Autism is a developmental disorder in which “theory of mind,” that is, the ability to represent others’ beliefs and feelings as distinct from one’s own, is disrupted. People with autism have trouble taking the perspectives of social partners, and acting on the reality that other people may want different things and focus on different goals than they themselves do. This failure of theory-of-mind is diagnostically and clinically one of the most significant disabilities in autism — but is it the most significant aetiologically? Or might theory-of-mind skills, and theory-of-mind deficits, emerge from more fundamental skills and deficits during the process of development?

At a neural level, attempts to pin autism to a single theory-of-mind region or subsystem within the brain or to a single cognitive module have been unable to capture the full range of autistic cognition, behaviour, and neurophysiology, because these models encompass only one sort of perspective-taking: in addition to the theory-of-mind deficit in taking others’ social perspectives, people with autism experience deficits in taking temporal perspective (what will happen next year versus what’s happening now), spatial perspective (what’s happening around the whole room versus what’s happening in front of me), and counterfactual or conditional perspective (suppose event X occurs, what happens then?). Our and others’ experimental work has shown that all these social and non-social perspective-taking activities are linked, psychologically and neurally, in people with autism spectrum conditions as in people in general.

What all these social and non-social perspective-taking abilities seem to have in common is a dependence on the rapid and automatic integration of many brain regions and cognitive subsystems — in effect, a domain-general ability to connect many separate, simultaneous channels of perceptual and cognitive events into an ordered structure: in a word, a narrative. This internal narrative ability depends not on any particular part of the brain, but rather on the degree to which widely separated parts of the brain connect with each other. Rather than a computationally and neurally local phenomenon, mental narrative is a network phenomenon — and narrative connectivity depends on neural connectivity. This neural and narrative capacity is not inherent, but must be developed. Its emergence relies on proper interactions between individual brain regions — interactions that can be guided by participation in understanding stories that link characters’ goals and beliefs to actions. This recognition of theory-of-mind as a developmental, emergent phenomenon opens a route to education and therapy.

A disconnected, autistic brain cannot bring separate channels together: one is unable to connect simultaneous actions and motivations, to connect one’s own perspective with the simultaneously represented perspective of a social partner, to connect a thought simultaneously with its expression in words, to connect an exogenous percept simultaneously with an endogenous concept, or even to connect simultaneous visual and auditory events. The key limitation is that of simultaneity: a person with autism can represent action or motivation, egocentrism or allocentrism, thought or syntax, external perceptual or internal mental events, and sights or sounds — just not at the same time. One possible workaround, then, involves removing the demand for simultaneity: what if the acts of knowing and imagining, self-representation and other-representation, thinking and phrasing, perceiving and conceiving, seeing and hearing didn’t have to occur at the same time? In other words, what if communication could take place asynchronously?

Typed text, unlike conversational speech, is an asynchronous communication medium: there is no demand to time utterances to maintain the flow of conversation, no demand for precise temporal sequencing of movements of the throat, tongue and lips, and no demand to perceive the spoken word in real time. The partially composed text itself serves as an external memory; thus there is no requirement to hold in mind the expression of a thought at the same time as one retains the thought itself. One can leave off in the middle of a word, interrupt, then resume. One can attend to the text and to other stimuli in turns. Thus the narrative structuring inherent in this textual medium can in some measure substitute for the lack of narrative connectivity in the autistic brain. In a study of typed communication as a therapeutic medium for children with autism who lack communicative speech, we have found that the textual medium increases communicative capacity. Also, tellingly, direct gaze at a subject of communication can actually decrease communicative capacity — a case of a salient percept interfering with a nascent concept. The same holds, more subtly, even for individuals outside the clinical autism spectrum, who manifest only subtle degrees of social communicative impairment compatible with typical development. These results suggest that the conceptualisation of autism as a deficit in the emergence of neural — and therefore narrative — connectivity is a productive one not only for theory but also for therapy, and that the act of narrative sequencing and perspective-taking in general may spur the development of the social perspective-taking that constitutes theory-of-mind.