Autism at the cognitive level: Developmental trajectory models

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

Autism at the cognitive level: Developmental trajectory models

AS WE HAVE seen, Theory of Mind (ToM) was originally conceptualised in a fairly modular fashion: as a skill which appeared at a specific developmental time point (ultimately marked by the ability to pass false-belief tasks) and that could either be present or absent. However, updated accounts of social cognitive development situate false-belief task success within a more complex developmental trajectory. This incorporates both early examples of ToM, such as implicit belief tracking and desire reasoning, and related behaviours such as joint attention and emotion recognition.

Rather than focus on a single deficit (failure to switch on ToM) marked out by a simple test (false-belief comprehension) developmental theories of autism posit that early, subtle differences in the child’s relationship with their environment place them on an unusual developmental pathway that leads to a pattern that makes up the diagnostic features of autism (Happé, 2015).

1. Early signs of autism

As mentioned in Chapter 3, diagnosis of autism is possible from early childhood. The average age of reported diagnosis in a large European survey was 42 months (Salomone et al., 2016), though the range varied from 34 to 50 months when comparing between European countries. In addition, this sample were only aged up to 7 years old, and many autistic people are diagnosed much later, including in adulthood. Diagnosis can also happen before 3 years old, but there are question marks over how reliable such diagnoses might be. Salomone and colleagues showed that being female and having better language are associated with later diagnosis, suggesting that those children diagnosed very early may represent only a sub-set of the autism constellation. This may be compounded by the influence of socio-economic and ethnic factors on access to specialist diagnostic services (Daniels & Mandell, 2014). Another possibility is that when diagnoses are offered very early in development, these are not stable over time (Bieleninik et al., 2017). At the time of writing, there is no reliable way to diagnose autism in the first year of life.

This is largely because, as we have seen, diagnosis at any age relies on observation of a pattern of behaviours across multiple contexts, ideally by a multi-disciplinary team. There is no single biological or behavioural marker with a robust positive predictive value for autism. Instead, there exists a range of early, parent-report screening tools (e.g. the Checklist for Autism in Toddlers, the Early Screening for Autism and Communication Disorders and the Autism Parent Screen for Infants) or direct observational protocols (e.g. the Autism Observation Scale for Infants, the Autism Diagnostic Observation Schedule — Toddler Module) to facilitate early screening. These tools rely on so-called ’red flags’, which are lists of behaviours such as following a point or responding to their name being called, that are reduced (or, for some behaviours, e.g. spinning objects, increased) in children who later receive an autism diagnosis relative to the general population.

Theoretical accounts of autism reviewed in this chapter interpret these early, small disruptions to the typical developmental process as having larger downstream effects, which are eventually recognisable as autism. Waddington’s epigenetic landscape, shown in Figure 7.1, is a useful visual metaphor for this process. Although the landscape was initially put forward as a way to understand the actions of genes across the lifespan, it can also be used to conceptualise how early subtle differences can result, over time, in a radically different approach to the world, such as found in autism. In the image, the ball can represent the child and his or her development, which unfolds like a slow, downhill roll. Small differences in the starting position, direction or speed of the ball, or tiny bumps along the way, could result in the ball following paths that diverge over time to become radically differentiated. The hills which separate each path can represent the significant effort required if one were to attempt to move, later in life, from one path to another.

Figure 7.1 Waddington’s epigenetic landscape

2. Studying early signs of autism

There is great interest in how autism is manifest in the first year of life and even pre-natally. For some, this effort is aimed at discovering a potential ’cure’ for autism or even introducing pre-natal screening, highly controversial and ethically debatable aims (see this chapter’s Current Debates section for more). However, the majority do not endorse this viewpoint but are still interested in early autism development, and there are many good reasons for this focus. Parents often report distress over the long, complex and sometimes upsetting process of seeking a diagnosis. In a recent international survey, Sue and colleagues uncovered large gaps between the age at which parents first raised concerns and mean age at which diagnosis occurred (Fletcher-Watson et al., 2017). In some cases, parents may struggle to have their child’s needs recognised by clinical services, especially if the child is cognitively able and/or skilled at masking. Likewise, autistic people who receive a diagnosis in later life often regret the time it took to identify their autism — in a recent qualitative study of the experiences of autistic women and girls, nearly every person interviewed wished they had known earlier (Sedgewick, personal communication). Robust, biological early diagnostic markers would permit rapid, reliable diagnosis early in life, with potential benefits for autistic people, their families and also for society in areas like planning service provision. Early diagnostic markers would also allow proper investigation of any putative environmental factors contributing to outcome, and help put paid to scaremongering media reports about the supposed causal role of screen time or radio waves in autism.

There are also more purely scientific reasons to search for concrete diagnostic markers to further our understanding of how different developmental domains interact. For example, as we will see in more detail next, there is ongoing discussion of how social cognitive skills relate to ’domain-general’ (i.e. not specifically social) abilities. Are differences in attention to social content in a stimulus related to general attentional processes? Do our social interactions in the real world rely on a fully operational attention system or, conversely, do they provide a training environment for that attentional system to develop? Likewise, in the previous chapter, we raised the question of how executive functions might act as a compensatory mechanism for some autistic people. In terms of autism, diagnostic markers might also provide a basis for starting to understand heterogeneity in autism, resolving the issue highlighted in Chapter 3 that division of autism into meaningful sub-groups is virtually impossible while we are reliant purely on behavioural diagnosis.

For a long time, the main method adopted to try to identify the earliest signs of autism was retrospective parental report and video analysis. In these studies, researchers asked parents about their child’s earliest months and years, and gathered home video footage of children who had later received an autism diagnosis. Attempts were made to standardise the samples, for example, by focusing on videos of a child’s first birthday party. The method was far from perfect, but despite the lack of control and objectivity, retrospective analysis produced some important findings about early features of autism, including reduced smiling and orienting to name around 12 months (Palomo et al., 2006). Other studies focused on clinical records, finding higher levels of reported concerns in routine infant and toddler screening tests, specifically in the social domain, for children who later received an autism diagnosis (Johnson et al., 1992).

More recently, the focus has switched to prospective studies, which permit the use of experimental methods and uniform classification of all participants. The most established method of prospective research is to recruit parents who already have an autistic child, who are now having a new baby. These younger siblings have about a 20% chance of receiving an autism diagnosis, meaning that researchers can recruit, for example, 100 families and reliably gather early years data from a sample of 20 children with autism. Given an estimated prevalence of 1% in the general population, it would require recruitment of 2,000 families to achieve the same group size if recruiting from the general population. The first of these autism-sibling studies began to yield data in 2005, when Lonnie Zwaigenbaum and colleagues reported on 65 younger siblings who had been — at that point — followed up to 24 months (Zwaigenbaum et al., 2005). More recently, the field has begun to be augmented by work with other high-likelihood groups, including children of autistic mothers and infants born premature. Specific findings from these studies will be mentioned next, as we consider a range of theoretical models in turn, but a recent review of results suggests that while many group differences have been found, the field requires more replications before any test could begin to achieve candidate marker status (Jones et al., 2014).

Although sibling studies, and work with other high-likelihood infant groups, represent a significant methodological improvement on the retrospective work that went before, this literature remains beset by limitations. One key issue is the potential generalisability of findings. It is unclear to what extent findings from an infant-sibling group would generalise to autistic children born into families without existing autism experience, or this pattern of genetic predisposition. The specificity of results is also a problem. To date, no published sibling studies have recruited a relevant control group, such as children with a learning disability or a genetic susceptibility to (for example) dyslexia or ADHD (although this latter group is currently being recruited and studied as we write). While some of the cohort may follow neurodivergent pathways without being autistic, permitting some comparison to be made between atypical outcomes, larger numbers are needed to determine the specificity of findings to later autism diagnosis. At the time of writing, when long-term infant cohort studies are still relatively few and far between, this issue is compounded by the drive to publish early — many papers report merely on cross-sectional comparisons of high-likelihood versus low-likelihood groups rather than wait for longitudinal data on diagnostic outcome. Even when outcome data are available, these will normally have been collected at the earliest reasonable time point (often 36 months, but sometimes as young as 18 months) — raising the likely possibility that these children actually represent a sub-set of the constellation, with a specific combination of autistic features amenable to early diagnosis. We know that many autistic children don’t stand out until much later, and as yet, we have little data from later diagnosed participants in infant-sibling studies, which may relate to variability within the autism constellation.

The field also needs to meet some steep statistical challenges before a robust early marker can be identified. It is well known that any early marker must demonstrate adequate sensitivity and specificity in a study with a large enough sample size to limit confidence intervals to an acceptable range. But less widely acknowledged is the importance of a test being both reliable and straightforward to administer, and the fact that the predictive value of any marker is adjusted according to the baseline prevalence of a condition in the general population. The percentage of false positives might be low, but if the number of people who do not have a particular condition is high — in this case, 99/100 individuals at current best estimates of prevalence — even a low false positive rate could translate into a large number of people being mis- diagnosed. These complex statistical concepts have been superbly summarised and helpfully modelled online on the Spectrum blog (see “Further Reading”) and are represented in Figure 7.2.

3. The social orienting hypothesis

We have seen that there is still no robust, reliable early marker of autism, either cognitive or biological. However, there are a number of theoretical accounts that make predictions about where such a marker might be found. The first of these is the social orienting hypothesis, which predicts that the earliest signs of autism should be a lack of preferential attention (across all sensory domains) to social content in the world. This account supposes that a basic mechanism that prioritises attention to social content, starting with simple orienting to faces and voices, is disrupted in autism. The child attends less to social content, missing out on opportunities to learn about social behaviour and develop language. Soon, a lack of comprehension of the social world starts to reinforce the attentional difference — children may choose to spend time and attention on non-social activities, which are more comprehensible and therefore enjoyable. This developmental pathway results in the social interactional and communication differences that are characteristic of autism.

This hypothesis was originally based on two different sources of data. Dawson and colleagues showed that autistic children at young ages were less likely than typical and learning disabled children to orient to social sounds — their name being called and clapping (Dawson et al., 1998). In the visual domain, Klin used eye-tracking with a small group of autistic adults to demonstrate that autism was associated with differences in looking patterns (Klin et al., 2002). Specifically, the autistic group showed less fixation on the social regions (e.g. face, eyes) of a complex stimulus (excerpts from a movie).

Figure 7.2 Statistical challenges of identifying a biomarker for autism

Panel A shows results of one attempt to identify autism using markers of early brain development (Hazlett et al., 2017). Here the data look like they have fairly good levels of precision. In Panel B, we see how different the data look when the number of incomplete scans are taken into account — this gives us a lot less confidence in the practical potential of the marker. Panel C extrapolates the data according to the population base rate. We can see that, if applied in practice, this test would result in unacceptably high numbers of false positives: infants diagnosed with autism who are not autistic. Provided by kind permission of Jon Brock.

What about data from infancy? It is uncontested that infants (including non-human infants, even chicks!) show a preference for looking at faces, or face-like stimuli, from birth. In fact, recent data have revealed that this may even be in evidence pre-natally, as infants in the womb orient to face-like light patterns projected onto the mother’s body (Reid et al., 2017) — though replications are needed to confirm this preliminary finding. Turning to autism, Klin and colleagues have replicated their adult finding with toddlers, showing less attentional bias towards faces compared with typical and learning disabled groups (Chawarska et al., 2010; Jones et al., 2008). In a study with typically developing twins, they also demonstrated that such looking patterns appear to be under genetic control (Constantino et al., 2017).

So far, so good, but there are complications. There are a number of findings that suggest intact social attention and orienting among adults and children in the autism constellation, casting doubt on the social orienting hypothesis (see Guillon et al., 2014 for a review). Indeed, the British Autism Study of Infant Siblings found no differences associated with later autism diagnosis in looking patterns to social and non-social content in a static image array (Elsabbagh et al., 2013). There are a number of ways we might probe the apparent disparities in the literature. Sue and colleagues found subtle differences in the timing of eye-movements when viewing naturalistic, but static, social stimuli (Fletcher-Watson et al., 2009). Other studies have suggested that the looking preferences of autistic toddlers are dictated by episodes of audio-visual synchrony rather than social content (Klin et al., 2009), though this finding, too, is contested (Falck-Ytter et al., 2017). A problem with much of this literature is that stimulus complexity is not well captured or controlled. There is evidence that the greatest group differences arise when more ecologically valid stimuli are used (Frazier et al., 2017), but this raises the question of whether the ultimate cause of different looking patterns is specific to the social domain, or rather an information-processing difficulty. The use of video stimuli that feature cuts between camera angles may be particularly problematic when it comes to interpretation of data, since a delay in oculomotor movements or in information processing could be mistaken for a difference in top-down attentional preference (Ames & Fletcher-Watson, 2010).

One way to attempt to resolve these equivocal findings is to probe differences in attention developmentally, rather than cross-sectionally — mapping the timeline of specific behavioural patterns and their effects. One recent high-profile paper charted looking patterns in an infant-sibling group with repeated testing, using eye-tracking, over the first six months of life (Jones & Klin, 2013). The authors reported significant differences between children who were, and were not, later diagnosed with autism in the trajectory of looking patterns over time. However, taking each data collection point independently, there were no group differences in looking times. Moreover, at the earliest time points, infants who later received an autism diagnosis actually showed higher proportions of looking time to social content — though this pattern was non-significant. This finding, combined with evidence that domain-general attentional differences are predictive of autism outcome, led Mark Johnson to argue that it was time to reject the social orienting hypothesis of autism (Johnson, 2014).

4. The social motivation hypothesis

A subtly different account of the developmental cascade resulting in autism places the emphasis not on attentional biases, but on reward value. The theory has a very similar structure to the social orienting hypothesis. Again, an innate mechanism is proposed, making social information inherently rewarding to infants. The earliest sign of autism in this scenario would be a reduction or absence of this tendency to find social information rewarding. This early small shift could result in reduced engagement with the social world over time, resulting in missed opportunities to learn normative interaction skills, and ultimately delivering the social and communication features of autism. As in the case of social orienting, an early bias might be self- sustaining — lack of social engagement, resulting in reduced social understanding, would undermine the already limited reward value of social information or contact.

This hypothesis continues to rely largely on data from adults, indicating fundamental differences in neural response to social stimuli. One argument emphasised by this theory is that the brain regions normally dedicated to developing social expertise become re-purposed in autism. A striking case study comes from a young man with autism who showed a neurological specialisation for recognising Digimon characters in regions typically reserved for recognition of human faces (Grelotti et al., 2005). However, it’s worth noting that shepherds learn to recognise their sheep using similar brain regions, without anyone implying this shows reduced interest in people! There is also some limited data from infancy, for example, indicating a reduced EEG response to direct gaze in infants who later get a diagnosis (Elsabbagh et al., 2012) and atypical white-matter development in the same population (Wolff et al., 2012). However, these data do not specifically invoke a social motivation account, as they do not demonstrate specificity of disruption to brain regions underpinning reward. Nor indeed is it clear that there is any such thing as a ’social reward’ system — meaning something distinct from the general reward system that can be activated by a range of stimuli; social, monetary, food and so forth (Lin et al., 2011). That said, recent data do show that reward value (indexed by rapidity of learning) varies with the social versus non-social nature of the stimulus and also whether it is more or less engaging for adults and children with typical development — a replication in autism would be of interest (Vernetti et al., 2018).

This theory clearly requires more direct testing in relevant populations, but even with more data, it is likely to be challenging to distinguish the social motivation hypothesis from the social orienting hypothesis. It is possible, perhaps even likely, that social attentional and social reward differences interact reciprocally across infancy. However, disentangling these is essential if we wish to understand the best way to support autistic children. Take, for example, a non-speaking child — what is the best way to help them develop their communication skills? The social orienting hypothesis suggests that iterative teaching might be appropriate. They have missed out on a number of natural learning opportunities, due to their different attentional focus at a critical developmental stage, but explicit teaching might enable them to expand their repertoire of communication skills, resulting in greater autonomy in the future. In contrast, the social motivation hypothesis suggests that the child’s lack of speech may reflect a lack of motivation to engage with others. In this context, repeated skill teaching might have limited value — especially in a framework where success is rewarded with social feedback, like smiles, praise and high fives. Instead, we might wish to structure the learning around the child’s personal interests, while providing an alternative way to communicate, such as a voice-output communication aid.

A challenge to the social motivation hypothesis comes from the reports of autistic people themselves, as well as independent observations. Children’s attachment to parents/caregivers is not affected by autism (Teague et al., 2017). Many autistic adults show a high motivation to engage with others, albeit in ways that do not always adhere to social norms. This is not a new observation — Lorna Wing originally characterised such children as “active but odd”. The phenomenon is also apparent when one considers the high rates of self-reported ’camouflaging’ among autistic people (Lai et al., 2017). Why would someone who does not find social engagement rewarding put so much effort into their relationships with other people? Difficulties navigating the social world, where these occur, often seem to arise from differences in social interaction style rather than a lack of motivation. One answer to this apparent conundrum might lie in considering the developmental timeline. If autistic people experience a reduced neurobiological reward in response to social content in infancy, this could have an impact on their developmental trajectory in relation to quantity, and therefore quality, of social interaction. However, this argument does not also require that an older autistic child or adult would still be un-motivated by social engagement. Indeed, individual differences in the period of time during which social reward systems were depressed could theoretically go some way towards explaining the sub-groups labelled aloof, passive and odd by Lorna Wing. However, firm evidence that infants later diagnosed as autistic lack social motivation is still wanting, and the absence of overt behavioural differences in social response in the first year of life is a striking finding from infant-sibling studies.

Returning to our criteria for what makes a good theory (see Chapter 5), both the social orienting and social motivation account run the risk of explaining too much. The developmental cascade does not allow, in any clear way, for the uneven profile of social abilities in autism. While autistic children often find it difficult to spontaneously intuit what someone else is thinking, they do show attachment to loved ones, emotion recognition and emotional empathy (unless they have an additional problem of alexithymia), as well some of the less noble aspects of human social processing, such as gender and racial stereotypes (Hirschfeld et al., 2007).

5. Intersubjectivity accounts

A very early theoretical model of autism, proposed by Peter Hobson, maintained that autism is primarily rooted in an affective and interpersonal difference (Hobson & Lee, 1998, 1999). Differences in the ability to perceive and respond to the affective expressions of others were hypothesised to lead to atypical social experiences in infancy and childhood, which in turn impacted later social understanding.

In common with other developmental models, this account is not consistently supported by the data: findings on differences between autistic and non-autistic people in viewing, identifying and responding to emotional stimuli are mixed at best. There are large individual differences between autistic people, and results change with age, intellectual level and specific study design (Uljarevic & Hamilton, 2013). Another source of evidence that may be relevant to intersubjectivity accounts of autism comes from studies of imitation. Like joint attention, imitation is thought to be a pivotal developmental skill underpinning explicit teaching and playing a role in forging social bonds. But, like emotion recognition, while imitation is frequently reported to be atypical in autism, methods and findings vary widely, and there are substantial individual differences (Vanvuchelen et al., 2011).

A significant question arising is whether we should interpret a lack of intersubjective identification between autistic people and non-autistic experimenters as a difference associated with autism, or as a typical manifestation of in-group effects? In Chapter 6, we discussed how some findings relating to imitation might plausibly be attributed to in-group and out-group status, rather than being autism specific. The same possibility applies to our interpretations of the emotion recognition literature. For example, Sheppard and colleagues created emotional stimuli by placing autistic and neurotypical people in one of four different social scenarios (Sheppard et al., 2016), such as being ignored or being told a joke. Their reactions were filmed and later these films were shown to non-autistic raters, who were asked to match the reactions to the scenarios. The results showed that raters struggled to match emotional responses of autistic people with scenarios, even though they rated those responses as equally expressive. In other words, this study reports a neurotypical deficit in recognising and interpreting the emotional expressions of autistic people. This echoes the work of Brewer and colleagues (2017, and see Chapter 6), which also demonstrated a mismatch between neurotypical and autistic expressions of emotional states — though in this case both autistic and non-autistic participants found autistic facial expressions harder to read. Moreover, Komeda and colleagues (2015) demonstrated that autistic people show a stronger neural signal of empathy towards autistic characters than to non-autistic characters — and the opposite pattern was true for neurotypical participants. These findings suggest that we might explain some of the behaviours considered to be distinctive features of autism as, instead, a typical manifestation of the tendency to more readily and deeply engage with people from the same social group. This finding corresponds with some new social theories of autism, as well as with reports from autistic people, and we will return to this theme in more depth in Chapter 9.

6. Early development outside the social domain

All three of the theoretical accounts presented here attempt to explain autism by identifying an early, subtle difference in the way that autistic infants relate to the social world. However, some of the most robust early markers linked to later diagnosis are in the non-social domain. For example, the gap- overlap task is a measure of the ability of an infant to switch their attention from a central target which has just disappeared, to a peripheral target which has just appeared. It also measures, via the ’overlap’ condition, how this ability is impacted when the peripheral target comes on while the central one is still present, and thus disengagement is required as well as switching. The association between gap-overlap task performance and later autism diagnosis has been replicated in different cohorts, though sensitivity and specificity of this marker has yet to be established (Jones et al., 2014).

Another domain receiving increasing attention is motor development. Autism in children and adults is associated with a number of distinctive gross and fine motor features including tip-toe walking, bouncy gait, dyspraxia and a range of repetitive behaviours in the motor domain (e.g. hand-flapping, body rocking, finger twiddling) often known colloquially as ’stimming’. Recently, early motor differences have also been detected at the level of eye-movements, with shorter mean fixation lengths in infancy being associated with later diagnosis (Wass et al., 2015). Micro-movements of the head, for example, as captured during neuroimaging, are being analysed in autism and other groups (Torres & Denisova, 2016). An innovative tablet-based study shows how distinctive motor movements in autism might be identifiable from game-play using machine learning (Anzulewicz et al., 2016). It is too early to draw firm conclusions from this literature, which is currently lacking large-scale and robust longitudinal data, but certainly the importance of early motor development should not be ignored in our efforts to understand autism.

How do we interpret these findings? One possibility is that the early features of autism are not social-specific. This feeds into a wider debate in development psychology on how skills required in the social domain relate to domain-general skills — we touched upon this in the review of executive function in autism in Chapter 6. On one hand, we might hypothesise that the social world permits intensive rehearsal of key domain-general switching (and other executive, or motor) skills. For example, attention switching and disengagement are required when rapidly moving your focus between your father’s face and the rattle he is shaking. Perhaps frequent exposure to this sort of situation acts as a ’training ground’ for the attention system. An infant without a great deal of interest in that situation, due to social attention, social motivation or intersubjective differences, might miss out on the chance to build up their attention skills. Alternatively, the causal direction could be the opposite. A child without an efficient attentional switching system might find that impedes their social interactions, leading over time to a divergent path in terms of social interactive behaviour. Another factor concerns the behaviour of the parent, of course. In the example noted earlier, if the child looks at the rattle but not her father’s face, how will her father react? His behaviour will be shaped by hers, and vice versa. Both people in the dyad create a shared environment which may promote certain types of skills and behaviours more than others.

A third explanation also exists — that both social interactive behaviours and domain-general skills are subsumed by a shared mechanism or brain region which is the source of the difference. Some have proposed that self-other switching, that is switching between representations to do with the self (own emotions, beliefs, motor intentions, etc.) and those to do with others, might be a candidate mechanism in the development of autism (de Guzman et al., 2016; Sowden & Shah, 2014). Bedford and colleagues have further shown that a combination of markers of domain-general attention with social neuro- response plus direct observation measures provide the most robust early identifier linked to later autism diagnosis at 7 years old (Bedford et al., 2017). Their data are interpreted as reflecting an additive mechanism, resulting in a predictor of autism that relies on the presence of separately measured cognitive capacities that combine to increase likelihood.

One final question is whether determining the early causal associations between different features of autism is really necessary? If interactive scenarios, like playing with a parent and a rattle, involve both social and non-social skills, we can provide opportunities to develop both by simply creating more such opportunities in a suitably enjoyable and motivating way. A recent attempt to do just that found that early, parent-delivered, play-based intervention led to increased attention to parents, with longer-term positive effects on parental directiveness and child communication (Green et al., 2015, 2017). Crucially, the authors emphasise “a strategy to mitigate developmental risks and modify prodromal symptom trajectories, rather than ’eliminate’ a condition” (2017, p. 1330, emphasis in original). Thus, while such early interventions require a robust ethical framework, they may have the potential — by delivering support in a critical window of development and enhancing key skills — to guide parents to tailor their behaviour, creating an optimal environment for the autistic child. This in turn could enable a higher proportion of autistic children to achieve greater autonomy and self-determination in the future.

7. Lifespan development in autism: beyond childhood

Theoretical accounts of autism reviewed here focus exclusively on early development in infancy but of course important developmental changes continue across the lifespan. Here we briefly mention two other key developmental stages: puberty/adolescence and old age (discussed at the behavioural level in Chapter 3). These are selected not because other life stages are unimportant — starting school, transitions from school to adulthood, finding a life partner and parenthood are all major life changes — but because these two stages are specifically associated with cognitive changes. The review is brief, again, not because these life stages are unimportant but because the published literature here is very limited.

At puberty, children experience changes across the biological, cognitive and behavioural levels. Biologically, there are well-known hormonal effects, but also neurological developments (Blakemore & Choudhury, 2006). Behavioural changes include transitions of school setting around the time of puberty, with new expectations, including the need to monitor a more complex timetable, maintain concentration for longer periods in class and manage independent study. Moreover, at this time, there are changing expectations of appropriate behaviour with regard to play, friendship and romantic relationships; for neurotypical teenagers, family often becomes less important and peer group relations much more influential. Occurrences of diagnosis of autism in later childhood and adolescence may be related to these contextual factors. As new, adult-like societal norms start to be applied, young people with autism may find their coping techniques stretched beyond capacity, revealing differences that were previously masked or accommodated. Qualitative work has revealed, for example, high rates of loneliness reported by autistic teenage boys (Lasgaard et al., 2010) and difficulties navigating sexual development (Dewinter et al., 2017). The challenges of the latter domain may be exacerbated by the high-prevalence of non-normative gender identification and sexual orientation among autistic people, and by the potential sensory impact of bodily changes. Adolescence is also a risky time in terms of mental health; social anxiety, eating disorders and depression often arise in the teens for neurotypical young people, and autistic young people are at higher risk for all of these (Simonoff et al., 2008).

For autistic adolescents who have significant support needs (perhaps due to learning disability and/or language disorder) and those who support them, this stage of life may produce different pressures. As non-disabled peers start to become more independent, parents and siblings can feel the stressful impact of having an autistic family member more acutely (Tsai et al., 2017). Hormonal changes may be related to increased frequency of self-harm or violence, made more challenging by increased size and strength. Beyond adolescence, as children grow into adults but support needs remain high, parents can start to worry about their child’s long-term care and well-being (Griffith et al., 2012).

Some studies report on so-called optimal outcome where children with an early diagnosis no longer meet diagnostic criteria on standardised instruments later in life (Fein et al., 2013). However, there is some evidence that those individuals do still experience significant mental health difficulties, raising concerns that apparent change in diagnostic status is due to effortful compensation (Livingston & Happé, 2017) rather than a genuine change in underlying differences. Recent attention to risk of suicidal behaviour in autism has revealed that feelings of “not belonging” and being a burden to others are among the factors contributing to suicidality in autistic people (Pelton & Cassidy, 2017). In that context, further research is needed to characterise so-called optimal outcome and the autistic experience of intervention, and to determine whether it is a legitimate goal for early supports.

Biological, cognitive and behavioural changes associated with ageing are also well established in the general population (Hedden & Gabrieli, 2004). Behaviourally, people may retire from full-time work, leading to opportunities for travel, hobbies and voluntary work, but also a loss of daily structure and work-based social networks. There may also be a need to care for grandchildren, a spouse or sibling and even, as life expectancies increase, a parent. Retirement can also put pressure on social relationships including life partnerships, as couples begin to spend more time together after many years of working outside the home. A number of adults coming for first diagnosis of autism in their 60s and 70s have long-standing autism traits that have not altered, rather the scaffolding and enabling environment may have been removed if an enjoyed and well-structured job ends, or an understanding and supportive partner dies.

Cognitively, as well as the risk of dementia, ageing is associated with normal declines in memory, reaction time and other fundamental cognitive skills. We know shockingly little about physical and mental changes in autism associated with ageing (Happé & Charlton, 2012). Follow-up of children diagnosed by the narrow autism criteria of the 1960s or ’70s, show continued need for substantial support, and even when social difficulties or repetitive behaviours are rated as having reduced, quality of life is disappointingly low (Howlin et al., 2014). Studies in the states suggest these adults have higher rates of almost all health problems (Croen et al., 2015). However, today’s children are diagnosed according to far broader criteria, and we hope they are being raised in a more supportive society — there is reason to be optimistic that their life journeys will be far more positive. A study of adults coming for first diagnosis of autism found that older adults reported more autistic traits than younger adults, but did somewhat better on some neuropsychological tasks (Happé et al., 2016). The first cross-sectional group studies of older autistic adults also suggest that age-related declines in some cognitive functions may be less steep in autism than non-autism groups (Lever & Geurts, 2016), although longitudinal studies are necessary to rule out cohort or selection effects.

A research focus on early development makes sense in regard to some scientific goals, such as understanding the developmental causal associations between different cognitive domains and key behaviours. Although there is no single pot of autism research funding that must be shared out between projects and disciplines, investment in such studies necessarily limits what is being done to provide evidence-based support for autistic people at other life stages. Given that aiming to identify a cure or preventative measure for autism itself represents a moral bankruptcy (even if improving communication or preventing epilepsy is desirable), we know that there will always be autistic people in our world. In which case, improving and increasing research into lifespan issues must be an important priority for the future.

8. Current debates


A group of theories arose, alongside and following attempts to characterise autism in relation to a fixed deficit in the social cognitive domain, which adopted a more developmental perspective on autism. We characterise them here as social attention, social motivation and intersubjectivity accounts. These theories attempt to capture small, early adjustments to cognitive processes, which set the child on a developmental pathway resulting in large differences at the age when diagnosis becomes possible. The developmentally increasing differences between the autistic and non-autistic child may be accompanied by differences in expectations from adults and peers, which contribute to the identification of specific behaviours as ’symptoms’ of autism. The search for consistent and specific early signs of autism continues but faces major statistical and methodological challenges. Meanwhile, autism research focusing on other life stages, such as old age, is in its infancy.

Big questions

The early autism research literature is beset by specific ethical issues (Fletcher-Watson et al., 2017b). It is essential to embed early autism research in a framework of engagement with community priorities. How will autistic children enrolled in longitudinal or early intervention studies feel when they grow up? If we find a reliable early marker of autism, what should we do with that information? Are justifications for early intervention before diagnosis legitimate, and if so, where do the boundaries lie between a support that enables autistic children to achieve their goals and a support that serves merely to suppress or ’normalise’ the autistic child?

Parallel with this debate is a concern regarding the ethics of intensive monitoring or early interventions in high-likelihood populations for those who do not go on to get an autism diagnosis. Are infant-sibling studies inadvertently changing parenting in enrolled families, and if so, how? What message do such studies send about the need to change the development of autistic children, and the acceptable costs of doing so? Rather than aiming for autism-specific early interventions, should we instead focus on ’generic’ supports that empower parents, enrich the early years environment and focus on global targets like school readiness and language?

The pressure for early findings from longitudinal studies may result in results that reflect only part of the autism constellation; will infant-sibling study results be very different once we know the outcome of the children at 8 or 18 years? How far do any such findings generalise to the autism constellation as a whole?

One final question for this field concerns the definition of “social” in the context of social orienting, social motivation and so on. Frequently, studies rely on looking patterns to static faces or videos of a single person talking or singing. In particular, looking at the eye region is interpreted as social, while looking at the mouth region is not. There is a risk here of a tautological argument, whereby any looking pattern exhibited by a child identified as autistic is automatically characterised as non-social, or even inferior. Researchers in this field need to be vigilant to ensure that results are interpreted without recourse to such logical fallacies and that group differences are recognised as differences, rather than being automatically interpreted through a set of normative values.


Could autism be a result of differences in development over time? As an autistic adult, working nationally as an adviser, this is a subject for personal reflection. Some of the theories of autism focus on the idea that autistic people are broken versions of other people. For example, suggesting that we’re bad at paying attention to social signalling from any others. In reality, I and my autistic colleagues find that we’re pretty excellent at paying attention to autistic social signalling and fairly clueless about how to interpret social signalling of non-autistic people. Likewise, non-autistic people seem to struggle to interpret us correctly. It appears to be a difference in communication, not a deficit. From my perspective, I find time with other autistic people socially rewarding. I am also very keen to attempt to socialise with non-autistic people, but am frequently met with a baffling response from them. A different communication system at work, rather than a deficit? Similarly, the idea that we struggle with empathy is a tricky one. I find it generally easy to spot the emotional state of autistic people and empathise accordingly. But, I struggle to interpret the signalling of non-autistic individuals, so take extra time to process that. By the time I’ve done so, and responded, they’re often already offended.

The question of how autistic people develop different skills over time is one we may not be close to answering, given the current lack of good research into adults. As a child, I was functionally non-verbal for the first ten years, rocking, flapping, spinning the wheels on toy cars. I was without doubt very autistic. I could sometimes speak words, but had no concept of their meaning. I could sometimes then speak phrases, with similar lack of clue as to what I was saying. Often, I could not speak at all, no matter how much effort I made, especially in any situation of sensory/social overload. Endless effort on my part has resulted in the communication skills we now see; I speak at conferences, study at the master’s level, run a company. And yet words are still not my natural language, and I still fail to use them well in any stressful situation. Spoken words remain exhausting, whereas technology has been hugely liberating. I’m the same person. Am I ’broken’ or not? I believe that autism is a difference and one that often has significant strengths for society.

Recommended reading

Beardon, L. (2017). Autism and Asperger syndrome in adults. London: Sheldon Press.

Blakemore, S. J. (2018). Inventing ourselves: The secret life of the teenage brain. New York, NY: Doubleday.

Fletcher-Watson, S., Apicella, F., Auyeung, B., Beranova, S., Bonnet-Brilhault, F., Canal-Bedia, R., … Farroni, T. (2017b). Attitudes of the autism community to early autism research. Autism, 21(1), 61—74.

Happé, F., & Charlton, R. A. (2012). Aging in autism spectrum disorders: A mini-review. Gerontology, 58(1), 70—78.

Johnson, M. H. (2014). Autism: Demise of the innate social orienting hypothesis. Current Biology, 24(1), R30—R31.

Jones, E. J., Gliga, T., Bedford, R., Charman, T., & Johnson, M. H. (2014). Developmental pathways to autism: A review of prospective studies of infants at risk. Neuroscience & Biobehavioral Reviews, 39, 1—33.