToM account of autism was its apparent ability to explain all of Wing’s triad. The hypothetical developmental sequence went as follows: autistic children start out with an inability to meta- represent, apparent in a lack of pretend play; social differences then arise because meta-representation is required to understand people as agents with independent minds; thus, the characteristic communicative profile of autism would follow from an inability to represent intentions, or to recognise utterances as conveying a speaker’s thoughts.

2. Variants of the Theory of Mind model

The ToM model of autism became extremely influential in autism research and continues to be a central component of a lot of teaching and training about autism. Regardless of the detail of whether its hypotheses have been upheld by the evidence (more on that next), there’s little doubt that talking about autism in terms of difficulties understanding what other people are thinking is a useful conceptualisation for neurotypical people new to the condition. Uta Frith’s influential 1989 book ’Autism: Understanding the Enigma’ brought the ToM account of autism to a wider audience, including parents and teachers. Following its initial break-through moment, ToM-based accounts began to diversify and competing versions arose. Many of these aimed at trying to identify the developmental precursors to false belief understanding, following Alan Leslie’s groundbreaking work (Leslie, 1987).

Simon Baron-Cohen’s monograph, Mindblindness (1997), presented a model of the cognitive components of ToM, incorporating first an eye-direction detector and an intention detector. In the model, these two psychological components caused children to attend to other people’s eyes and intentions, which was necessary in order to develop an understanding of mental states in dyadic interaction contexts. Together these skills combined to provide the basis of a Shared Attention Mechanism, enabling triadic interactions between two people and a third shared focus (e.g. a father and daughter looking at a butterfly together). This mechanism provided the foundation of the ultimate ToM Mechanism required for understanding of mental states. For Baron-Cohen, the root of autism was a failure to develop and display shared attention, and this was supported by evidence that lack of shared attention was a reliable pre-diagnostic marker (Baron-Cohen et al., 1992).

Other variants on the ToM model asked how a child with typical development built this set of skills. Many rival theories arose (see Carruthers & Smith, 1996 for a range of relevant chapters). Theory-theory purported that children acted like little scientists, making hypotheses about the actions of those around them and testing them against reality. In this way, children would learn to predict how people would behave under certain conditions and extrapolate a complex set of rules about behaviour and beliefs. Others argued that the young age at which children understood false beliefs, and the relative universality of that early understanding, was incompatible with conscious reasoning as the route to ToM. Instead, Alan Leslie and others proposed that children were innately predisposed to develop the meta-representational ability underlying pretending and reasoning about others’ mental states. Meanwhile, simulation-theory suggested that the basis of ToM was an identification between the child and those around them — an awareness that other people are “like me”. This enabled the child to simulate another’s behaviour by asking themselves what they would do in a particular situation.

3. Challenges to the Theory of Mind model: universality

In every study using ToM tests, including the original 1985 paper, some autistic people are found to pass. How can we explain this test success? The first question is whether those autistic participants who pass ToM tests are using the same sort of automatic and intuitive mentalising as do neurotypical participants, or instead figuring out what others think via another route? If the latter, then difficulty intuitively representing other’s mental states may still be universal to autism, and a good explanation for social and communication differences.1 If, on the other hand, we believe that some autistic people can represent mental states, but still show the social and communication features of autism, then a second possibility is that a delay in acquisition of mentalising abilities reverberates across the lifespan, causing persistent differences in interaction. Delay could disrupt the normal coordination of the mentalising system with other abilities during development and potentially interfere with years of formative social experience. A parallel might be drawn with learning a second language in early versus later life; the latter rarely results in totally fluent, accent-free use. The second language analogy is helpful, perhaps, because speaking a second language learnt late can be enjoyable or frustrating, can be exhausting and harder when you are tired; all things that may apply to the social understanding that autistic people develop through their own efforts later in life.

1It seems important to reiterate here that an “inability to represent mental states” for an individual, or an entire population, should not be taken to imply that an autistic person does not care about or have an interest in other people. We simply mean that the underlying, automatic representational mechanism used by the non-autistic population for ’mind-reading’ may not be ’wired in’ for autistic people.

A highly intelligent young woman with autism told Francesca that, when tested on the Sally-Anne task, she had wanted to answer ’box’, where the ball really was, but remembered that when psychologists ask questions there are always tricks, so she said the opposite, ’basket’, thereby passing the test! Many autistic people describe consciously working out what is going on in social situations and say that this feels like doing mental arithmetic. There is evidence that some autistic people have managed to “hack” a solution to ToM tasks, thanks to experience, and using general problem-solving skills (Frith et al., 1991). Such hacking may be relatively inflexible, allowing success on simple, artificial ToM experiments, but without the flexibility and automaticity required in real life. Various attempts have been made to create ToM tests that are more ecologically valid, less easy to hack and therefore more representative of the real-life social difficulties many autistic people experience.

The Strange Stories (Figure 6.1) were an early attempt to present complex real-world vignettes that require mentalising in order to understand the speaker’s intention behind a non-literal utterance (Happé, 1994); they have more recently been turned into video-based tasks (Brewer et al., 2017; Murray et al., 2017). The Frith-Happé Triangles Animations, inspired by the work of Heider and Simmel ask the participant to interpret the actions of abstract moving shapes (Abell et al., 2000). The descriptions participants produce are analysed for evidence that the animations have been spontaneously interpreted as intentional, marked by use of mental state terms. These tasks have revealed a persistent reduction in mentalising in autistic adults without learning disability, though exceptions to the rule remain. Further evidence of subtle differences between autistic and non-autistic approaches to mentalising comes from implicit ToM assessment via eye-tracking (Schneider et al., 2012). In these experiments, people watch a sequence of events from a Sally-Anne type scenario, while their eye-movements are recorded. When the ’Sally’ character returns to collect her hidden toy, neurotypical people tend pre-emptively to fixate where they expect her to look first, based on her false belief. Autistic participants tend not to do this, even when they pass the explicit verbal false-belief question (Schuwerk et al., 2015).

We have reviewed, two possible interpretations for the fact that many autistic people pass ToM tasks: one is that people may grow out of differences seen in childhood and another is that people can learn strategies to tackle ToM problems. While we know that some delay in ToM seems inevitable (there is no evidence to date of pre-schoolers with autism passing ToM tasks), so far there is no evidence available that can categorically distinguish between the “different strategy” and the “developmental delay” interpretation. Both accounts predict that autistic people should be able to solve ToM tasks when they are simple, but that a dis-fluency in application of mentalising should be apparent in more complex tasks and everyday life. One way to disentangle these accounts is to use longitudinal data which can tell us whether ToM task performance is predicted by age or intellectual ability. If ToM task success comes with age, that suggests a developmental delay account, but if ToM task success is more closely related to IQ then we might lean more towards an alternative strategy account. That said, delayed and different are not strictly alternatives — for example, early delays in ToM might lead to the development of different strategies later in life.

Figure 6.1 Some examples of the Strange Stories

When there is evidence that IQ influences ToM scores, this also invites a third interpretation — that mentalising is hard for autistic people because of an additional processing difficulty. We know that the intellectual capacity required to mentalise is not especially high — in typical development a mental age of four years is sufficient to pass standard ToM tests and to demonstrate this ability across a range of different tasks (Gopnik & Astington, 1988). In fact, implicit tasks are passed much earlier (see Chapter 7). Participants with learning disability, too, pass these tasks despite relatively impaired general intellectual and problem-solving abilities (Baron-Cohen et al., 1985). Data from the false-photograph task (see Chapter 5 and Figure 6.2 — in this task, the child has to hold in mind the dual representation of a doll’s clothes now and in a photo taken earlier) seemed to show that autistic children who struggle with mentalising can manage meta-representation outside the social domain. On the other hand, other work comparing verbal, non-verbal and non-social meta-representational tasks suggests that autistic children may have a generalised difficulty extending beyond mentalising specifically (Iao & Leekam, 2014). Consideration of domain-general information processing differences in autism will be addressed further in Chapter 8.

Figure 6.2 The false photo task

Reproduced by kind permission of the artist, Axel Scheffler.

Why does all of this matter? We want to be clear on why determining the root of the apparent, if not consistent, autistic difficulty with mentalising is important to psychologists. First, let us assume that being able to envisage, interpret and act upon the mental states of others is useful (this is not necessarily the case — more on this later in the chapter). Second, let us assume that the autistic community agree it is desirable for children to learn to apply this skill, in order to get by in a largely non-autistic world. In that case, we would want to understand the best way to teach mentalising skills to children on the spectrum. The best way to teach would probably be to help them replicate a strategy which had already been applied and honed by autistic people before them; which means understanding how that strategy works is a pre-requisite for teaching. One starting point is to examine more directly the ways in which autistic people think about mental states. In one innovative study, Edey and colleagues (2016) showed that, when asked to produce animations representing different mental states, the films made by autistic people were not just different to those created by neurotypical people, but also that neurotypical viewers struggled to interpret those films. This kind of work may be the key to supporting young autistic learners but also helping non-autistic people to confront their own role in the social communication barriers between people with and without autism.

4. Challenges to the Theory of Mind model: specificity and primacy

As well as problems with the universality of mentalising difficulties in autism, it has also been suggested that ToM difficulties are found in other groups apart from autism. If mentalising can be disrupted without resulting in the behavioural features of autism, this would seem to threaten its proposed causal role. Acquired or later life mentalising problems have been reported in many other clinical groups, including those with strokes leading to brain damage in the right hemisphere, and individuals with schizophrenia, who show over-attribution of mental states that may relate to paranoid delusions. In addition, three key groups in whom traditional false-belief failure has been shown developmentally are typically developing children under 3 years, children with learning disability (but not autism) and children born deaf into non-signing homes.

In all three cases, unlike in autism, failure on false-belief tasks is not commensurate with social difficulties in everyday life. For this reason, it is plausible that task failure is not due to problems attributing mental states per se, but to aspects of the task demands, such as executive control. After all, 2-year-olds talk about mental states, and deaf and learning disabled children make friendships in the usual way. Importantly, when task demands are reduced, for example using implicit false-belief tests measuring spontaneous looking behaviour, typically developing toddlers show evidence of tracking mental states as young as 15 months (Scott & Baillargeon, 2017; Setoh et al., 2016). Simple tests like the penny-hiding game — where a child gets to guess which hand an adult is hiding a coin in and then has a go at being the hider — also show that ToM operates more typically in those with intellectual disability but not autism, compared with those with both (San José Cáceres et al., 2015).

The originators of the ToM account set themselves a high bar by setting out to identify a primary deficit. Even if mentalising is universally challenging for autistic people, is this really the primary difference? One challenge is that for ToM to have a primary role in explaining autism, it needs to explain other features as well as social and communication differences. Some authors have proposed that RRBIs infiltrate the cognitive space that would otherwise be occupied by a rich database of social knowledge, built on early mentalising abilities. Keen interests in, say, Finnish architecture or bicycle maintenance simply absorb the resources that most non-autistic people give over to recording and processing details about other people. Another possibility, reviewed already in Chapter 3, is that RRBIs are a way to manage the anxiety that results from having to operate in a hard-to-comprehend social world. One prediction that results from these proposals is that RRBIs should be at their most intense when ToM difficulties are most profound. While some papers do find associations between ToM and RRBIs (e.g. Jones et al., 2018), the general lack of association offers further support for the ’fractionated triad’ approach to understanding autism at the cognitive level (Brunsdon & Happé, 2014).

Another challenge for ToM models is to determine the point in the developmental process at which the autistic and non-autistic pathways divide. As the ToM model was extrapolated — both in the study of autism and in investigations of typical development — a richer developmental sequence was revealed, making it harder to determine this point (see Figure 6.3). A significant factor in this process is the diversification of ways to measure ToM. Although the false-belief task remains the ’gold standard’, a range of experiments reveal ToM-relevant abilities that precede false belief in development. An example is desire-reasoning tasks in which children as young as 18 months can correctly recognise that someone else might have different wants — in this case a preference for broccoli rather than biscuits — than their own (Repacholi & Gopnik, 1997). Other studies have illustrated implicit mentalising-in-action when children from two-and-a-half years are able to lay a false trial in order to deceive others (Chandler Friz & Hala, 1989). Vasu Reddy pioneered research into early humour, showing how toddlers can make jokes and tease people in a way that seems reliant on a rudimentary mentalising ability (Reddy, 1991). Thus the capacity to represent the mental states of both oneself and others, or of reality and appearance, is evident from very early in development.

Figure 6.3 An approximate developmental sequence of milestones relating to Theory of Mind

In addition, mentalising skills have been associated with other early developmental milestones, such as attending to people, joint attention with others, imitation and recognising emotions (Baron-Cohen, 2000). Joint attention has been described as a ’pivotal skill’ in development that is often significantly delayed in autistic children (Charman, 2003). This term describes the three-way, or triadic, attention that can take place between two people and a shared external object. For example, when a mother points to an airplane in the sky and says, “Look!”, and her son looks up at her and then at the plane, and she looks back at her son and so forth … that’s joint attention. Likewise, imitation has been hypothesised to play a key role in development, providing a foundation for much learning but also a way to connect emotionally with others (Meltzoff, 1990). Figure 6.3 shows an approximate timescale of ToM- related abilities in typical development. As ever, the causal relationships here are hard to disentangle — a series of skills become apparent over a narrow developmental period and the extent to which one might be dependent on another remains uncertain.

This enrichment of our concept of ToM has given rise to some developmental models of autism. When mentalising was judged only by standard false-belief task success, the primacy of mentalising problems in autism seemed problematic; after all, children often show signs that they are autistic before they are old enough to take ToM tests. However, over recent years the use of looking-based tests of so-called implicit false-belief attribution suggests mental state tracking at 18 months (Senju et al., 2011) and even younger (see review by Scott & Baillargeon, 2017, but also Heyes, 2014 for a debate regarding interpretation of these findings). Now the emergence of mentalising no longer looks too late to play an early and causal role in, at least, the social and communication features of autism. On the other hand, work with cohorts of infants with a high-likelihood of later autism diagnosis has demonstrated few reliable behavioural markers of autism before age 12 months. Whether a narrow focus on meta-representation can explain the rich and complex developmental sequence in autism will be explored in more detail in the next chapter, where developmental models of autism are considered.

5. Alternatives to the Theory of Mind model

One other cognitive capacity that has been proposed as the primary deficit explaining autism is executive function. Executive functions are subject to a range of subtly different definitions, but they encompass those ’oversight’ abilities, reliant on frontal lobe function, which pervade any complex/novel task. Examples of executive functions, include planning, inhibition and working memory. There is considerable evidence that many autistic people have executive function difficulties (Hill, 2004), but are these a candidate for a “primary deficit”?

To be convincing in this field, an executive (dys)function model of autism must explain the social difficulties apparent in the real world, as well as those operationalised in false-belief tasks and other ToM measures, by recourse to a more general process. After all, executive functions operate across both social and non-social contexts. Some authors propose that an autistic child’s performance on false-belief tasks does not reflect a mentalising deficit, but rather a specific difficulty in overcoming the perceptual salience of the object in the real location. One test of this hypothesis examined the effect of a competitor on performance on a “windows task”. In this task, the child simply had to point to one of two boxes, into which she alone could see, via a small window. On each trial a sweet was placed in one of the boxes, and if the child indicated the empty box, she won the sweet. In the competitor version, an ignorant second player searched in the indicated box and kept any sweets they found — so that the same response (pointing at the empty box) resulted in the child being rewarded with a sweet, but this time in the social context of “deceiving” the competitor. Hughes and Russell (1993) found that autistic children struggled with the ’windows task’ in both conditions and concluded that their difficulties with deception did not spring from an inability to mentalise, but from a failure to inhibit action to the object.

Subsequent studies (e.g. Hughes et al., 1994) showed that on a number of non-social tasks — Tower of Hanoi, detour reach task — most children with autism do show some difficulty with acting away from the object or inhibiting a prepotent (previously rewarded) response. However, the debate about the primacy of executive function versus mentalising difficulties in autism is not so easily resolved (for an interesting examination of the developmental relationship between these abilities, see Wade et al., 2018). One meta-analysis of studies in typical development suggests that executive functions have a role to play in false-belief understanding, but not other aspects of ToM (Devine & Hughes, 2014). Early individual differences in executive function have been found to predict later false-belief performance, but not vice versa, which may suggest a developmental primacy of executive function — assuming that measures of the two constructs are of equal developmental sensitivity. However, the presence of ToM difficulties along with intact executive function has been found in autistic children, suggesting that executive difficulties do not underpin mentalising difficulties in this population (Pellicano, 2007).

The relation between executive function difficulties and social interactive behaviours in the real world (as opposed to ToM task performance) has not been adequately probed, but there is some evidence that executive function problems do link to social interactive and communication difficulties beyond the lab (Happé et al., 2006). On the other hand, this same paper showed that executive function difficulties are more severe and persistent in children with ADHD than children with autism. One requirement of a ’primary deficit’ model for autism is that it ought to be able to explain what is unique about the autistic profile — something traditionally undermined by the presence of overlapping but more intense executive function difficulties in a distinct condition. However, if autism is a composite of cognitive features (as suggested in the fractionated triad account), not all the ’ingredients’ need be unique to autism.

Can executive function difficulties explain other features of autism? Certainly many autistic people report difficulties with real world tasks that rely on executive function, such as planning their personal schedules or dealing with changes to a regular journey. Attempts have also been made to trace the roots of RRBIs to an executive function cause. One such is a suggestion that autistic people find generalisation and generativity difficult — these are the cognitive underpinnings for the transfer of knowledge across contexts and spontaneous generation of new content and solutions. It is proposed that such limitations might translate into a tendency to engage in the same behaviour repeatedly, across multiple settings. However, a recent study exploring ToM, executive function and diagnostic features of autism in a large group (n = 100) of diagnosed adolescents demonstrated links between ToM and autism profiles (including RRBIs), and between ToM and executive function, but not between executive function and autism behavioural profiles independently (Jones et al., 2018). One possibility is that executive function difficulties are not a core part of autism at the cognitive level, but that these may co-occur with autism quite often and that such executive function challenges make it harder for an autistic person to mask, strategically accommodate or compensate for their social and communication differences. Conversely, an autistic person with good executive functions might be getting by at school or at work using a series of rules, switching between these in response to changing contexts, relying heavily on working memory, attention, flexibility and planning. It is possible that competence in executive functions is one of the things that separates those who aren’t readily identified as autistic — including many women and girls — from their autistic peers. The same compensatory phenomenon might be apparent in a range of other conditions that diverge from typical development, as suggested by Mark Johnson (Johnson, 2012). Of course, the negative consequences of some such compensatory efforts, such as camouflaging, can be exhaustion, burn-out or worse (Hull et al., 2017).

6. The role of primary deficit models

Neither of the two models presented here has fully achieved what it set out to do — to provide a simple, cognitive-level explanation for a diversity of biological and behavioural features found in autism. Autistic performance on the various measures created to assess ToM in particular, but also executive function, remains complex. It has not yet been possible to disentangle the various ways in which an autistic person might address a ToM task. One of the reasons that data are lacking on this is that, since the introduction of the ToM model, the science has begun to focus on more nuanced, developmental theories of autism. While many of these have their roots in ToM, incorporating mentalising as part of an important array of social cognitive abilities, the focus on false belief in particular has lessened.

Despite the terminology of ’deficit’ and the determination to encapsulate the variety of autistic experience in a single cognitive label, we must acknowledge the beneficial descendants of this body of work. Mentalising remains a useful shorthand for a non-autistic person encountering autism for the first time. Considering that someone else might not easily comprehend your intentions or beliefs (even though they are interested in and concerned for you) can help neurotypical people to adapt their behaviour in ways that autistic people may find useful — e.g. be more explicit about their thoughts and feelings, make their language more literal and precise.

A desire for parsimony — with authors postulating the minimum number of underlying psychological features necessary to account for a behavioural pattern — is prevalent throughout psychology. However, it may be more convincing in the case of autism to propose multiple primary differences, at the psychological level as at the genetic and behavioural levels (Happé et al., 2006). A co-occurrence of multiple differences, unconnected at the psychological level might be explained by the spatial proximity of their biological substrates in the brain or by a single but pervasive difference in brain function or connectivity. If this is the case, it may be futile to look for a single psychological marker of autism. The ’fractionated triad’ account argues for different psychological factors impacting different aspects of the behavioural phenotype of autism, underpinned by largely independent genetic influences (Brunsdon & Happé, 2014).

7. Questioning assumptions about cognitive performance in autism

There are also other ways altogether to interpret these attempts to determine the psychological features of autism. One recent study examined the role of audience effects — the performance-enhancing impact of being watched by another person — on false-belief task scores in children with and without autism (Chevallier et al., 2014). The study revealed that the usual ’performance gap’ between typically developing and autistic children was eliminated when there was no observer. The authors conclude that some behaviours that have been considered characteristic of autism may derive from a lack susceptibility to audience effects rather than an underlying difficulty with the task itself.

Further, indirect evidence for this interpretation comes from parallels between patterns of data derived from studies with autistic children, and studies of in-group and out-group effects in typical development. For example, multiple studies find evidence that autistic children show less spontaneous imitation, and less fidelity in imitation, than children without autism (Williams et al., 2004). However, the same effect is seen when typically developing children are organised into two arbitrary groups (a “minimal group paradigm”) — children as young as 14 months old are less likely to imitate peers from the other group (e.g. Buttelman et al., 2013). When we use this result to frame our interpretation of autism research, suddenly we can see that a lot of autistic behaviour could potentially be understood in the context of feeling separate from, or different to, the experimenter. And of course, because experimenters in the field are rarely autistic themselves, this is a perfectly accurate and reasonable judgement of the situation.

8. Current debates


Late 20th century models of autism strove to identify a single core cognitive feature of autism that could explain all surface behaviours observed. These models were to some extent victims of their own determination to explain autism in one neat theory, as they could not adequately address the variability among people with a diagnosis, or explain features of autism whose importance researchers were only just beginning to acknowledge (e.g. sensory sensitivities). These theories, ultimately, explained too little.

In addition, much of this work adopted a view of autism that is no longer an acceptable approach, describing autism in terms of deficit and disregarding the strengths of, and contributions made by, autistic people. Nevertheless, these models have driven the field forward by drawing attention to fascinating aspects of typical and autistic development (how do we understand other people’s minds?). By identifying features of autism which were outside the reach of these attempted explanations, these researchers exposed the heterogeneity and complexity which is now a focus of much research and community interest.

Big questions

What is next for ToM? This construct is still used widely in research, often to explore connections with other variables — ToM and gender, ToM and employment, theory of mind and quality of life. Given the limitations of the construct as an explanation for the wide range of experiences within the autistic constellation, is it still useful? Or is a focus on ToM now merely a habit in the field?

Can autistic performance on ToM tasks be explained largely in terms of in-group and out-group or audience effects? And, if we adopt this framework, what does it mean for our understanding of typical social behaviours? It is reasonable to question whether mentalising and its consequences are necessarily a positive influence in our society. For example, in-group effects can lead to prejudice and unconscious bias, and obligatory mentalising mechanisms may be counter-productive for originality.

Some autistic people report having too much insight into other people — detecting every detail of someone’s behaviour, as well as feeling empathy very strongly. If so, it is possible that we have our models of autistic and non-autistic social skills entirely the wrong way around. Perhaps non-autistic people are relatively poor observers of human behaviour, and this gives us the ability to respond quickly and fluently, because we are only using the gist of the social situation to guide our decisions. Meanwhile autistic people are burdened with an unfiltered representation of every detail that prevents them from making handy generalisations or using holistic approximations to guide their social behaviour. We will return to this theme in Chapter 8.


While I struggle understanding peoples’ thoughts and actions, I have also dedicated a lot of my time to ’learning social’, much like learning a language as described in this chapter. Therefore, in the same way that a newcomer to English might have more knowledge of prepositions and the passive voice than a native speaker, I often see intricacies in interactions that others miss because of this active learning process. But I will always have an ’accent’ preventing me from fully integrating, and the constant practise and repetition of this learning is indeed frustrating and tiring.

The primary deficit model was damaging, especially in schools, as it paved the way for interventions aimed at making autistic individuals ’normal’. When I read about the early deficit models, I feel angry, as these have left an imprint in people’s minds. The idea that something is ’missing’ infers that it can be added or fixed at a later date, and although the neurodiversity model is working to amend these preconceptions, they have spawned many inappropriate interventions such as applied behavioural analysis (ABA). When people tell me they are engaging their children in these kinds of therapies, or in diet treatments like the gluten-free, casein-free approach, what I hear is that they feel that their child, and I, are not good enough as we are. That we need to be cured, fixed, tempered, basically changed to be more acceptable as human beings.

The primary deficit assumption that autistic individuals ’overreact’ or ’misinterpret’ all situations must be addressed as this could lead to serious problems. When I was at school I raised a potentially very serious child protection issue with a teacher; I was later told that I had overreacted to the information I had heard because I was autistic and that it was not being followed up because of this. Now, I don’t know if there was anything behind the ’banter’ I had heard in the common room, but I do know that my instincts were being questioned because of my diagnosis.

I welcome the idea of neurotypical ToM being taught by autistic individuals who have already mastered it. I think I would have gained a lot more from someone sitting me down at nine years old and telling me: there’s nothing wrong with you; you’re just different. These are the problems you are going to face, but it will get better.

Recommended reading

Baron-Cohen, S. (2000). Theory of mind and autism: A fifteen year review. Understanding Other Minds: Perspectives from Developmental Cognitive Neuroscience, 2, 3—20.

Beardon, L. (2008). Is Autism really a disorder part two — theory of mind? Rethink how we think. Journal of Inclusive Practice in Further and Higher Education, 1, 19—21.

Johnson, M. H. (2012). Executive function and developmental disorders: The flip side of the coin. Trends in Cognitive Sciences, 16(9), 454—457.

Leslie, A. M. (1987). Pretense and representation: The origins of “theory of mind”. Psychological Review, 94(4), 412.

Scott, R. M., & Baillargeon, R. (2017). Early false-belief understanding. Trends in Cognitive Sciences, 21(4), 237—249.