DSM-5 has been conceived as both conservative and progressive; it attempts to keep important diagnostic traditions intact while it introduces clinically important paradigmatic shifts. Rejecting a categorical understanding of autism (with its all-or-nothing approach to diagnosis) and replacing it with a dimensional model is a considerable improvement over DSM-IV. It is understood in both DSM-IV and DSM-5 that children who meet the criteria for autism often have symptoms of other disorders (anxiety disorders, affective disorders, attention deficit hyperactivity disorder, specific language disorders, and intellectual disability in particular). The presence of these other symptoms should be noted, and, if necessary, the patient should be treated. If, however, the symptoms are sufficient to meet criteria for other disorders, then the patient should be diagnosed as having two or more disorders. It is good that DSM-5 allows for this, but the popularity of this practice reflects the still limited understanding we have of pathophysiology in autism and other syndromes defined on purely behavioral grounds.

While many things about autism, such as key diagnostic features, remain similar in DSM-5, other aspects of the diagnosis involve major changes in DSM-5. The name of the broader category in which autism is situated is no longer pervasive developmental disorder; it is autism spectrum disorder (ASD). In fact, ASD encompasses a number of syndromes that were individually classified in DSM-IV; Asperger's disorder and pervasive developmental disorder not otherwise specified are gone. Whereas DSM-IV included three sets of symptoms for autism, in the draft version of DSM-5, social and communication problems have been melded to form “social/communication deficits” and “fixated interests and repetitive behaviors.” The criteria for both sets must be met for the diagnosis of ASD to be made.

Although DSM-5 uses the term “spectrum” to characterize the disorder, one might ask: a spectrum of what? The most immediate suggestion is that the autistic spectrum captures variations on social communication problems. In the initial autism literature, twin and family studies showed that symptoms of “socially odd behaviors” were more likely to be observed in the relatives of people with autism than in the general population, suggesting a spectrum of social communication deficits. When social communication itself was revealed to be a developmental progression that begins at birth and continues into adulthood, autism was better understood as an arrest or deviation in this developmental trajectory. The normal development of social communication has been thoroughly described in many studies over the past 30 years (1).

Infants come into the world with a behavioral propensity to look at complex rather than simple visual stimuli. The face of an infant's mother (or primary caretaker) is a complex stimulus, and the newborn will look at it more and more over the first few weeks. Soon an infant learns to engage in affective reciprocity (i.e., to recognize and read early social signals mediated through facial expression, prosody, gesture, and body language, not only with his or her mother but with an expanded range of caretakers). In the second year of life, the child develops joint attention skills; by 15 months, the child will check the mother's facial expression or tone of voice to learn what to do in an ambiguous situation. This is observed at a time when the child has little spoken language.

By 3 or 4 years old, children begin to show an understanding of the thoughts and motives of others (theory of mind and mentalizing) and learn to use this knowledge to govern social interactions. Their play is marked by the imitation of activities that they see adults around them doing or that they see on TV. Through group play, children learn even more how to read the intentions of others. By adulthood, each person should have developed a mastery of social knowledge and skills, but any arrest or deviance in this process may result in a social communication abnormality. We recognize this abnormality as being on the autism spectrum. People with profound autism may scarcely master the earliest stage of affective reciprocity; those with mild autism may have good affective reciprocity and joint attention skills, but they may have poor mentalizing and pragmatic skills. They are able learn factual information (often of an arcane nature) and to lecture endlessly on it, but they may not understand that they should tailor their lecture to their listener's interests and may be puzzled if their listeners turn away or look askance. So, where exactly does the boundary lie between autism spectrum disorder and merely peculiar social behavior?

Twin studies (2, 3) have shown that the capacity to master social communication is genetically influenced, and epidemiological studies show that this capacity is distributed in a continuous fashion in the general population, but results from these studies raise further questions about the boundaries separating illness from the extremes of normality. People whose social communication skills fall on the extreme abnormal end of this continuum can readily be diagnosed, using “gold standard” clinical instruments, as meeting the criteria for autism. Our understanding of autism as a spectrum disorder is reinforced because social and communication skills have a continuous distribution. This distribution also suggests how, in future iterations of DSM, we might deal with the severity issue. First, identify suitable diagnostic instruments for the measurement of social communication skills and determine the distribution of social communication scores in the general population using statistics such as the mean and standard deviation. Second, assign cutoff points for severe and moderate autism spectrum disorder that are chosen using clinically appropriate criteria. The diagnostic instruments would vary depending on the age of the child, although the cutoff points should not. It would even be possible to rename moderate autism spectrum disorder as Asperger's disorder, satisfying those who decry the disappearance of this diagnosis from the DSM-5. Lesser degrees of social and communication disorder might not need labeling, although the presence of a mild impairment could be noted. The question of where to place the boundary of social oddity and mild autism could be controversial, and it will require considerable discussion by expert clinicians, parents, and persons with mild autism themselves. We may need a requirement that the symptoms result in a “current impairment in adaptive functioning” (the language used in DSM-IV to define mild mental retardation) to define the diagnosis of mild autism.

While accepting the social communication spectrum understanding of autism, some have suggested that we should try to go behind the social phenomena to see if we can understand why social communication skills are poor. They propose that certain failures in executive functioning could lead to the social communication deficits observed in autism. Executive functioning encompasses a number of phenomena such as central coherence, working memory, and cognitive inhibition. I am not optimistic that executive functioning research will soon lead to an understanding of ASD; there is too much of a gap between the specific cognitive skills defined by executive functioning and the social deficits of autism. Yes, people with autism do fail to see the big picture in social situations (failure of central coherence in executive functioning terms), but knowing this does not provide me with useful clinical or treatment information. Someday when the database of executive functioning has advanced, this knowledge may be of greater help.

My personal preference, then, is to opt for a failure of “social communication development” as the core factor for the disorder. Clinically, I find that parents, educators, and colleagues find the explanation that autism is a disorder in which the unfolding of “social communication” skills has been arrested or attenuated is reasonable and makes sense. It is what the parents have reported to me in the Autism Diagnostic Interview–Revised, it is how they can understand what may be wrong, and it is what informs treatment. So, for me, the model works.

A DSM-5 diagnosis of ASD also requires that the specific fixated interests and repetitive behaviors criterion be met. Although these symptoms have been required for the diagnosis of autism from DSM-III onward, there have been few in-depth studies of these behaviors. When they have been studied, it is often assumed that they represent a common factor, but this may not be true. Studies have shown that fixated interests and repetitive behaviors are frequently present in people with autism and Asperger's disorder. But three factor analysis studies (4–6) suggest that there may be two categories of fixated interests and repetitive behaviors: one associated with “insistence on sameness” and the other with repetitive sensorimotor behaviors. If someone has social communication deficits and poor mentalizing ability, then he or she might be expected to wish that social and environmental factors remain predictable. Sensorimotor repetitive behaviors tend to be correlated with less-developed intellectual skills, specific language deficits, and younger age. To demand that sensorimotor fixated interests and repetitive behaviors be present for the diagnosis of ASD could exclude many people who do have significant social communication symptoms but lack the sensorimotor symptoms. However, there needs to be confirmation of this two-factor model or identification of a better model.

There is a final, and rather unexpected, feature of DSM-5 on which I must comment. There are two diagnoses in DSM-5 that appear quite similar: autism spectrum disorder and social communication disorder. The latter is listed in the new diagnostic category of language impairment. Persons with social communication disorder have an “impairment of pragmatics” and impairment in the “social uses of verbal and nonverbal communication.” The presence of fixated interests and repetitive behaviors is required for ASD, but it is an exclusionary factor for social communication disorder. In the past 20 years, social communication disorder has been studied extensively by speech and language specialists. My review of this literature convinces me that ASD and social communication disorder are the same disorder, except for the required DSM-5 presence of fixated interests and repetitive behaviors in ASD. I wonder if social communication disorder was included as an attempt to better define mild autism. However, to introduce a fairly similar category to autism spectrum disorder, more than 60 years after Kanner's and Asperger's landmark publications, could be clinically disruptive. The diagnosis of autism (i.e., pervasive developmental disorder in the past three versions of DSM) has long been incorporated into a network of official research, public health, school, and reimbursement systems. How will these systems deal with ASD and social communication disorder? There is no discussion of this in DSM-5, but perhaps the field trials will illuminate any conflicts.