PFI_12mmX178mm.pdf + eps format

Journal of Child Psychology and Psychiatry 61:3 (2020), pp 218–232 doi:10.1111/jcpp.13176

Annual Research Review: Looking back to look forward – changes in the concept of autism and implications for future research

Francesca Happe, 1 and Uta Frith2 1Social, Genetic and Developmental Psychiatry Centre, Institute of Psychiatry, Psychology and Neuroscience, King’s College London, London, UK; 2Institute of Cognitive Neuroscience, University College London, London, UK

The concept of autism is a significant contribution from child psychiatry that has entered wider culture and public consciousness, and has evolved significantly over the last four decades. Taking a rather personal retrospective, reflecting on our own time in autism research, this review explores changes in the concept of autism and the implications of these for future research. We focus on seven major changes in how autism is thought of, operationalised, and recognised: (1) from a narrow definition to wide diagnostic criteria; (2) from a rare to a relatively common condition, although probably still under-recognised in women; (3) from something affecting children, to a lifelong condition; (4) from something discreet and distinct, to a dimensional view; (5) from one thing to many ‘autisms’, and a compound or ‘fractionable’ condition; (6) from a focus on ‘pure’ autism, to recognition that complexity and comorbidity is the norm; and finally, (7) from conceptualising autism purely as a ‘developmental disorder’, to recognising a neurodiversity perspective, operationalised in participatory research models. We conclude with some challenges for the field and suggestions for areas currently neglected in autism research. Keywords: Autism spectrum disorders; Asperger disorder; neurodevelopmental disorders; social cognition.

Association’s 3rd edition of the Diagnostic and Sta- Introduction tistical Manual (DSM-III; APA, 1980), which included Just over half a century after autism was first named ‘Infantile Autism’ as a separate diagnosis for the first and described, there has been a recent explosion of time, listed as one of six criteria, ‘Pervasive lack of interest in the history of the diagnosis. This is responsiveness to other people’. In contrast the manifest notably in books addressing a general equivalent criterion for ‘Autism Spectrum Disorder’ audience, such as those by Grinker (2008), Feinstein (ASD) in the latest edition of DSM-5 (APA, 2013), (2011), Silberman (2015), Donvan and Zucker ‘Persistent deficits in social communication and social (2017), and Evans (2017). Given the ready availabil- interaction across multiple contexts...’, can be met in ity of historical accounts covering a variety of social a wide range of ways. Lorna Wing’s suggested autistic and clinical aspects, here we draw on our personal social typology of ‘aloof, passive, and active-but-odd’ experience of the changing face of autism research can be seen as an early attempt to move from ‘lack of over the last 30 years. responsiveness’ to social difficulties that varied widely Without looking back, one might imagine that in manifestation (Wing & Gould, 1979). However, the autism has always been conceptualised as it is perception remained in the 1980s that a child who today, a lifelong neurodevelopmental condition with was, for example, over-friendly was not autistic. a spectrum of manifestations and high rates of co- Two more of DSM-III’s six diagnostic criteria for occurring mental health difficulties. However, look- Infantile Autism were ‘Gross deficits in language ing back over the past decades of autism research development’ and ‘If speech is present, peculiar reveals dramatic reconceptualisations with far- speech patterns such as immediate and delayed reaching implications, both theoretical and practi- echolalia, metaphorical language, pronominal rever- cal. We focus below on seven major changes in the sal’. Note the focus on language rather than commu- conception of autism and in each section attempt to nication, and expectation that many autistic children draw out the implications for future research. would show no speech. Language disorder was seen as central to autism, and perhaps even explanatory for the difficulty in relating to people and objects in The changing concept of autism: 1. From the usual way. Much attention continued to be paid narrow to wide to delayed and atypical speech in autism through the In the 1980s, when psychological research started to early 1980s such that, in 1989, Frith reflected that, intensify, the concept of autism was far narrower than ‘More has been written on the language of autistic notions today. For example, the American Psychiatric children – the peculiar forms of their speech as well as their difficulties in comprehension – than any other of Read the Commentary on this article at doi: 10.1111/jcpp. their psychological disabilities’ (p. 20). 13212 Against this background, researchers and clini- Conflict of interest statement: No conflicts declared. cians were surprised when some children acquired

© 2020 Association for Child and Adolescent Mental Health Published by John Wiley & Sons Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main St, Malden, MA 02148, USA doi:10.1111/jcpp.13176 Changes in the concept of autism 219 useful and even fluent language and yet remained theory – although much debated – has been so autistic. Hans Asperger (1944) had described such influential in the field that it is easy to forget that cases, who were otherwise very similar to those prior to this specific account, social difficulties in described by Kanner (1943). Indeed, this is one of the autism were generally rather amorphously charac- main reasons why Asperger’s syndrome was eagerly terised as a lack of sociability or interest in others. taken up after Wing’s 1981 paper and Frith’s English The precise ToM hypothesis allowed two important translation of his work in 1991, first informally areas of progress in research: neuroimaging investi- amongst clinicians, and then formally in ICD-10 gations of the neural underpinnings of key social (WHO, 1990) and DSM-IV (APA, 1994). Added as a processing differences in autism, and the delineation ‘disorder of uncertain nosological validity’, Asperger of intact social abilities in autism, including emo- syndrome in ICD-10 (and Asperger disorder in DSM- tional empathy (see Happe, 2015 for a wider review IV) differed from autism only ‘in the fact that there is no of this topic). general delay or retardation in language or in cognitive development’. This new category gave rise to a dra- From narrow to wide: implications for future matic change in research attention. Almost overnight, research researchers were galvanised to find out more about a group of autistic children who puzzlingly showed no An unintended consequence of the focus on ‘pure’ language delay at all. Many did not come to clinical autism has been the neglect of intellectual disability attention in early childhood, because they spoke well and developmental language disorder in recent and did not have intellectual impairment. Up until this research. Russell et al. (2019), examining autism time, the majority of participants in autism research research papers published in 2016, covering 301 had intellectual disability/language impairments; studies and more than 100,000 participants, esti- through the 1990s that began to change, so that the mated that 94% of all ASD participants had IQ in the vast majority of studies came to be carried out with average range, and 80% of studies showed selection verbal participants of average or above average IQ. bias against participants with intellectual disability. The focus primarily on autism without accompany- The authors contrast this with an estimated 50%– ing intellectual or language disability could be partly 55% of the autistic population that have intellectual attributed to simply practical issues: research with disability. This selection bias towards verbal and this group is far easier than devising tasks accessible intellectually able autistic participants obviously to minimally verbal individuals. However, we believe threatens the generalisability of research findings, this focus must also be seen in the historical context of but will also undoubtedly have consequences for the autism’s relative recency as a diagnostic category, in direction of future theories, as well as shaping public addition to the more general move in psychiatry to a understanding of autism. multiaxial classification system. In the last decades of Minimally verbal participants are rarely involved the 20th century, there was still a need to show that in research studies, and this is particularly obvious studying autism as a developmental condition, dis- in brain imaging studies (Jack & Pelphrey, 2017). tinct from intellectual disability, language disorder Yet, arguably, we have never had more opportunity and so forth, was valid. Showing that psychological to assess and investigate the cognitive processes of differences could be demonstrated in individuals with these ‘hard to reach’ groups, thanks to rapid ‘pure’ autism (i.e. without the confound of intellectual advances in technology. Wearable sensors and EEG impairment and/or language disorder) was part of allow monitoring of physiological and neural mark- that effort. Where Hermelin and O’Connor (1970) had ers unobtrusively. Functional near-infrared spec- introduced the practice of comparing autistic groups troscopy (fNIRS) provides opportunities for with IQ or mental age-matched comparison (nonautis- measuring neural activity in naturalistic contexts tic) groups (often intellectually disabled, but some- (Pinti et al., 2018). Virtual reality brings the outside times with, e.g., sensory disabilities), now researchers world into the laboratory, for controllable, replicable focused on so-called ‘high-functioning’ autism and yet realistic encounters. Eye-tracking technology Asperger syndrome in order to understand what part gives a read-out of interest and attention without of the behavioural and cognitive presentation of requiring verbal instruction or response. And the autism was distinct from intellectual and/or language pervasive availability of tablets and smartphones, disability. with touchscreen response and intuitive app devel- While the behavioural diagnostic criteria widened, opment, means that most young people know how to it should be noted perhaps that the psychological interact with gamified tests, from their earliest years. characterisation of autism became significantly As work by Laurie et al (2019) shows, autistic more focused in the 1980s. The ‘Theory of Mind’ children and adults across the full IQ range, interact (ToM) deficit hypothesis proposed that failure in the with technology for much the same purposes, and ability to (meta)represent mental states in oneself with much the same preferences, as their neurotyp- and others is the cognitive cause of the characteristic ical counterparts. These new technologies should autistic behavioural difficulties in social interaction lead to a revolution in research and herald the and reciprocal communication (Frith, 1989). This inclusion of intellectually impaired and minimally

© 2020 Association for Child and Adolescent Mental Health 220 Francesca Happe and Uta Frith J Child Psychol Psychiatr 2020; 61(3): 218–32 verbal children and adults in studies (e.g. Tager- estimates from different survey methodologies and Flusberg, et al., 2017). their likely impact on apparent rise in numbers have been eloquently discussed by Fombonne (2018). In parallel with the rise in numbers of autism The changing concept of autism: 2. From rare diagnoses, there has been a striking rise in research to common interest in autism over the last 30 years. In 1988, The first autism prevalence estimate, by Lotter when the first author began her PhD, keeping abreast (1966), was approximately 4 in 10,000, and Wing, of published papers (hard copy only, of course) on who introduced the notion of the ‘autism spectrum’, autism was easy; 190 publications that year included gave an estimate of 22 in 10,000 from her epidemi- ‘autis*’ in the title, and there were no more than 2,600 ological study of children known to special educa- such publications in total across all previous years. A tional needs services in Camberwell (Wing & Gould, Web of Science search for the topic of ‘autis*’ in 2018 1979). Even in the 1980s, autism was considered to returned more than 6,000 publications, and the total be a very rare condition. More recently, the median of corpus of autism papers at the time of writing exceeds prevalence estimates of ASD worldwide was given as 68,000. This publication explosion reflects, amongst 62/10,000 (Elsabbagh et al., 2012). A ballpark other things, increases in research funding for autism. figure of 1 in 100 is widely accepted today, with In the United States, this was fuelled in part by some estimates being even higher. influential parents lobbying for improved services There was a widespread sense, in the 1980s, that and, in some cases, starting charities that specifically autism was underdiagnosed, and more specialist sought to bring into the field scientists who had diagnostic services were needed. Today, some authors previously had no interest in autism. In Europe, too, have suggested that autism may be overdiagnosed. autism research has received some major funding For example, Gillberg and colleagues reported no awards in recent years, most notably EU-AIMS (Aut- increase in levels of (parent-reported) autistic traits ism Innovative Medicine Studies), funded by the (at 9 or 12 years) in a Swedish population-based European Commission for c.£30 million, and the sample of children born from 1993 to 2003, but a follow-on AIMS-2-TRIALS receiving c. £55 million from significant increase in the rates of autism diagnosis the European Union plus matched charity and indus- over this 10-year period (Lundstrom€ et al., 2015). They try contributions in kind, currently the largest single also found a decrease in level of parent-reported autism grant in the world (Loth et al., 2017). autism (and related) symptoms and impairment for children receiving an autism diagnosis at age 7– No longer rare, but still underdiagnosed in females? 13 years, over the period 2004–2014 (Arvidsson et al., 2018). The authors suggest the bar for diagnosis In the 1980s, the estimated male:female ratio in is now significantly lower – perhaps too low. autism was 4:1, rising to perhaps 10:1 for those with The huge rise in numbers of autism diagnoses has Asperger syndrome. The rule of thumb then was that led some groups to speculate on possible environmen- females might be more rarely affected by autism but, tal factors; the change was thought to be too rapid to when affected, were ‘hit harder’ (e.g. had autism plus reflect genetic factors (although secular changes, such intellectual disability). This pattern led to the as older parenthood, could in principle have some hypothesis of a ‘female protective effect’; some aspect small impact; Wu et al., 2017). However, despite the of biology that meant that females required a higher US Centre for Disease Control (CDC) reporting a 78% aetiological ‘load’ in order to manifest autism. Evi- increase in autism prevalence rate (in 8-year-olds, dence in support of the female protective effect for from record review) from 2004 to 2008, there is reason autism has been reported; siblings (male or female) to doubt whether actual incidence of autism has of autistic girls show higher rates of autistic traits/ increased. American state-to-state differences in diagnosed autism than siblings of autistic boys (e.g. prevalence and rise in numbers are striking and Robinson et al., 2013). Some molecular genetic highlight the wide range of factors impacting diagnosis studies find evidence of higher rates of relevant rates (Sheldrick & Carter, 2018). Diagnostic substi- mutations (e.g. deleterious autosomal copy-number tution has been suggested; as autism diagnoses have variants; CNVs) in autistic females versus males – increased, a parallel drop can be seen in diagnoses of even after adjusting for IQ differences (e.g. Jacque- intellectual disability, in some places. As discussed mont et al., 2014). The high male preponderance has above, the widening of diagnostic criteria and intro- also led to a number of specific aetiological theories duction of Asperger’s Disorder and ‘Pervasive Devel- about autism, notably Baron-Cohen’s ‘extreme male opmental Disorder – Not Otherwise Specified’ in the brain’ account and work on prenatal androgens and DSM-IV (in 1994) undoubtedly affected prevalence. their putative association with social and nonsocial Taylor et al.’s (2013) analysis of information from the autism-related traits (e.g. Baron-Cohen et al., 2019; UK GP research database found a fivefold increase in for debate, see Ridley, 2019). the annual incidence rates of autism during the 1990s However, there is also the possibility that the lower in the United Kingdom, but that rates had held steady numbers of females, and especially of females with since. The many factors affecting prevalence good intellectual and language skills, might reflect

© 2020 Association for Child and Adolescent Mental Health doi:10.1111/jcpp.13176 Changes in the concept of autism 221 poor recognition of autism in these groups. Older give spurious false-positive and as well as false- estimates of the male:female ratio were largely based negative findings. on numbers from clinical or special educational needs An emphasis on increasing statistical power, and (SEN) registers. More recent studies, with epidemio- replication, means a move towards collaboration and logical samples and using active ascertainment, find consortia. While we would like to think that a single considerably lower male preponderance. A recent researcher, with little funding, can still make a meta-analysis by Loomes et al. (2017) estimated a conceptual breakthrough supported by small scale male:female ratio of 3:1 and importantly found that results from a cleverly designed and low tech experi- this did not differ greatly by intellectual dis/ability. If ment – this is clearly not the current trend. Consortia current diagnostic methods and criteria are male- have proved vital for genetic research, to collect the biased, this ratio may still be somewhat inflated. hundreds of thousands of samples required. The The discrepancy between estimates from passive AIMS-2-TRIALS network has collected cognitive data versus active ascertainment suggests that we have from hundreds of participants across several sites been missing or misdiagnosing large numbers of across Europe, alongside potential biomarkers (Loth autistic women and girls. Evidence is accumulating & Evans, 2019). Consortia require agreed shared that females are diagnosed later than males, and protocols and standardised tasks to pool psycholog- require higher symptom expression for diagnosis ical data. Currently, many of the measures routinely (see Carpenter, Happe & Egerton, 2019 for a fuller used in psychological research on autism lack good treatment of this topic). Diagnostic overshadowing psychometric data/properties, and this must change. probably also plays a part in under-recognition of Future research will also make use of ‘big data’ autism in women. For example, where eating disor- from registers and record linkage of routine data (e.g. ders are the presenting problem, clinicians may not electronic health and education records). These sorts consider an autism diagnosis. Sedgewick et al. of big data have typically been ‘shallow’, for example (2019) reported that at least 20% of women with noting diagnoses, hospital admissions or medica- anorexia passed cut-offs for an autism diagnosis, tions prescribed. However, technology for ‘deep phe- and elevated rates were still found in women who notyping’ could allow detailed, personal data to be had recovered from anorexia (so were not merely due collected at scale (e.g. from mobile phones). Ethical to very low weight). concerns, and fears about privacy, have to be Why might autism be overlooked in females? There addressed and may require a shift in public atti- are several reasons. First, research has often excluded tudes. Attitudes within the research community also female participants, and thus, research evidence need to change; there is the challenge of how to disproportionately reflects male autism. This research reward collaboration and data sharing, especially for forms the basis for diagnostic criteria and tools and early career researchers trying to make their names has historically shaped our notion of autism. Second, within very large consortia. there is a general bias to think of autism as predom- Regarding autistic females, researchers have inantly a male condition; thus parents, teachers and begun to study how some autistic women (and clinicians may not think of autism when they see a girl men) modify their behaviour to fit in and pass as who is struggling socially. Third, autism may look ‘neurotypical’ through camouflaging and compensa- different in females. There is as yet little robust tion (Hull et al., 2019; Livingston & Happe, 2017). research to tell us whether/how autism looks different These concepts can be operationalised as the dis- in females, in part because of reliance on diagnosed crepancy between ‘external’ levels of autistic traits samples who by definition meet current criteria (see (e.g. social behaviour as measured by the Autism discussion in Carpenter et al., 2019). Diagnostic Observation Schedule, ADOS; Lord et al., 1989) and ‘internal’ status (e.g. as self-rated on the Autism Spectrum Quotient, AQ, or as tapped by From rare to common: implications for future performance on ToM tests). The neural basis of research camouflaging, and possible sex differences in this, In the 1980s, when autism was still considered rare, are just beginning to be explored (Lai et al., 2018). very small sample sizes were acceptable in research. One clear implication of the past under-represen- Now, with much higher prevalence estimates, far tation of females in research samples is that future larger scale studies are possible. This coincides with studies must include all genders. Empirical research increasing recognition of the importance of replica- is needed to identify whether current diagnostic tion and large sample sizes. In the 1980s and even criteria and processes are gender-fair, or how to 90s, many researchers believed that our small make them so. Until they are, it is important to avoid sample sizes might prevent us from obtaining a the circularity of including only volunteers with an statistically significant result, but that where we did existing diagnosis; population-based samples, find significant group differences, these were likely to actively ascertained, are needed e.g. to establish be valid. A better understanding of the meaning of whether the female protective effect applies specifi- statistical significance has led to wider awareness, cally to autism or across all neurodevelopmental more recently, that small sample sizes are likely to conditions. Such research should also consider

© 2020 Association for Child and Adolescent Mental Health 222 Francesca Happe and Uta Frith J Child Psychol Psychiatr 2020; 61(3): 218–32 previously neglected topics of specific relevance to suffer high rates of mental and physical problems females, such as the autistic experience of adoles- (Croen et al., 2015). Using population-wide data cence, pregnancy, motherhood and menopause. from the 2011 Scottish census, for example, Ryd- zewska et al. (2018) found that 47% of autistic adults (identified as 0.2% of the total population) had poor The changing concept of autism: 3. From general health, about twice the rate in the general childhood to life span population, and that older adults and women had Why have we been so slow to realise that most especially elevated rates of ill health. autistic people are adults? Kanner and Asperger’s first descriptions were, of course, of children – and as From childhood to life span: implications for future child psychiatrist and paediatrician, respectively, research the term ‘Infantile Autism’ reflected their disciplines. Autism continued to be the preserve of child psychi- So little research exists addressing ageing in autism atry in both research and clinical practice for that the implication for future research is rather decades. It was not until 1971 that Kanner pub- simple; the developmental trajectory of later life must lished his 30-year follow-up of his 11 original cases be studied in autism with the same rigour applied to in adulthood. Even in the 1980s, autistic adults were child, and recently prospective infant, studies. Does largely invisible to research (with some notable the characteristic cognitive profile (e.g. strength in exceptions, e.g. Lund & Jensen, 1989; Schopler & detail-focused tasks, impairment in specific sociocog- Mesibov, 1983). As described above, the majority of nitive tasks) change with age in autism? Preliminary autistic people also had intellectual disability, and findings of preserved skills in some areas, relative to most adults would be in institutions or group homes. neurotypical ageing, are intriguing, but require fur- As the diagnosis widened in the 1990s to include ther study (Lever & Geurts, 2016; Zivrali-Yarar, 2017). Asperger syndrome, so awareness of adults rapidly The variability in life trajectories for autistic adults increased. Digby Tantam was one of the few adult is huge. The traditional markers of ‘good outcome’ psychiatrists interested in autistic adults (and, specif- (living independently, having a job and friends) are ically, similarities or differences versus ‘schizotypal’ increasingly questioned, with greater recognition adults; Tantam, 1988), and provided case studies that that quality of life needs to be defined in broader likely encouraged interest from other adult psychia- terms through consultation with autistic people and trists. Descriptions of highly intelligent ‘Asperger’ their families (e.g. McConachie et al., 2019). adults led to a new phenomenon; parents of recently We need to know whether rates of age-related diagnosed autistic children would recognise them- mental/physical illness are elevated in autism and, if selves when reading around the topic. At the same so, why. Possible reasons include shared genetic time, in research laboratories, the advent of neu- predisposition, or ‘phenotypic’ links mediated by, for roimaging was a driver to identify and recruit intellec- example, stress, isolation, reduced help-seeking and tually able autistic adults for research studies, using social disadvantage. Establishing the causes of any first PET and then (f)MRI. For the very first neuroimag- elevation in age-related ill health is the first step ing study of ToM in autism, in the early 1990s, such towards preventing this health inequality for older adults were still so hard to find that we collaborated adults on the autism spectrum. internationally and even flew volunteers over to the To accomplish this research, epidemiological United Kingdom specially to take part (Happe et al., approaches are needed. It is not currently clear where 1996)! to find the presumably 1% of the ageing population The Office of National Statistics household survey meeting criteria for autism. Establishing retrospective (Brugha et al., 2011) showed that the prevalence of diagnosis in, for example, dementia clinics, with little autism in adults was the same as in children, and likelihood of gathering developmental history, will be suggested that most autistic adults at that time were challenging. Longitudinal studies following adults undiagnosed. It is still the case that many adults are from midlife into old age are necessary, with account coming for first ASD diagnosis late in life, and the taken of changes in diagnostic criteria that create DSM-5 criteria explicitly allow for late recognition of differences between those adults diagnosed in child- characteristics that have been present since early hood (in the 1960s) and those receiving first diagnosis development but ‘may not become fully manifest by the current broader criteria. until social demands exceed limited capacities’. Although challenging, research into ageing in Although ‘Infantile Autism’ has become ‘Autism autism is important, not only for practical reasons Spectrum Disorder’, there is still a paucity of (planning services, preventing poor health), but research with adults and particularly with older potentially also for theoretical and fundamental adults (Roestorf et al., 2019). There is still little scientific insights. Studies of infant brain trajectories awareness of autism amongst most adult services, in autism, suggesting ‘overgrowth’ in the first 4 years especially old age psychiatry. Raising awareness is of life in a subset of children, have given rise to vital, both to understand the life course in autism, theoretical accounts of the neurobiology of autism and also because of indications that autistic adults (e.g. failure of typical synaptic pruning). Brain

© 2020 Association for Child and Adolescent Mental Health doi:10.1111/jcpp.13176 Changes in the concept of autism 223 changes in later life may also give important clues to first researcher to report empirical data on social the biology of autism, or indeed of other conditions, differences in parents of autistic children (Wolff, just as the study of early-onset dementia in Down Narayan & Moyes, 1988), although both Kanner and syndrome contributed to neurobiological under- Asperger noted unusual social manner in some standing of Alzheimer’s. parents of the children they diagnosed. However, it was not until the end of the 20th century that these subclinical autistic traits, the ‘broad(er) autism phe- The changing concept of autism: 4. From notype’, were systematically measured (Piven et al., discrete to dimensional 1997; Pickles et al., 2000; for review, see Rubenstein The original conception of autism was as a distinct & Chawla, 2018). entity, distinct from typical development and from A further extension of the broad autism phenotype other conditions. Indeed, in the opening sentence of beyond those with a genetic connection and into the Kanner’s famous 1943 paper, he remarks, ‘there have general population began with instruments such as come to our attention a number of children whose the AQ (Baron-Cohen et al., 2001) and Social Respon- condition differs so markedly and uniquely from siveness Scale (SRS; Constantino et al., 2000). These anything reported so far, that each case merits...a questionnaires began as ‘screeners’ to identify indi- detailed consideration of its fascinating peculiarities’ viduals who might warrant diagnostic assessment, (p. 217). Early research focused, for example, on but quickly gained popularity as measures of individ- identifying the unique (‘pathognomonic’) features ual differences. In the last ten years, there has been a distinguishing autism from other disorders, in order sharp increase in studies ‘of autism’ that have to help clinicians make a categorical judgement of included no diagnosed individuals at all, rather ‘autism’ or ‘not autism’. Interestingly, Wing’s work as examining correlates of high autism trait scores in early as 1969 compared parental report of social, neurotypical groups. One limitation of most of this ‘executive’ and other features in autistic children and research has been the sole use of self-report mea- children with sensory (e.g. visual) impairment, lan- sures. It is interesting, however, that some studies of guage disorder or Down syndrome, and concluded self-rated subclinical autistic traits have found a that although aspects of ‘autistic’ behaviour could be relationship with objective markers highlighted in found in 2- to 5-year-olds in these other groups, studies of diagnosed autism; for example, higher autism could be ‘clearly differentiated...by consider- (subclinical) AQ scores predicted more accurate pitch ation of the complete clinical picture’ (Wing, 1969). In and temporal processing (Stewart et al., 2018), and the same paper, however, Wing discusses whether lower performance on tests of social cognition and inclusion of less clear or borderline cases of autism executive function (Gokc€ ßen et al., 2016). would have altered these results. Indeed, Wing is Autistic trait measures such as the AQ show a widely credited with introducing the notion of the smooth continuum between diagnosed autism and ‘autism spectrum’, with a parallel intended with the subclinical individual differences; there is a normal spectrum of coloured light, which is heterogeneous distribution of traits, rather than a bimodal distri- but also continuous. bution (although see Abu-Akel et al., 2019 for a Measuring autistic traits or symptoms quantita- different modelling approach with large scale self- tively was necessary to measure differences within report data that supports both dimensional and the spectrum, as well as to track developmental categorical conceptions). While it should be born in change or treatment effects. Perhaps the first mind that the same behaviour can have different detailed interview schedule specifically designed for underpinnings, it does appear that, at the beha- autism was Wing and Gould’s ‘Children’s Handicaps, vioural level at least, one can be ‘a bit autistic’. Behavior and Skills’ (HBS) schedule, which was used At the genetic level, too, it appears that the genetic in the original Camberwell study (Wing & Gould, influences on subclinical autistic traits largely over- 1978). This can be seen as the predecessor to the lap with those on diagnosed autism, based on both development in the late 1980s of widely adopted behavioural genetic (Colvert et al., 2015; Robinson diagnostic scales, such as the Autism Diagnostic et al., 2011) and molecular genetic approaches Interview (ADI; LeCouteur et al., 1989) and ADOS (Massrali et al., 2018). Indeed, for the majority, the (Lord et al., 1989), which have been important in the genetics of autism is just like the genetics of height; attempt to establish comparability of research sam- their autism is the result of many common genetic ples across studies. variants, each of miniscule effect. We all carry many The notion that autism behavioural dimensions of these variants, and so, a dimensional character- might extend beyond autism is a relatively recent isation of autism is also plausible genetically. Only in idea, albeit one mirrored in many other diagnoses a minority of autistic people are rare genetic muta- (e.g. ADHD). Interest in the characteristics of first- tions of high penetrance, relevant to their autism – degree relatives of autism probands grew in the wake and even in these cases, genetic ‘background’ in of twin studies, which showed high concordance for terms of those common variants remains relevant. autism and also for a wider set of cognitive and Polygenetic scores, which add up the weighted language difficulties. Sula Wolff was possibly the effects of those hundreds or thousands of common

© 2020 Association for Child and Adolescent Mental Health 224 Francesca Happe and Uta Frith J Child Psychol Psychiatr 2020; 61(3): 218–32 variants, are found to be elevated even in families Lawson, Rees and Friston (2014) posit an imbalance whose autistic child shows a de novo genetic muta- of the precision ascribed to sensory evidence relative tion associated with autism (Weiner et al., 2017). to prior beliefs. Those conducting early studies of brain structure in autism, initially postmortem, then using increas- Discrete to dimensional: implications for future ingly sophisticated neuroimaging in vivo, expected to research find striking and specific neural differences under- lying the often dramatically different behavioural The failure, to date, to find qualitatively distinct phenotype identified as autism in the late 20th ‘diagnostic’ biomarkers distinguishing autism from century. However, recent large meta-analyses and nonautism, has led to a shift to search instead for studies pooling MRI scans across hundreds of stratification biomarkers, markers that might pre- autistic participants reveal surprisingly little by dict differences in, for example, prognosis or treat- way of qualitative differences from neurotypical ment response, within groups of autistic individuals samples (e.g. Pua, Bowden & Seal, 2017). However, (Loth & Evans, 2019). It will be interesting to see there are multiple quantitative differences in struc- whether huge studies such as AIMS-2-Trials will tural findings, as reviewed by, for example, van Rooij reveal distinct biological subtypes. In principle, et al. (2017). qualitatively different causes can underlie a smooth Behaviourally, genetically and neuroanatomically, continuum of quantitative differences in behaviour. then, a dimensional characterisation of autism As sample sizes in autism genetic consortia rise, appears warranted. What about the cognitive level? polygenic scores for autism may begin to explain a Cognitively autism may present a qualitative differ- meaningful proportion of variance in autistic traits. ence in some respects and not in others. The original It is important to note that such scores will not be ToM deficit account posited a qualitative difference; useful for individual prediction of autism diagnosis; autistic people do not meta-represent others’ propo- the probabilistic nature, the relatively weak signal sitional attitudes, while neurotypical people do and the low base rate of autism in the population (Frith, Morton & Leslie, 1991). More recent versions mean that any attempt to screen for autism would distinguish implicit from explicit ToM and place not only be ethically problematic but also practically emphasis on lack of spontaneous, automatic (and doomed to failure. Polygenic scores for autism are perhaps effortless) tracking of mental states in likely, however, to have a profound effect on autism autism (e.g. Schuwerk et al., 2016). There is scope, research. While established in discovery samples of then, for either a qualitative or quantitative distinc- hundreds of thousands, polygenic scores can be tion to be drawn. Individuals on the autism spec- applied in studies of hundreds of participants and trum also differ from one another in their ToM task will be feasible to include in psychological as well as performance, and this can change with age; whether biological research. Genomic structural equa- these differences should be interpreted as reflecting tion modelling, for example, allows the interrogation differences in core sociocognitive characteristics or of the genetic structure (overlap, independence) of ‘compensation’ is an interesting question addressed different traits or conditions. Already, Warrier et al. elsewhere (Livingston & Happe, 2017). (2017) have been able to collect phenotypic and Other cognitive characteristics of autism may genetic information on more than 50,000 people via more easily fit a dimensional approach. Executive commercial genotyping service 23andMe. Using dysfunction is described in autism, and many other Baron-Cohen’s Systemizing Quotient and Empathiz- groups, as quantitatively different from typically ing Quotient self-report questionnaires, their work developing samples. The ‘weak central coherence’ supports the distinction between social and nonso- account described quantitative differences in detail- cial aspects of autism, behaviourally and genetically focused cognitive style with an explicitly dimensional (Happe, Ronald & Plomin, 2006; see below). approach in which both extremes (configural bias A challenge for the future is that the huge volumes and featural bias) may have advantages for different of data collected, and the astonishing advances in tasks (Happe & Frith, 2006). This account also genetic and analytic methods, will only produce addressed sensory differences in autism (e.g. failure information as good as the measures employed. to habituate to stimuli; Frith, 1989), which have only There is a need for new methods to facilitate deep recently been recognised in the diagnostic system phenotyping at scale and to supplement current despite their significant impact on autistic people’s reliance on self-report questionnaires. For example, everyday lives. exploring sensory sensitivities with physiological The more recent Bayesian explanations of sensory measures and establishing agreement with subjec- and perceptual atypicalities in autism stem in part tive self-report will be important (Kuiper et al., 2019). from an intensified interest in sensory issues. These Sensitive assays that can distinguish between social accounts also lend themselves to quantitative inter- difficulties with different causes (e.g. autism versus pretations; for example, Pellicano and Burr (2012) social anxiety) and that target distinct and separable suggest that cognitive and sensory differences in aspects of social processing (e.g. ToM versus emo- autism result from attenuated Bayesian priors, while tional empathy versus social motivation) are needed;

© 2020 Association for Child and Adolescent Mental Health doi:10.1111/jcpp.13176 Changes in the concept of autism 225 to be optimally useful in the new era of big data and absorption of Asperger’s Disorder into ASD in DSM- polygenic scores, these need to be user-friendly, 5 (Happe, 2011). remotely administered and automatically coded. Despite increasingly large samples and sophisti- cated statistical methods, studies have largely failed to ‘cut nature at the joints’ with any greater intricacy The changing concept of autism: 5. From one than dividing autism with versus without intellectual to many disability (in the past unhelpfully referred to as ‘high’ Although autism was historically conceptualised as or ‘low-functioning’), and autism with versus without a discrete, categorical diagnosis based on a coherent language disorder (Gillespie-Lynch et al., 2012). syndrome of co-occurring symptoms, current con- Both of these aspects, which are additional and in ceptions question the unity of autism in two key some respects orthogonal to the core features of senses: first, a growing realisation that autism in autism, have profound impact on prognosis and different individuals likely has different causes or degree of support needs. aetiologies; and second, that even in a single indi- Despite this failure to find behavioural subtypes, vidual, different core symptoms may have different current consensus is that there are many different origins. biological routes to autism, many different aetiolo- gies, represented by the use of the term the ‘autisms’ (Coleman & Gillberg, 2012). From one to many: the ‘autisms’ The original conception of autism was as a single From one to many: unitary to ‘fractionable’ entity, and scientists searched for a single cause. There was from the start, however, a clear recogni- The second sense in which autism has changed tion of the behavioural heterogeneity; even the cases conceptually from one to many, concerns the sug- Kanner described showed a wide range of levels of gestion that the symptoms that define autism may adaptive and intellectual functioning. Rutter, in his have separable causes even in a single individual. 1968 review, describes the variation in intellectual The ‘fractionated triad’ hypothesis suggests that the and language ability, as well as in developmental social, communicative and rigid/repetitive aspects of trajectories, and the relevance of this for various autism have separable underpinnings at the genetic, possible accounts of the ‘primary defect’ in autism. neural and cognitive levels (Happe, Ronald & Plomin, Over the years, there has been a concerted effort to 2006). The idea that autism is a ‘compound condi- parse the huge behavioural heterogeneity of the tion’ is not new; Wing and Wing (1971) suggested autism spectrum into meaningful subgroups (e.g. that autism is best understood as ‘a combination... Zheng et al., 2019). This in part reflects a concern of impairments...’, based on the observation that that relative lack of progress in understanding the ‘isolated fragments of the full clinical picture fre- neurobiology of autism might be due to biological quently occur...’. However, relevant evidence from heterogeneity in study samples. Identifying distinct population-based samples was not collected until (and more homogeneous) behavioural subgroups Ronald and colleagues began to examine the beha- was expected to advance our understanding – in vioural and genetic association between autistic much the same way as the current trend for studying traits, measured in the Twins Early Development genetically homogeneous subgroups (see below). Study (TEDS). They found that parent-reported However, while there have been huge efforts of social and nonsocial autism traits correlated only subtyping, both by behaviour (e.g. regression) and modestly in children in the general population, and by biology (e.g. macrocephaly, serotonin levels), they even in subgroups with high traits and/or autism have had rather little success to date. spectrum diagnoses (see Happe & Ronald, 2008 for The inclusion of ‘Asperger’s Disorder’ as a separate review, and, e.g., Kim et al., 2018 for replication). diagnostic category from Autistic Disorder in DSM-IV While social, communication and rigid/repetitive (see above) was one such attempt, to make sense of difficulties did co-occur somewhat above chance, heterogeneity in early language and cognitive devel- many children showed pronounced difficulties in opment in the autism spectrum. Despite more than only one of the ‘triad’ of autistic traits. Using twin 100 papers comparing those diagnosed with Asper- modelling, Ronald and colleagues were also able to ger’s Disorder versus ‘high-functioning autism’, the show distinct genetic influences on different autism results were largely negative (no group differences) or symptom domains, in the general population and circular (differences on variables included in the amongst high trait/diagnosed autism samples (re- diagnostic process, such as clumsiness). Indeed, an viewed in Happe & Ronald, 2008). As mentioned influential study by Lord et al. (2012) showed that above, recent studies using polygenic scores in huge even comparing specialist clinics in the States, the samples of adults support this conclusion, finding best predictor of what diagnosis an individual distinct genetic signals for the social versus nonso- received (Asperger, Autism, PDD-NOS) was not any cial dimensions of autism (Warrier et al., 2017). characteristic of the person, but rather which clinic At the cognitive level, too, it has been suggested they attended. These considerations led to the that autism may be characterised by a compound of

© 2020 Association for Child and Adolescent Mental Health 226 Francesca Happe and Uta Frith J Child Psychol Psychiatr 2020; 61(3): 218–32 deficits/differences (e.g. impaired ToM, executive lack of it, such as lower intelligence or perhaps poor dysfunction and detail-focused processing bias) executive function. In addition, if a sizeable percent- and that attempts to find a unitary psychological age of children do indeed show autism-like symp- explanation for social and nonsocial behavioural toms in only one of the traditional triad domains, it features have largely failed (Brunsdon et al., 2015; would be important to know whether these are Happe & Ronald, 2008). The particular pattern of qualitatively different from the (e.g. social) difficulties cognitive strengths and weaknesses should, in prin- seen in autism, and what if any clinical needs such ciple, be reflected in everyday skills, and perhaps ‘single deficit’ individuals have. Lastly, the beha- help explain behavioural heterogeneity (Brunsdon vioural genetic work suggesting largely nonoverlap- et al., 2015). Although tracing such links is complex, ping genetic influences on social and nonsocial some examples exist; Jones, Simonoff and Baird aspects of autism would argue for polygenic scores et al. (2018) found direct links between ToM (but not to be created not for autism as a whole, but for executive function) task performance and parent- individual differences in social skills/deficits, and rated symptoms in autistic adolescents. separately for rigid/repetitive traits.

From one to many: implications for future research The changing concept of autism: 6. From pure With regard to parsing the ‘autisms’, the search for to complex stratification biomarkers to discover autism sub- In medical tradition, some diagnoses ‘trump’ or groups for ‘personalised’ interventions is already override others. In the 1980s, there was debate as underway (Loth & Evans, 2019). There is also to whether a diagnosis of autism was trumped by an considerable research effort focused on understand- aetiological medical diagnosis; if a brain lesion was ing the biological path to autism in genetically discovered or Tuberous Sclerosis diagnosed, argued homogenous subgroups (e.g. chromosome 22q13 some, the autism diagnosis should be removed. deletion). Clinical groups defined by aetiology, such ‘Idiopathic’ autism was considered by some as ‘real’ as Fragile X, have been viewed as a tractable target autism, as distinct from autism that was secondary and a step towards finding a putative ‘final common to a known neural or genetic basis. This view was pathway’ relevant also to ‘idiopathic’ autism. How- part of the historical search for a single biological ever, it remains unclear whether ‘autistic’ symp- characterisation of autism, discussed above – and tomatology seen in these groups is qualitatively the stands in sharp contrast to today’s conception in same as that in autism; early studies of FraX which autism is quintessentially a behavioural diag- mistook eye-gaze avoidance for social impairment, nosis that can accompany a wide range of biological and recent neuropsychological studies of children conditions. There is still, however, some diagnostic with 22q13 deletion have suggested social difficul- overshadowing, and some clinicians may be slower ties may be less important than language difficulties to diagnose autism in children with a known genetic (Laura et al., 2018). Some have argued that the study syndrome (e.g. Down syndrome; Wester Oxelgren of specific aetiological groups in which autism is et al., 2019) than those without. common has established the final common pathway Despite some early voices flagging the high rates of as perturbation in synaptic functioning and excita- co-occurring physical and psychiatric conditions tory/inhibitory imbalance (e.g. Oliveira et al., 2018). accompanying autism, most notably Gillberg (re- However, it may be that these fundamental and viewed in Gillberg & Billstedt, 2000), only recently complex neural processes can be disrupted in so has there been widespread awareness that autism many different ways, that this ‘final common path- rarely occurs alone. Historically, an autism diagno- way’ is so broad as to represent, in effect, a plethora sis trumped a wide range of psychiatric diagnoses, of distinct ‘roads’. including anxiety, ADHD and, problematically, Regarding the ‘fractionated triad’ notion, does this Asperger’s Disorder (in DSM-IV). Only in DSM-5 imply that autism, per se, does not exist? We would (APA, 2013) were multiple diagnoses allowed in suggest not; the particular mixture or compound of combination with ASD for the first time. It is now causes (genetic, neural, cognitive) may be qualita- clear that additional difficulties are common in tively more than the sum of its parts. Thus, although autism, and although clinic samples are naturally autism may share genetic or cognitive (e.g. executive enriched for complex and comorbid cases, popula- dysfunction) aspects with other clinical groups (e.g. tion-based studies also show elevated rates of many ADHD), these may interact with the other aspects of physical and mental health problems. A recent meta- autism to create a unique and distinct condition. The analysis by Lai et al. (2019) of 83 studies produced fractionated triad hypothesis would, however, sug- the following estimated rates of co-occurring psychi- gest that transdiagnostic studies, comparing differ- atric conditions in autism: ADHD 33%, anxiety ent neurodevelopmental groups, will be worthwhile. disorders 23%, sleep–wake disorders 13%, depres- It will also be important, in future work, to consider sive disorders 12%, obsessive–compulsive disorder how to distinguish core cognitive characteristics of 10%, disruptive/impulse-control/conduct disorders autism from aspects that relate to compensation or 10%, schizophrenia spectrum disorders 5%, and

© 2020 Association for Child and Adolescent Mental Health doi:10.1111/jcpp.13176 Changes in the concept of autism 227 bipolar disorders 5%. The seriousness of mental occurring condition. Examples might include health problems in autism is becoming clear; popu- reduced resources for compensation, due to socioe- lation-based data on suicide (from >27,000 adults) conomic disadvantage or executive dysfunction. suggested an odds ratio >7 for autistic adults, with Fourth, there may be shared aetiology, environmen- suicide risk being especially high for autistic women tal, genetic or both (Tick et al., 2016). Knowing which and those without intellectual disability (Hirvikoski of these different possibilities underlies high rates of et al., 2016). co-occurring conditions in autism matters because it Alexithymia (difficulty identifying and talking may inform intervention; for many autistic people, it about your own feelings) is an important co-occur- is not the autism but the co-occurring anxiety, ring trait that appears to affect about half of autistic depression, epilepsy or sleep problems that most adults (see Kinnaird et al., 2019 for a meta-analysis). impair quality of life. Much more research is needed First studied by Hill et al. (2004), later work by Bird to understand these, and less well-recognised and colleagues established links between high alex- co-occurring problems such as catatonia (motor ithymia and difficulty recognising others’ emotions ‘freezing’, temporary muteness, difficulty initiating and responding empathically to them. Bird’s work, movement; Shah, 2019). Establishing the reasons comparing autistic and nonautistic groups high and for high rates of co-occurring problems in autism is low in alexithymia, suggests that it is co-occurring an important aim for research. Longitudinal studies alexithymia and not autism itself that is associated are often suggested as the best way to establish with reduced recognition of and empathic response causality, but while they can (sometimes) establish to others’ emotions (Bird & Cook, 2013). Alexithymia order of emergence of conditions or traits, moving is also common in many other clinical conditions, beyond association is complex (Happe, 2001) and including depression, substance abuse, psychosis may require, for example, intervention designs. and eating disorders. Interestingly, historically It is now clear that studies of autism will not be Rastam et al. (1997) noted high levels of alexithymia representative if researchers exclude participants in women with anorexia who also showed ‘empathy with common mental health problems such as disorders’, a wider term that for Gillberg included anxiety, in search of ‘pure’ autism. Perhaps less autism. clear in the field is that this, debatably, means we must also stop screening out common mental health conditions in our comparison ‘typically developing/ From pure to complex: implications for future ed’ control groups. Considering co-occurring condi- research tions, and their possible confounding effects, in A James Lind Alliance survey in 2018 by research autism vs. comparison group differences, is essen- charity Autistica, and a similar large survey of tial. Does low self-esteem/depression inflate corre- stakeholders in North America (Frazier et al., lations between self-report measures (including 2018), found that mental health was a top priority autism trait self-ratings) that focus on deficits and research area identified by autistic adults. Why is difficulties? The presence of co-occurring alex- autism so often accompanied by mental health, ithymia may be particularly important to note in difficulties? A number of possibilities exist, which research studies, where it may help explain hetero- are by no means mutually exclusive (and which geneity in (emotional) task performance, and in might be marked with more nuanced terminology; clinical work, where it may affect treatment engage- see, Rubenstein & Bishop-Fitzpatrick, 2019). First, ment and response. apparent ‘comorbidity’ can be due to selection bias, Given the high rates of additional difficulties, the and additional problems may raise the likelihood of surprising subgroup becomes those autistic children an individual requiring clinical services. One analy- and adults without mental health problems. sis of twin data (Tick et al., 2016) suggested a Population-based studies may be needed to find phenotypic link from raised hyperactivity to autism, these (perhaps undiagnosed) individuals, who may interpreted by the authors as hyperactive behaviour give us insight into resilience factors (individual or making diagnosis of autism more likely (when environmental) that are important for living a happy another child with the same level of autistic traits autistic life. might not be diagnosed if quiet and not disruptive). Second, real phenotypic causal associations may exist between autism and co-occurring conditions; The changing concept of autism: 7. From for example, social exclusion or bullying may lead to ‘developmental disorder’ to neurodivergence anxiety, depression and even PTSD. Communication Over the last 30 years, and rapidly in the last difficulties central to autism may reduce help-seek- decade, concepts of autism in many places have ing, with negative effects on health. Unusual eating shifted from a purely medical model to a more social patterns, restricted diets and sensory sensitivities model of disability (Shakespeare, 2017; see also may have adverse gastrointestinal effects. Associa- discussion in Fletcher-Watson & Happe, 2019). The tions may be due to a third factor affecting both traditional notion that autism is a disorder defined autism (manifestation/diagnosis) and the co- purely by deficits inherent to the person, has been

© 2020 Association for Child and Adolescent Mental Health 228 Francesca Happe and Uta Frith J Child Psychol Psychiatr 2020; 61(3): 218–32 challenged. Instead, autism may be considered a attracting to the field scientists not previously inter- difference (‘neurodivergence’) that constitutes a dis- ested in autism. More recently, research charities ability in the context of the demands of the neu- such as Autistica have embraced the new era of rotypical world. This change to a neurodiversity stakeholder- and self-advocate-led research perspective has been driven by autistic voices; for a (Fletcher-Watson et al., 2018). New participatory full history of autistic self-advocacy and the neuro- research models challenge nonautistic researchers diversity movement, the reader is referred to the to collaborate with autistic people at every stage of edited volume by Kapp (2019). It is notable that the research, from identifying key questions, designing first – and highly influential – autistic autobiogra- methods, recruiting participants, interpreting find- phy, by Temple Grandin, was published in 1986 and ings, to dissemination and public engagement. Such that today there are hundreds of first-person ways of working open new research avenues; sensory accounts. issues would have been a major scientific focus Under this new conception, the old talk of ‘curing’ much sooner if researchers had been working more autism is no longer applicable or acceptable. There is closely with autistic people. Although sensory differ- sometimes a tension, however, between autistic ences had been addressed in the ground-breaking adults promoting autism as a part of neurodiversity, psychological experiments of Hermelin and and parents of those individuals who are severely O’Connor (1970), it took the direct input from disabled by intellectual impairment/language disor- autistic individuals to give them the prominence der/epilepsy. The research priorities that grow from they now have, and inclusion in DSM-5. these different experiences of autism are naturally The challenge for the future will be how to ensure different. Of course, nobody disagrees that many of that all the diverse voices within the autism spec- the co-occurring conditions common in autism are trum, with as well as without intellectual and/or valid treatment targets. No one wants to keep their language disability, are heard. There is also the depression, crippling anxiety, sleep problems or need, as in all science, to allow ‘blue skies’ and gastrointestinal conditions. Similarly, interventions theoretical work alongside research with obvious for intellectual disability and language disorder and direct practical ‘impact’, recognising that con- would be welcomed and do not stigmatise autism, crete benefits often emerge unplanned from unex- or threaten its characterisation as a different, not pected lines of research. deficient, way of being. As discussed above, the dimensional conception of autism has no natural cut-off point where high Conclusions: challenges and opportunities autism traits become ‘autism’. In DSM-5, an ASD The concept and diagnosis of autism has seen diagnosis requires that the autistic traits ‘cause gradual as well as dramatic changes in the last few clinically significant impairment in social, occupa- decades, and it continues to evolve. We have drawn tional, or other important areas of current function- on personal recollection to trace some of the major ing’. If ‘impairment’ is a function of the interaction changes in the last 30 years or so and captured between the person’s characteristics and the these changes under seven headings, considering demands of the environment/context, this means the implications for future research. Our review has that an autism diagnosis based on current beha- identified a number of future challenges for vioural assessments could potentially come and go. researchers. We believe that the change from narrow That is, the same person may live happily with their to wide can account for the increased prevalence of autistic traits in one context or at one age when they autism spectrum disorders. The change from rare to find their niche, but be impaired by them in another common is still ongoing, with questions about the context that is less accommodating. Are we ready to under-representation of females. The change from consider autism as something that comes and goes? childhood to lifespan highlights the need for research Do we need a term for autism/high autistic traits on ageing. The change from discrete to dimensional that are not impairing; not a diagnosis, then, but a poses new questions regarding nonimpairing traits. cognitive style or personality type, perhaps? If so, The change from one to many requires that we this might be the most dramatic reconceptualisation consider the fractionation of autism. The change of autism yet. from pure to complex acknowledges that additional mental health problems are common and may be confounds in research. Finally, the change from From ‘developmental disorder’ to neurodivergence: ‘developmental disorder’ to neurodiversity requires implications for future research collaborative approaches to research with the very Traditionally, the autism research agenda has been diverse autism community. scientist- or funder-led, with some notable (autism) In the course of looking back to look forward, some parent scientists making a major contribution his- neglected topics emerged. There is little research on torically (e.g. Wing and Rimland) and currently. In intellectual impairment, evidence-based educational the States in particular, parent-led charities raised approaches and technological aides for the many funds and influenced the direction of research by autistic people with ID. Language, once such a focus

© 2020 Association for Child and Adolescent Mental Health doi:10.1111/jcpp.13176 Changes in the concept of autism 229 for autism research, is now relatively little studied, Future research is likely to be increasingly dom- but many important questions remain; for example, inated by big data, but it would be a serious mistake how are some autistic children able to acquire for researchers to lose sight of individuals. Students language apparently without delay or atypicality, and early career researchers exploring autism in given the apparently vital role of social processing secondary analysis of huge datasets need also to (e.g. recognition of speaker’s intentions) in word work directly with autistic people, to really under- learning? What is the role of motor impairment, or stand their experiences and concerns. Furthermore, some as yet unquantified deficit in volitional action, the power of sharing information between research- in the language impairment of ‘minimally verbal’ ers and community members, as well as making autistic people? joint decisions about research priorities, is invalu- By contrast with these neglected areas, some areas able. Participatory research models and codesign of autism research are flourishing around the world. with autistic people and the wider autism commu- These include the groundbreaking infant-sibling nity can ensure that autism never becomes just a studies, which track from birth children at elevated variable in a spreadsheet. genetic probability of autism; early intervention For those just setting out on their own journey in programmes are increasingly being held to the same autism research, this is an exciting time; our under- standards of evidence as traditional medical trials; standing of autism has changed so much over the and genetic consortia with open-access data are last few decades, and it is almost impossible to reaching critical mass for major discoveries. The full imagine what our concept of autism will be in 2060. impact of these will be felt in the coming decades. Autism research has typically focused on white males in high-income countries, and it is only very Acknowledgements recently that researchers are recognising that most F.H. is part-funded by the National Institute for Health autistic people live in low- and middle-income coun- Research (NIHR) Biomedical Research Centre at South tries. How culture, ethnicity and socioeconomic London and Maudsley NHS Foundation Trust and status affect not only pathway to diagnosis, but also King’s College London. The views expressed are those the manifestation of autism and developmental of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health and Social Care. adaptation, has yet to be properly explored. The authors would like to thank Isabel Sinha for their While autism so far remains a purely behavioural kind help with references and Chris Frith for useful diagnosis, important questions arise about how we discussion. The authors have declared that they have recognise it in women, in the elderly, and how far we no competing or potential conflicts of interest. would allow the phenotype to stray from our current notion and still be called ‘autism’. Will newly devel- oped cognitive tests or polygenic scores help by Correspondence taking us beyond behaviour; could they ever aid Francesca Happe, Social, Genetic & Developmental diagnostic decision-making? Could cognitive assays Psychiatry Centre, Institute of Psychiatry, Psychology assist recognition of autism that looks different, & Neuroscience (PO 80), King’s College London, De perhaps uncovering layers of compensation or cam- Crespigny Park, Denmark Hill, London SE5 8AF, UK; ouflage? Email: [email protected]

Key points

 Originally, and even as late as the 1980s, autism was conceptualised as a rare and overwhelmingly male disorder of childhood, categorically distinct from typical development and other disorders, and usually accompanied by intellectual and language disabilities.  Now autism is thought of as a relatively common and dimensional condition/neurodivergence, a lifelong condition, underdiagnosed in females, and usually accompanied by co-occurring mental health conditions.  Research approaches have changed accordingly, with greater awareness of heterogeneity, increasing emphasis on large sample sizes and use of autism trait measures with subclinical groups.  Future research should make use of new technology to include autistic individuals with intellectual/language impairments, address the historical imbalance in participant gender, study autism in LAMICs and balance big data with deep phenotyping.  A participatory research approach that includes the views of the autism community is needed to ensure that research priorities are constantly monitored and reviewed.

© 2020 Association for Child and Adolescent Mental Health 230 Francesca Happe and Uta Frith J Child Psychol Psychiatr 2020; 61(3): 218–32

References Fletcher-Watson, S., & Happe, F. (2019). Autism: A new Abu-Akel, A., Allison, C., Baron-Cohen, S., & Heinke, D. introduction to psychological theory and current debate. (2019). The distribution of autistic traits across the autism London: Routledge. spectrum: evidence for discontinuous dimensional subpop- Fombonne, E. (2018). The rising prevalence of autism. Journal – ulations underlying the autism continuum. Molecular Aut- of Child Psychology and Psychiatry, 59, 717 720. ism, 10, 24. Frazier, T.W., Dawson, G., Murray, D., Shih, A., Sachs, J.S., & American Psychiatric Association (1980). Diagnostic and sta- Geiger, A. (2018). Brief Report: A survey of autism research tistical manual of mental disorders: DSM-III. (1994). DSM-IV. priorities across a diverse community of stakeholders. â – (2013) DSM-5 . Washington: American Psychiatric Publish- Journal of autism and developmental disorders, 48, 3965 ing. 3971. Arvidsson, O., Gillberg, C., Lichtenstein, P., & Lundstrom,€ S. Frith, U. (1989). Autism: Explaining the enigma. Oxford: (2018). Secular changes in the symptom level of clinically Blackwells. diagnosed autism. Journal of Child Psychology and Psychi- Frith, U., Morton, J., & Leslie, A.M. (1991). The cognitive basis atry, 59, 744–751. of a biological disorder: Autism. Trends in Neurosciences, 14, – Asperger, H. (1944). Die “Autistischen Psychopathen” im Kin- 433 438. desalter. European Archives of Psychiatry and Clinical Neuro- Gillberg, C., & Billstedt, E. (2000). Autism and Asperger science, 117,76–136. English translation by U. Frith (1991), syndrome: Coexistence with other clinical disorders. Acta – Asperger, H. (1991). ‘Autistic psychopathy’ in childhood. In U. Psychiatrica Scandinavica, 102, 321 330. Frith (Ed.), Autism and Asperger syndrome (pp. 37–92). Gillespie-Lynch, K., Sepeta, L., Wang, Y., Marshall, S., Gomez, Baron-Cohen, S., Tsompanidis, A., Auyeung, B., Nørgaard- L., Sigman, M., & Hutman, T. (2012). Early childhood Pedersen, B., Hougaard, D.M., Abdallah, M., ... & Pohl, A. predictors of the social competence of adults with autism. – (2019). Foetal oestrogens and autism. Molecular Psychiatry. Journal of Autism and Developmental Disorders, 42, 161 https://doi.org/10.1038/s41380-019-0454-9 174. € ß Baron-Cohen, S., Wheelwright, S., Skinner, R., Martin, J., & Gokcen, E., Frederickson, N., & Petrides, K.V. (2016). Theory Clubley, E. (2001). The autism-spectrum quotient (AQ): of mind and executive control deficits in typically developing Evidence from Asperger syndrome/high-functioning autism, adults and adolescents with high levels of Autism traits. – males and females, scientists and mathematicians. Journal Journal of Autism and Developmental Disorders, 46, 2072 of Autism and Developmental Disorders, 31,5–17. 2087. Bird, G., & Cook, R. (2013). Mixed emotions: The contribution Grandin, T. (1986). Emergence, labeled autistic. Novato, CA: of alexithymia to the emotional symptoms of autism. Trans- Academic Therapy Publications. lational Psychiatry, 3, e285. Grinker, R.R. (2008). Unstrange minds: Remapping the world of Brugha, T.S., McManus, S., Bankart, J., Scott, F., Purdon, S., autism. New York: Basic Books. Smith, J., ... & Meltzer, H. (2011). Epidemiology of autism Happe, F. (2001). Social and non-social development in spectrum disorders in adults in the community in England. autism: Where are the links? In J.A. Burack, T. Charman, Archives of General Psychiatry, 68, 459–465. N. Yirmiya & P.R. Zelazo (Eds.), Perspectives on development – Brunsdon, V.E.A., Colvert, E., Ames, C., Garnett, T., Gillan, N., in autism (pp. 237 253). Mahwah, NJ: Lawrence Erlbaum Hallett, V., ... & Happe, F. (2015). Exploring the cognitive Associates. features in children with autism spectrum disorders, their Happe, F. (2011). Criteria, categories, and continua: autism co-twins, and typically-developing children within a popu- and related disorders in DSM-5. Journal of the American – lation-based sample. Journal of Child Psychology and Psy- Academy of Child & Adolescent Psychiatry, 50, 540 542. chiatry, 56, 893–902. Happe, F. (2015). Autism as a neurodevelopmental disorder of – Carpenter, B., F. Happe, & J. Egerton (Eds.) (2019). Autism mind-reading. Journal of the British Academy, 3, 197 209. and girls. London: Taylor & Francis/Routledge. Happe, F., Ehlers, S., Fletcher, P., Frith, U., Johansson, M., ... Coleman, M., & Gillberg, C. (2012). The autisms (4th ed.). New Gillberg, C., & Frith, C. (1996). ‘Theory of mind’ in the York: Oxford University Press USA. brain. Evidence from a PET scan study of Asperger syn- – Colvert, E., Tick, B., McEwen, F., Ames, C., Curran, S., drome. NeuroReport, 8, 197 201. Woodhouse, E., ...& Bolton, P. (2015). Heritability of autism Happe, F., & Frith, U. (2006). The weak coherence account: and autism spectrum disorder in a UK twin sample. JAMA Detail-focused cognitive style in autism spectrum disorders. – Psychiatry, 72, 415–423. Journal of Autism and Developmental Disorders, 36,5 25. Constantino, J.N., Przybeck, T., Friesen, D., & Todd, R.D. Happe, F., & Ronald, A. (2008). ‘Fractionable Autism Triad’: A (2000). Reciprocal social behavior in children with and review of evidence from behavioural, genetic, cognitive and – without pervasive developmental disorders. Journal of Devel- neural research. Neuropsychology Review, 18, 287 304. opmental and Behavioral Pediatrics, 21,2–11. Happe, F., Ronald, A., & Plomin, R. (2006). Time to give up on a – Croen, L.A., Zerbo, O., Qian, Y., Massolo, M.L., Rich, S., single explanation for autism. Nature Neuroscience, 9, 1218 Sidney, S., & Kripke, C. (2015). The health status of adults 20. on the autism spectrum. Autism, 19, 814–823. Hermelin, B., & O’Connor, N. (1970). Psychological experi- Donvan, J.J., & Zucker, C.B. (2017). In a different key: The ments with autistic children. Oxford: Pergamon. story of autism. New York: Broadway Books. Hill, E., Berthoz, S., & Frith, U. (2004). Brief report: Cognitive Elsabbagh, M., Divan, G., Koh, Y.J., Kim, Y.S., Kauchali, S., processing of own emotions in individuals with autistic Marcın, C., ... & Fombonne, E. (2012). Global prevalence of spectrum disorder and in their relatives. Journal of Autism – autism and other pervasive developmental disorders. Autism and Developmental Disorders, 34, 229 235. Research, 5, 160–179. Hirvikoski, T., Mittendorfer-Rutz, E., Boman, M., Larsson, H., € Evans, B. (2017). The metamorphosis of autism. A history of Lichtenstein, P., & Bolte, S. (2016). Premature mortality in child development in Britain. Manchester: Manchester autism spectrum disorder. British Journal of Psychiatry, – University Press. 208, 232 238. Feinstein, A. (2011). A history of autism: Conversations with Hull, L., Mandy, W., Lai, M.C., Baron-Cohen, S., Allison, C., the pioneers. Hoboken, NJ: John Wiley & Sons. Smith, P., & Petrides, K.V. (2019). Development and valida- Fletcher-Watson, S., Adams, J., Brook, K., Charman, T., tion of the camouflaging autistic traits questionnaire (CAT-Q). – Crane, L., Cusack, J., & Pellicano, E. (2018). Making the Journal of Autism and Developmental Disorders, 49, 819 833. future together: Shaping autism research through meaning- Jack, A., & Pelphrey, K.A. (2017). Annual Research Review: – ful participation. Autism, 23, 943–953. Understudied populations within the autism spectrum

© 2020 Association for Child and Adolescent Mental Health doi:10.1111/jcpp.13176 Changes in the concept of autism 231

Current trends and future directions in neuroimaging systematic review and meta-analysis. Journal of the Amer- research. Journal of Child Psychology and Psychiatry, 58, ican Academy of Child & Adolescent Psychiatry, 56, 466– 411–435. 474. Jacquemont, S., Coe, B.P., Hersch, M., Duyzend, M.H., Lord, C., Petkova, E., Hus, V., Gan, W., Lu, F., Martin, D.M., ... Krumm, N., Bergmann, S., ... & Eichler, E.E. (2014). A & Risi, S. (2012). A multisite study of the clinical diagnosis of higher mutational burden in females supports a “female different autism spectrum disorders. Archives of General protective model” in neurodevelopmental disorders. Ameri- Psychiatry, 69, 306–313. can Journal of Human Genetics, 94, 415–425. Lord, C., Rutter, M., Goode, S., Heemsbergen, J., Jordan, H., Jones, C.R., Simonoff, E., Baird, G., Pickles, A., Marsden, A.J., Mawhood, L., & Schopler, E. (1989). Autism diagnostic Tregay, J., ... & Charman, T. (2018). The association observation schedule: A standardized observation of com- between theory of mind, executive function, and the symp- municative and social behavior. Journal of Autism and toms of autism spectrum disorder. Autism Research, 11,95– Developmental Disorders, 19, 185–212. 109. Loth, E., Charman, T., Mason, L., Tillmann, J., Jones, E.J., Kanner, L. (1943). Autistic disturbances of affective contact. Wooldridge, C., ...& Banaschewski, T. (2017). The EU-AIMS Nervous Child, 2, 217–250. Longitudinal European Autism Project (LEAP): Design and Kanner, L. (1971). Follow-up study of eleven autistic children methodologies to identify and validate stratification originally reported in 1943. Journal of Autism and Childhood biomarkers for autism spectrum disorders. Molecular Aut- Schizophrenia, 1, 119–145. ism, 8, 24. Kapp, S. (Ed.) (2019). Autistic community and the neurodiver- Loth, E., & Evans, D.W. (2019). Converting tests of fundamen- sity movement: Stories from the Frontline. London: Palgrave tal social, cognitive, and affective processes into clinically MacMillan. useful bio-behavioral markers for neurodevelopmental con- Kim, H., Keifer, C.M., Rodriguez-Seijas, C., Eaton, N.R., ditions. Wiley Interdisciplinary Reviews: Cognitive Science, Lerner, M.D., & Gadow, K.D. (2018). Structural hierarchy 10, e1499. of autism spectrum disorder symptoms: An integrative Lotter, V. (1966). Epidemiology of autistic conditions in young framework. Journal of Child Psychology and Psychiatry, children: I. Prevalence. Social Psychiatry, 1, 124–137. 59,30–38. Lund, J., & Jensen, J. (1989). Dimensional approach to Kinnaird, E., Stewart, C., & Tchanturia, K. (2019). Investigat- infantile autism in mentally retarded adults. Acta Psychi- ing alexithymia in autism: A systematic review and meta- atrica Scandinavica, 80, 389–394. analysis. European Psychiatry, 55,80–89. Lundstrom,€ S., Reichenberg, A., Anckarsater,€ H., Lichtenstein, Kuiper, M.W., Verhoeven, E.W., & Geurts, H.M. (2019). Stop P., & Gillberg, C. (2015). Autism phenotype versus registered making noise! Auditory sensitivity in adults with an autism diagnosis in Swedish children: Prevalence trends over 10 spectrum disorder diagnosis: Physiological habituation and years in general population samples. British Medical Jour- subjective detection thresholds. Journal of Autism and nal, 350, h1961. Developmental Disorders, 49, 2116–2128. Massrali, A., Brunel, H., Hannon, E., Wong, C., Baron-Cohen, Lai, M.C., Kassee, C., Besney, R., Bonato, S., Hull, L., Mandy, S., & Warrier, V. (2018). Integrated genetic and methylomic W., ... & Stephanie, H. (2019). Prevalence of co-occurring analyses identify shared biology between autism and autis- mental health diagnoses in the autism population: A tic traits. Molecular Autism, 10, 31. systematic review and meta-analysis. Lancet, Available at McConachie, H., Wilson, C., Mason, D., Garland, D., Parr, SSRN: https://ssrn.com/abstract=3310628 J.R., Rattazzi, A., ... & Magiati, I. (2019). What is important Lai, M.C., Lombardo, M.V., Chakrabarti, B., Ruigrok, A.N., in measuring quality of life? Reflections by autistic adults in Bullmore, E.T., Suckling, J., ... & MRC AIMS Consortium four countries. Autism in Adulthood. (2018). Neural self-representation in autistic women and Oliveira, B., Mitjans, M., Nitsche, M.A., Kuo, M.F., & Ehren- association with ‘compensatory camouflaging’. Autism, 23, reich, H. (2018). Excitation-inhibition dysbalance as predic- 1210–1223. tor of autistic phenotypes. Journal of Psychiatric Research, Laura, P., Marie, G., Romuald, B., Catherine, B., Sylvie, R., 104,96–99. Arnold, M., ... &Fred erique, B.B. (2018). 22q13 deletion Pellicano, E., & Burr, D. (2012). When the world becomes ‘too syndrome: communication disorder or autism? evidence real’: A Bayesian explanation of autistic perception. Trends from a specific clinical and neurophysiological phenotype. in Cognitive Sciences, 16, 504–510. Translational psychiatry, 8, 146. Pickles, A., Starr, P., Kazak, S., Bolton, P., Papanikolaou, K., Laurie, M.H., Warreyn, P., Uriarte, B.V., Boonen, C., & Bailey, A., ...& Rutter, M. (2000). Variable expression of the Fletcher-Watson, S. (2019). An international survey of autism broader phenotype: Findings from extended pedi- parental attitudes to technology use by their autistic chil- grees. Journal of Child Psychiatry & Psychology, 41, 491–502. dren at home. Journal of Autism and Developmental Disor- Pinti, P., Tachtsidis, I., Hamilton, A., Hirsch, J., Aichelburg, C., ders, 49, 1517–1530. Gilbert, S., & Burgess, P.W. (2018). The present and future Lawson, T.P., Rees, G., & Friston, K.J. (2014). An aberrant use of functional near-infrared spectroscopy (fNIRS) for precision account of autism. Frontiers in Human Neuro- cognitive neuroscience. Annals of the New York Academy science, 8, 302. of Sciences. https://doi.org/10.1111/nyas.13948. LeCouteur, A., Rutter, M., Lord, C., Rios, P., Robertson, S., Piven, J., Palmer, P., Jacobi, D., Childress, D., & Arndt, S. Holdgrafer, M., ...& McLennan, J. (1989). Autism diagnostic (1997). Broader autism phenotype: Evidence from a family interview: A standardized investigator-based instrument. history study of multiple-incidence autism families. Ameri- Journal of Autism and Developmental Disorders, 19, 363– can Journal of Psychiatry, 154, 185–190. 387. Pua, E.P.K., Bowden, S.C., & Seal, M.L. (2017). Autism Lever, A.G., & Geurts, H.M. (2016). Age-related differences in spectrum disorders: Neuroimaging findings from systematic cognition across the adult lifespan in autism spectrum reviews. Research in Autism Spectrum Disorders, 34,28–33. disorder. Autism Research, 9, 666–676. Rastam, M., Gillberg, C., Gillberg, I.C., & Johansson, M. Livingston, L., & Happe, F. (2017). Conceptualising compen- (1997). Alexithymia in anorexia nervosa: a controlled study sation in neurodevelopmental disorders: Reflections from using the 20-item Toronto Alexithymia Scale. Acta Psychi- autism spectrum disorder. Neuroscience & Biobehavioral atrica Scandinavica, 95, 385–388. Reviews, 80, 729–742. Ridley, R. (2019). Some difficulties behind the concept of the Loomes, R., Hull, L., & Mandy, W.P.L. (2017). What is the ‘Extreme male brain’ in autism research. A theoretical male-to-female ratio in autism spectrum disorder? A review. Research in Autism Spectrum Disorders, 57,19–27.

© 2020 Association for Child and Adolescent Mental Health 232 Francesca Happe and Uta Frith J Child Psychol Psychiatr 2020; 61(3): 218–32

Robinson, E.B., Koenen, K.C., McCormick, M.C., Munir, K., Tantam, D. (1988). Annotation: Asperger’s syndrome. Journal Hallett, V., Happe, F., ...& Ronald, A. (2011). Evidence that of Child Psychiatry & Psychology, 29, 836–840. autistic traits show the same etiology in the general popu- Taylor, B., Jick, H., & MacLaughlin, D. (2013). Prevalence and lation and at the quantitative extremes (5%, 2.5%, and 1%). incidence rates of autism in the UK: Time trend from 2004– Archives of General Psychiatry, 68, 1113–21. 2010 in children aged 8 years. British Medical Journal Open, Robinson, E.B., Lichtenstein, P., Anckarsater,€ H., Happe, F., & 3, e003219. Ronald, A. (2013). Examining and interpreting the female Tick, B., Colvert, E., McEwen, F., Stewart, C., Woodhouse, E., protective effect against autistic behavior. Proceedings of the Gillan, N., & Rijsdijk, F. (2016). Autism spectrum disorders National Academy of Sciences of the United States of and other mental health problems: exploring etiological America, 110, 5258–5262. overlaps and phenotypic causal associations. Journal of Roestorf, A., Bowler, D.M., Deserno, M.K., Howlin, P., Klinger, the American Academy of Child & Adolescent Psychiatry, 55, L., McConachie, H., ...& Geurts, H.M. (2019). “Older adults 106–113. with ASD: The consequences of aging”. Insights from a series van Rooij, D., Anagnostou, E., Arango, C., Auzias, G., of special interest group meetings held at the International Behrmann, M., Busatto, G.F. ... & Buitelaar, J.K. (2017). Society for Autism Research 2016–2017. Research in Autism Cortical and subcortical brain morphometry differences Spectrum Disorders, 63,3–12. between patients with autism spectrum disorder and Rubenstein, E., & Bishop-Fitzpatrick, L. (2019). A matter of healthy individuals across the lifespan: Results from the time: The necessity of temporal language in research on ENIGMA ASD Working Group. American Journal of Psychi- health conditions that present with autism spectrum disor- atry, 175, 359–369. der. Autism Research, 12,20–25. Warrier, V., Toro, R., Chakrabarti, B., Borglum, A., Grove, J., Rubenstein, E., & Chawla, D. (2018). Broader autism pheno- Hinds, D., ... & 23andMe Research Team (2017). System- type in parents of children with autism: A systematic review izing is genetically correlated with autism and is genetically of percentage estimates. Journal of Child and Family Stud- distinct from social autistic traits. bioRxiv, 228254. ies, 27, 1705–1720. Weiner, D.J., Wigdor, E.M., Ripke, S., Walters, R.K., Kosmicki, Russell, G., Mandy, W., Elliott, D., White, R., Pittwood, T., & J.A., Grove, J., ... & Werge, T. (2017). Polygenic transmis- Ford, T. (2019). Selection bias on intellectual ability in sion disequilibrium confirms that common and rare varia- autism research: A cross-sectional review and meta-analy- tion act additively to create risk for autism spectrum sis. Molecular Autism, 10,9. disorders. Nature Genetics, 49, 978. Rutter, M. (1968). Concepts of autism: A review of research. Wester Oxelgren, U., Aberg, M., Myrelid, A., Anneren, G., Journal of Child Psychiatry & Psychology, 9,1–25. Westerlund, J., Gustafsson, J., & Fernell, E. (2019). Autism Rydzewska, E., Hughes-McCormack, L.A., Gillberg, C., Hen- needs to be considered in children with Down syndrome. derson, A., MacIntyre, C., Rintoul, J., & Cooper, S.A. (2018). Acta Paediatrica, 108, 2019–2026. Prevalence of long-term health conditions in adults with Wing, L. (1969). The handicaps of autistic children—A com- autism: observational study of a whole country population. parative study. Journal of Child Psychology and Psychiatry, British Medical Journal Open, 8, e023945. 10,1–40. Schopler, E., & Mesibov, G.B. (1983). Autism in adolescent and Wing, L. (1981). Asperger’s syndrome: A clinical account. adults. New York: Plenum Press. Psychological Medicine, 11, 115–119. Schuwerk, T., Jarvers, I., Vuori, M., & Sodian, B. (2016). Wing, L., & Gould, J. (1978). Systematic recording of behaviors Implicit mentalizing persists beyond early childhood and is and skills of retarded and psychotic children. Journal of profoundly impaired in children with autism spectrum Autism and Childhood Schizophrenia, 8,79–97. condition. Frontiers in Psychology, 7, 1696. Wing, L., & Gould, J. (1979). Severe impairments of social Sedgewick, F., Kerr-Gaffney, J., Leppanen, J., & Tchanturia, interaction and associated abnormalities in children: Epi- K. (2019). Anorexia nervosa, autism, and the ADOS: How demiology and classification. Journal of Autism and Devel- appropriate is the new algorithm in identifying cases? opmental Disorders, 9,11–29. Frontiers in Psychiatry, 10, 507. Wing, L., & Wing, J.K. (1971). Multiple impairments in early Shah, A. (2019). Catatonia, shutdown and breakdown in childhood autism. Journal of Autism and Childhood autism. A psycho-ecological approach. London: Jessica Schizophrenia, 1, 256–266. Kingsley. Wolff, S., Narayan, S., & Moyes, B. (1988). Personality char- Shakespeare, T. (2017). The social model of disability. In L.J. acteristics of parents of autistic children: A controlled study. Davis (Ed.), The disability studies reader (5th edn, pp. 195– Journal of Child Psychology and Psychiatry, 29, 143–153. 203). Abingdon, UK: Routledge. World Health Organization (WHO) (1990). International statis- Sheldrick, R.C., & Carter, A.S. (2018). State-level trends in the tical classification of diseases and related health problems prevalence of Autism Spectrum Disorder (ASD) from 2000 to (ICD) - 10th revision. Geneva, Switzerland: World Health 2012: A reanalysis of findings from the autism and develop- Organization. mental disabilities network. Journal of Autism and Develop- Wu, S., Wu, F., Ding, Y., Hou, J., Bi, J., & Zhang, Z. (2017). mental Disorders, 48, 3086–3092. Advanced parental age and autism risk in children: a Silberman, S. (2015). Neurotribes: The legacy of autism and systematic review and meta-analysis. Acta Psychiatrica how to think smarter about people who think differently. Scandinavica, 135,29–41. Sydney, NSW: Allen & Unwin. Zheng, L., Grove, R., & Eapen, V. (2019). Spectrum or Stewart, M.E., Griffiths, T.D., & Grube, M. (2018). Autistic subtypes? A latent profile analysis of restricted and repet- traits and enhanced perceptual representation of pitch and itive behaviours in autism. Research in Autism Spectrum time. Journal of Autism and Developmental Disorders, 48, Disorders, 57,46–54. 1350–1358. Zivrali-Yarar, E. (2017). Ageing and autism spectrum disorder. Tager-Flusberg, H., Plesa Skwerer, D., Joseph, R.M., Unpublished PhD thesis, King’s College London. Brukilacchio, B., Decker, J., Eggleston, B., ... & Yoder, A. (2017). Conducting research with minimally verbal partic- ipants with autism spectrum disorder. Autism, 21, 852– Accepted for publication: 12 November 2019 861. First published online: 28 January 2020

© 2020 Association for Child and Adolescent Mental Health