Autism as a diagnosis creates a dichotomous distinction: to have autism or not, from what is essentially a multi-dimensional condition. Each dimension is a continuous scale from mild symptoms in the general population through to severe cases. Bentall [2] distinguishes this dimensional nature of psychopathology as follows:

“Abnormal behaviours and experiences are related to normal behaviours by continua of frequency (the same behaviours and experiences occur less frequently in non-psychiatric populations), severity (less severe forms of the behaviour and experiences can be identified in non-psychiatric populations) and phenomenology (nonclinical analogues of behaviours can be identified as part of normal life).”

Today, autism is conceived as a true spectrum, where autistic traits have a normal distribution in the general population and an arbitrary cut-off point determines who is considered to be on the spectrum and who is not. Constantino and Todd [2] measured autistic traits in a large community sample, and found no discontinuity between normality and psychopathology, as would be evidenced by bimodal distribution. These findings were repeated in a Scandinavian study [3]. Individual symptoms of autism occur in the general population and appear not to associate highly, without a sharp line separating pathological severity from common traits [4]. Our own work suggests the symptoms or behaviours that underpin autism have a normal distribution in the population [5,6], and a cut-off or threshold determines which children are given the autism label or not (Figure 1). The exact position of the cut-off point has been culturally and historically determined. The imposition of a cut off between normality and abnormality is therefore ‘an arbitrary but convenient way of converting a dimension into a category’, as Goodman and Scott [7] point out.

There are several studies that have sought to uncover what the dimensions that underlie autism are, best known in the form of the triad of impairments [8]. Results from factor analysis have shown between two and seven dimensions in autistic behaviour, encompassing social-communication deficits, restricted interests, and repetitive behaviour, lack of role play and hypo-or hyper-sensitivity [9]. The DSM-5, meanwhile, now considers just two dimensions underpin the condition: social communication impairment and repetitive behaviour/ sensory issues, a change from the familiar triad of impairments listed in previous versions.

Figure 1: Threshold for diagnosis of ASD.

The diagnostic criteria assume that there is some underlying deficit or pathology that underlies autism. But the boundaries around the symptom clusters that are considered to lie within this class have shifted with each successive revision of the DSM. ‘Autism’ is therefore what the DSM says it is. Each revision to diagnostic criteria has led to a slightly different set of children being classified as having autism. Modern definitions are more inclusive; that is, more children are classified with the diagnosis, and this is one reason why the prevalence of autism as diagnosed has shot up over the last 20 years as many studies have shown [10]. It is also possible that increased autism rates are due also to newer environmental risk factors, as alluded to in DSM-5 [11]:

“It remains unclear whether higher rates reflect an expansion of the diagnostic criteria of DSM-IV to include sub-threshold cases, increased awareness differences in study methodology or a true increase in the frequency of autism spectrum disorder”.

Children with an autism diagnosis, that was thought to be rare 40 years ago, are now frequently encountered in the mainstream classroom. One reason is that the axes of behaviour that underlie autism, those multi-dimensions or factors, are iteratively determined by the new diagnostic criteria, and these have shifted over time to include less severe cases. ‘Autism spectrum disorders’ becomes whatever behaviours at whatever severity we choose to put the boundary around. In this way the diagnostic criteria define the phenotype for research, and research defines the revision to the criteria. This circularity means that any disorder will always be a product of the time in which it is defined.

A second criticism is that the nature of what is considered ‘disordered behaviour’ is to some extent a reflection on the values of society at any time point. So lack of social skills only becomes a problem in a context where having good social ability is paramount. Some have argued it is the context the educational institutions that has driven the rise of Asperger’s syndrome as a diagnosis [12].

Today’s criteria for autism diagnosis can be seen as pragmatic. The diagnostic criteria themselves and their classification into the larger picture are always in a state of flux. As Goodman and Scott point out:

“Fashion continues to be important in classification, and there are likely to be minor and major revisions in the schemes for many years yet ….Our current ideas are like maps of largely unexplored territorybetter than nothing provided you don’t take the details too seriously”

However, when used in practice the diagnostic criteria are often presented as scientific ‘truths’. Through the process of diagnosis, the classification system remains largely invisible to parents and patients. From a parent’s point of view, it is worth remembering that the criteria that underpin a child’s autism diagnosis do not exist in a vacuum. The DSM can be seen as a tool for categorization contingent on the historical and moral values of the time.

This work was funded by the UK Economic and Social Research Council (ESRC) as part of the Secondary Data Analysis Initiative.