Autism at the cognitive level: What makes a good theory?

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

Autism at the cognitive level: What makes a good theory?

AT THE START of this book, we framed our approach in terms of different levels of explanation. We have so far attempted to give a fairly concise, descriptive account of the behavioural and biological features of autism as currently conceived. However, the bulk of the book focuses on the intermediate level — the cognitive level. Cognitive theories are the explanatory models that attempt to link biological heterogeneity with behavioural heterogeneity via cognitive simplicity (if only!). Given the vast complexity and variability in the autism constellation at both biological and behavioural levels, the cognitive level may be our best bet for encapsulating autism in a meaningful and concise manner. For example, differences in social interaction can manifest very differently depending on age, presence of intellectual disability (or not), environment, personality, gender and so on. And yet we may attempt to explain all these multiple presentations via shared cognitive roots.

Here we will consider three, key, observable domains of autism, exploring how these have been represented at a cognitive level, in order to introduce key terms, methods and concepts. Later chapters will interrogate the evidence for and against different theories, but here we focus merely on defining and describing the cognitive level of explanation. We will also provide an overview of the potential value of cognitive models and a framework for critical appraisal of psychological theory. Note that we use ’cognitive’ and ’psychological’ virtually interchangeably — cognitive here encompasses the processing of emotional information as well as learning processes and skills such as executive functions, memory and language.

1. Understanding social interaction and communication at the cognitive level

Social interaction and communication challenges are often characterised at the cognitive level as manifestations of a difficulty with ’mentalising’ or ’Theory of Mind’ (ToM). These terms both describe the everyday ability to recognise, represent and interpret the mental states (e.g. beliefs and intentions) of others. The published literature gives strong primacy to this domain, as a focus of both experimental research and as a support target. Despite the well- recognised complexity and multi-factorial consequences of autism, frequently, it is viewed in terms of social differences. As a result, a significant majority of experimental and qualitative research, and of autism-specific supports, focus on the social domain.

Mental states encompass a wide range of beliefs, thoughts and feelings which we have taxonomised as shown in Figure 5.1. The category of mental state cognitively underpinned by ToM is the propositional attitude — examples include believing, intending and pretending. These are important in psychological accounts, because keeping track of other people’s propositional attitudes requires meta-representations: representations of representations. Meta-representations allow us to form a mental picture of the world not just as it is, but as it might be in the mind of another person (more on these in Chapter 6). However, there are other types of inner state which are important for social interaction, but don’t require meta-representation. Emotional states include happiness, sadness or anger, and also more complex feelings like jealousy, embarrassment and tenderness. Recognising emotional states does not require ToM — but thinking about why someone seems angry might require inferences about their beliefs. Seeing that someone is angry, doesn’t require meta-representation, but working out that “she is angry that I forgot her birthday” and also “I don’t need to apologise because actually I have a surprise party planned, but she doesn’t know that”, does need ToM. Hard-to-interpret emotional cues like conflicting facial expression and speech (e.g. “I really detest this cake” while raising one eyebrow, smiling and reaching for another slice), or ambiguous emotional signals might be elucidated using meta-representational information about what the person thinks or believes.

Figure 5.1 A taxonomy of mental states

Underneath emotional states we find emotional drives, which can be further sub-divided into simple (e.g. wanting a coffee) and complex (e.g. wanting that the next café serves a good espresso). Again, interpreting others’ emotional drives might involve meta-representation (e.g. “she is walking that way because she thinks there is a coffee shop there, but I know that there isn’t”). Emotional drives overlap with emotional states but are more closely tethered to specific behaviours. They, in turn, are often founded on the most basic type of internal state — an interoceptive state, or physical sensation, such as hunger, pain or tiredness. It is important to remember that the psychological term ’ToM’ in its original sense refers only to those mental states that require meta- representation. As we can see, this involves only one category of inner state, though in the real world understanding propositional attitudes interfaces with our interpretation of a much wider range of behaviours and feelings.

Chapter 6 includes a more detailed examination of how difficulties understanding mental states may explain a range of social communication differences in autism. In brief, one possibility is that difficulty spontaneously constructing or holding meta-representations in mind prevents autistic people from readily understanding other people’s beliefs and intentions. We will see how this might in turn interfere with a range of social interactive skills — pretending during play, understanding referential aspects of language (e.g. the difference between “my cup” and “your cup”) or telling a story.

2. Measuring social cognition

A range of tasks has been used to measure mentalising in the psychological literature, most of them requiring the participant to track someone’s mistaken or false belief (Wellman et al., 2001). This has been considered the litmus test for mentalising because it requires attribution of a belief that is different from the participant’s own belief (and reality). The classic ’false-belief’ test is the Sally-Anne task, illustrated in Figure 5.2. In this task, the experimenter tells a story using two puppets, Sally and Anne. Anne moves Sally’s ball while Sally is absent from the scene. The participant must identify where Sally will look for her ball when she returns — the correct answer is that she looks in the place where she left the ball: Sally has a false belief that it is still in the basket, even though (as the participant knows) it is actually in the box.

A wide range of other tasks have been used to test mentalising. These include false-contents tasks, like the Smarties Task, where a child finds out that a tube of sweets actually contains crayons, but must correctly predict that someone else would guess it contained sweets. The false-photograph task was developed to capture meta-representations in a non-mentalising context. In this task, the child has to hold in mind the dual representation of a doll’s clothes now (red dress) and in a photo taken earlier (green dress). For older children there are second- order false-belief tasks. These involve stories where one character has a false belief about another character’s belief — e.g. Mary thinks that Jon doesn’t know where the ice-cream van has gone (but actually Jon does know). For adults, tasks like the Strange Stories present realistic scenarios (in written vignettes, Happé, 1994, or more recently short films; Murray et al., 2017) which require mentalising to interpret the non-literal utterance or ambiguous action of the characters, such as telling a white lie about an unflattering haircut or using sarcasm.

Figure 5.2 The Sally-Anne task

Reproduced from Uta Frith, 1989 with kind permission of the author and of the artist, Axel Scheffler.

3. Mentalising and emotion

What about differences in perceptual exploration and identification of emotional expressions from faces, body language and stories? If we find differences here, this suggests that there’s something going on in autism that extends beyond meta-representation specifically. While there is evidence of autistic differences, many studies report no difference (Uljarevic & Hamilton, 2013), and there is burgeoning evidence to suggest that some emotion recognition difficulties are due to co-occurring alexithymia — difficulty recognising what emotion you are feeling — rather than to autism (Bird & Cook, 2013). Furthermore, it is not clear whether emotion recognition in autism relies on different mechanisms to those found in neurotypical groups, nor whether such differences are fundamental to autism, or merely a result of reduced or different interpersonal experience (Harms et al., 2010).

Recent data have also begun to reveal that neurotypical people struggle to recognise and interpret the emotional reactions of autistic people, even while rating them as equally intense and expressive (Brewer et al., 2016; Sheppard et al., 2016). At the interoceptive level, there is evidence that some autistic people may experience even fundamental states, such as hunger and pain, differently from the general population (DuBois et al., 2016; Moore, 2015). Alexithymia has been linked to altered interoception, as part of a wider difficulty identifying how you feel inside (Murphy et al., 2018). Understanding more about interoception in autism may be important for elucidating sensory sensitivities and experiences, as well as differences in the social domain. Given the fact that there are probably differences between autistic and non-autistic people in how they experience, express or perceive emotional states and interoceptive states (even though we haven’t quite pinned these down), it seems that our cognitive explanations need to explore more than just meta-representation to explain the social features of autism.

At the same time, it is essential to emphasise that if people with autism have problems knowing what other people are thinking, that doesn’t mean they do not care how they feel. There is a dissociation between the skills required to comprehend the mental states of other people (mentalising), and emotional empathy. An example is illustrated in Figure 5.3. A mother was crying after the family cat had died. Her autistic son went and searched through all of his toys and found a hard, plastic figurine with a human body and a cat’s head, from the children’s cartoon show, Thundercats. In an attempt to comfort his crying mother, he prodded her gently with the figurine and, when she ignored him, finally held it up very close to her eye, pressing it into her face. This didn’t have the effect of comforting the mother, but it seems clear that the boy was motivated by empathy and a strong desire to help, even though his solution (while logical) didn’t really work.

Thus, we can see a dissociation in autism between knowing what someone thinks, which may be hard, and caring and feeling with them, which comes naturally. The same distinction is apparent in a very different group; those with psychopathy. Psychopaths are the opposite of autistic: they are good at telling what you are thinking and may use that to manipulate you, but don’t give a damn about your feelings. In autism, feeling empathy must also be distinguished from expressing empathy — the latter may prove challenging for some autistic people, especially if expected to act in narrowly defined, normative ways to show they care. Finally, we should note that it is also possible that many, if not all, of the social domain differences observed in autism, may in fact be underpinned cognitively by non-social explanations — this theoretical angle is explored in Chapter 8.

Figure 5.3 An illustration of mental state mis-understanding, with good emotional empathy, in real life

Reproduced by kind permission of the artist, Axel Scheffler.

4. Understanding restricted and repetitive behaviours and interests at the cognitive level

The first point to reiterate here is that the term “restricted and repetitive behaviours and interests” (hereafter, RRBIs) is not popular among the autistic community, for understandable reasons. It pathologises a set of behaviours that vary widely in complexity and function, and give great pleasure and comfort to many. It is true that some RRBIs can present serious disadvantages — self-injurious behaviour falls into this diagnostic domain, but even in that case, simply forcing someone to stop a self-injurious behaviour can do more harm than good. In other cases, it is important to reflect on whether our judgements of RRBI result from adherence to unnecessary and to some extent arbitrary social expectations of behaviour. Despite these considerations, we will use the term RRBI in this book to enable correspondence between our discussion and the wider psychological and clinical literature.

The cognitive level explanation of RRBIs is (even) less clear cut than for social and communication differences associated with autism. Lorna Wing characterised these as a consequence of a failure of imagination (notably pretend play) in autism — for her, repetitive behaviours traced their roots to a failure to think of anything else to do. Another possibility was that RRBIs were the flipside of reduced engagement with the social world. They filled the space that would otherwise normally be taken up by interpersonal interactions. In both accounts, RRBIs take root through lack of an alternative. However, this is no longer the prevailing explanation. More recent characterisations of RRBIs seem to have reversed the hypothesised causal link between imagination (itself a poorly defined and hard to measure construct in psychological terms) and these behaviour patterns. It is now recognised that RRBIs are themselves motivating and self-sustaining, and may promote a particular form of creativity. In fact, there is increasing interest in, and awareness of, the imagination and creative talents of autistic people (Diener et al., 2014; Pring et al., 2012).

Why might RRBIs be so rewarding for someone with autism? One possibility is that they are a behavioural manifestation of anxiety — perhaps a response to the experience of operating in a social world which is hard to comprehend. RRBIs might serve to add an element of control to the world, thus reducing anxious feelings. There is some evidence supporting this account. Measures of anxiety correlate highly with measures of RRBIs in autistic groups, and both RRBIs and anxiety may be linked to an underlying psychological construct called intolerance of uncertainty (Wigham et al., 2015). If autistic people have high intolerance of uncertainty, this could cause anxiety, especially in social situations. This is because, if you struggle to represent other people’s mental states and thus predict their behaviour, social situations probably involve high levels of uncertainty. To reduce uncertainty and anxiety, you might engage in predictable, familiar actions or routines. Comforting repetition may also not become boring if, as psychophysical studies suggest, habituation to familiar or repeated stimuli is reduced or even absent in autism (Sinha et al., 2014). Reduced habituation might mean that both pleasant and unpleasant sensations retain their initial fascination/repellence for longer, resulting in both RRBIs and sensory sensitivities.

RRBIs may also be related to patterns of ability, seen when running neuropsychological batteries of assessments with autistic people. A so-called uneven profile of abilities is often apparent, with enhanced or preserved skills in aspects of performance IQ relative to scores on verbal IQ tests. An area of autistic strength (e.g. attention to detail) may be instrumental in developing a particular interest (e.g. collecting thermostats). It is thought that enhanced visuo-spatial skills may explain some common areas of interest and expertise in autism, such as STEM subjects (science, technology, engineering, maths) or train spotting. Alternatively, an intense and focused interest might create the enhanced skill via a process of repeated practice and specialisation. Thus the causal primacy of cognitive abilities like attention to detail and RRBIs remains unclear. Another factor which muddies the waters in terms of determining causal precedence, is that having enhanced attention to detail seems a necessary pre-requisite for high levels of ’insistence on sameness’. After all, it’s hard to get stressed about small changes in your environment if you don’t notice them. To put it another way, a high level of attention to detail might result in a manifestation of insistence on sameness that is particularly hard to accommodate, since even tiny changes to the environment will be detected and cause distress.

In the future, it might help our cognitive models if we divide RRBIs into two factors that reliably emerge at the behavioural level:

Repetitive sensory motor (RSM) behaviors and insistence on sameness (IS) behaviors … RSM behaviors are characterized by repetitive use of objects, complex motor mannerisms, and sensory seeking behaviors … IS behaviors include rituals, compulsions, and resistance to change in routines.

(Hundley et al., 2016, p. 3449)

Hundley and colleagues point out that while repetitive sensory-motor behaviours are associated with younger ages and lower IQ, insistence on sameness is either increased with age, or independent of that factor. Another way to further our understanding of RRBIs is to adopt more qualitative and creative methods to understand how RRBIs operate and are experienced in the lives of autistic people (e.g. Dickerson et al., 2007).

5. Understanding sensory, and other features, at the cognitive level

Sensory aspects of autism include both hyper (over) and hypo (under) sensitivity. At the cognitive level they may be understood within an attentional framework by positing a relationship between detail focus and patterns of sensory sensitivity. For example, a focus on the details of a sensory input — like the presence of a specific frequency in a complex audio stimulus like the sound of a hoover — might make that sound especially unbearable. As with other domains of interest, the causal links in this case are opaque: is attention to detail a manifestation of sensory sensitivity, or a consequence thereof?

Sensory symptoms can also be linked, at the cognitive level, with some RRBIs, especially physical ones like flapping or rocking. This explanatory framework characterises RRBIs as a way to screen-out unwanted input by overwhelming the system with self-stimulatory behaviour: e.g. I can’t hear the humming of the lights if I am totally focused on the sensation of flicking my fingers in front of my eyes. The same system may be used by people who experience chronic hypo-sensitivity and use self-stimulation to provide more input to an under-stimulated system. Some autistic people have talked about needing weighted vests or blankets, or tight clothing to help them identify the edges of their body in space.

One interesting line of research suggests that autism may be characterised by increased perceptual capacity, which — at first sight paradoxically — leads to greater distraction by task-irrelevant stimuli (Remington et al., 2009). The idea is that we are less distracted when all our perceptual capacity is fully engaged by a task with high ’perceptual load’ (e.g. spotting a target among very similar items), but when the task becomes simpler (the targets are easy to spot), distraction by irrelevant stimuli increases. High perceptual capacity in autism might result in irrelevant sensory stimuli intruding on attention more easily, leading to unwanted distraction by sensory stimulation.

There are multiple other experiences, common to many people within the autistic community, that do not have their own cognitive explanation. In fact, it is a frequent objection among autistic people, and those who support them — like their parents — that some of the most pressing issues in their lives are under-studied by scientists. Where, they ask, is the research on potty training? On sleep? On employment skills? And they are right that a tiny minority of research focuses directly on these issues, which have an unquestionably strong impact on people’s lives. On the one hand, some of these issues may not be autism specific and so are addressed by research in other domains (e.g. trials of melatonin for sleep). On the other hand, one would hope that all autism research that improves our psychological theories will contribute to understanding and addressing these practical issues. Taking sleep as an example, again, understanding why autistic people struggle with sleep may involve understanding anxiety and stress. If these, in turn, are a result of operating in a social world organised according to neurotypical expectations that are often bewildering, any research that attempts to understand and better accommodate the social differences between people with and without autism is actually relevant (in the long run) to improving sleep problems in autism.

6. What to look for in a psychological theory

This chapter has introduced some of the key terminology and concepts that we use to describe features of autism in cognitive terms. Subsequent chapters will take an in-depth look at different sorts of theories that have been put forward to account for autism, examining the evidence for and against in each case. One final step before we proceed however, is to agree what to look for in a psychological theory. What work must a theory do for us? How can a theory open our eyes to facts, and how can we avoid being blinded by our preconceived notions? Table 5.1 summarises what we might look for in a good theory of autism.

Think back to Chapter 1, when we described how autism could be a condition with multiple biological and behavioural manifestations, united by a common cognitive dimension. If several features co-occur reliably, the most parsimonious explanation is that they are caused by the same underlying differences. Despite the huge variety within the autism constellation, and changes in our formulations of the diagnosis, autism is defined by the co-occurrence of social and communication differences with RRBIs. Thus it is possible, perhaps even probable, that a single cognitive characterisation could underlie these diverse features of autism.

It is worth highlighting that, while a good theory should give a causal account it should not be construed that this means these theories advocate for autism ’cures’. Understanding what comes first can help with understanding and support and does not imply a preventative agenda. A good theory should also be supported by high-quality evidence. It is important to be clear that this doesn’t preclude the use of qualitative methods — it is essential in research that the selected method is a good match with the question being asked. For many questions, qualitative research methods are the most appropriate and can provide unique opportunities to probe, for example, the nature of the social construct of autism (O’Reilly et al., 2017) or practical ways to make a community setting more autism-friendly (Fletcher-Watson & May, 2018). In other cases, qualitative research offers rich complementary information about phenomena identified experimentally, or raises new questions for quantitative investigation (e.g. Losh & Capps, 2006 on experiences of emotion in autism). However, when quantitative techniques are required, it is right to expect the highest standard of evidence. Anything less risks selling the autistic community short, by providing a wobbly foundation for policy and practice.

Table 5.1 What to look for in a good autism theory


Concrete predictions, which generate rigorous tests


An interpretation, not just a simple description, of the evidence


Detailed explanation of the pattern of characteristics in the autism constellation


A causal account


Alignment with basic scientific truths, including what we know about typical development


Informed by community perspectives and priorities

In 1994, when this book was first published, adequate sample sizes, comparisons across syndromes and high-quality methods, such as RCTs and longitudinal studies, were few and far between. Now that the research field has grown-up, we will be focusing on high-quality studies and especially systematic reviews and meta-analyses wherever these are available. At the same time, we recognise that new ideas are coming up in autism research all the time. These are often initially tested in small experimental studies, which can be strengthened by the use of clever experimental methods, including well-designed control tasks, and measurement of potential confounders.

Where good quality evidence is available to support a robust theory, this can influence both behavioural and biological levels of understanding, and in turn shape best practice in schools, clinics and communities. Psychological theories can provide a supporting framework for the interpretation of biological data (neuroimaging, genetic profiles) and help develop translational measures that increase the impact of research with animal models. A strong theory can flag what behavioural features are distinctive to autism and important for diagnosis or measurement over time. Where there are challenges, psychology directs our attention to the underlying problem (e.g. anxiety) rather than placing the emphasis on the surface behaviour (e.g. head-banging). If we move towards comprehension of the deep underpinnings of autism this can lead to greater empathy, patience and understanding from would-be allies. In addition, good quality research can highlight what’s not different in autism — what’s common to all of us and how we can use this information to develop shared understanding and opportunities?

A key part of the quest for a psychological theory is robust and unbiased examination of assumptions. It is instructive to ask yourself (if, like us, you are a neurotypical Brit), is life for an autistic person a bit like life for me if I found myself plonked down in rural Japan? What would I do there that would be typically autistic? Probably fail in your attempts to communicate and misunderstand people a lot, sometimes cause offence, be highly stressed, struggle to make friends or get a job, adopt fixed routines and copy others, as coping strategies. And if you also look obviously different, you would have it relatively easy. The Japanese folk around you would realise you haven’t a clue, forgive your faux pas as due to ignorance rather than bad intentions and adapt their behaviour to accommodate your needs. Autism, as a hidden disability, confers no such opportunity. Throughout this book, we try to adopt neutral terminology, emphasising difference rather than deficit — while recognising the disadvantages that being autistic may bring in a neurotypical world. In any attempt to explain autism that follows, we ask readers to bear in mind that we work within a paradigm defined by a non-autistic majority who have developed social rules and traditions that work for them. Reflecting on this position is an essential part of ensuring that we are not too quick to translate an autistic difference into an autistic deficit, simply because it doesn’t fit with those norms.

7. Current debates


The diagnostic and behavioural features of autism can be interpreted via cognitive frameworks to provide simpler models to make sense of a seemingly diverse set of behaviours in a (relatively) unified way. Examples include: describing social features of autism in terms of meta-representation, linking restrictive and repetitive behaviours with attention to detail and understanding sensory sensitivities in the context of increased perceptual capacity. A robust psychological theory can connect lived experiences and biological data into a coherent model with implications for best practice.

Big questions

There has been, historically, an overwhelming focus on the social and communication domain in psychological literature. Is it warranted? Does this correspond with the primary needs of people with autism? Or is it more aligned with researcher interests — for example, using autism as a ’window on typical development’?

How can we link cognitive research more clearly to community priorities? Are the connections we might write about in academic papers likely to come to fruition? How can we ensure that psychological research follows through, by delivering theoretically informed and evidence-based supports for the community?

What are the causal relationships between different features of autism? Is it possible, or desirable, to identify features that are drivers versus consequences of autistic development? For example, are RRBIs a typical response to living in a social world that is confusing and stressful? Or a fundamental part of a uniquely autistic experience?


There are more things in heaven and earth Horatio, than are dreamt of in your philosophy.

(Shakespeare, 1603)

Upon completion of this chapter, the prudent reader may engage in a critical reflection on the merits and short-fallings of Autism theory. In so doing, one should consider the ubiquitous challenges and limitations associated with theorising about human cognition, emotion and behaviour. Not only do human beings abound with rich diversity in culture, language and belief systems, we are, by our very nature, inherently (biologically, psychologically and socially) complex. A conclusion may be that it is not advantageous to presume the essence of Autistic-ness can be captured and reduced to a single unitary psychological theory.

Theory can offer useful definition, structure and fixed boundaries of categorisation. By all means, let us profess that (the access to, and knowledge of) theory can be a convenient pragmatic tool and a comforting lynchpin when systematising large chunks of complex information. In my everyday professional practice, theoretical models of Autism can, for example, enable identification of thresholds of cognitive states, which can simplify the process(es) of universal recognition, and communication in both clinical practice and research settings. However, the narrow and finite dimensions of psychological theory cannot speak to the magnitude of autistic heterogeneity. There is a real threat to the fidelity of autism research and practice if we blindly subscribe to theory. For instance, expressed beliefs such as “autistic people can’t feel empathy”, or “autistic people don’t have the ability to mentalise” and “it’s impossible to be in a caring profession if you’re autistic” are damaging to the autistic community and likely derived from out-dated autism theory.

A rudimentary glance over the past 70 years of autism research uncovers a history that can be labelled as progressive at best, and turbulent at worst. Our collective knowledge and understanding of autism, as expressed through current psychological theory, is perhaps most sensibly categorised as transient, momentary and imperfect. The authors wisely highlight “new ideas are coming up in Autism research all the time” and so, amidst the frequent evolution of novel research hypotheses, experimental paradigms, and theoretical perspectives there is but only one certainty of which we can be sure: The human-ness of autistic people (their families, friends and carers) transcends anything a theory has to offer. Only by opening up to the lived experiences and lifelong narratives of autistic people may we achieve an understanding of what it means to be autistic.

Recommended reading

Adams, J. (2017). Active but odd: An unswerving ’compulsion’ to create. Blog post published at Museum for Object Research,

Baron-Cohen, S., Tager-Flusberg, H., & Lombardo, M. eds. (2013). Understanding other minds: Perspectives from developmental social neuroscience. Oxford: Oxford University Press.

Frith, U. (2008). Autism: A very short introduction (Vol. 195). Oxford: Oxford University Press.

Grinker, Roy Richard. (2007). Unstrange minds. New York: Basic Books.