Discoveries in autism research, as in all scientific endeavours, are determined not just by the data that we scientists observe but by the questions that we ask. Empirical science is inseparable from such rhetorical scaffolding, which simultaneously supports and limits both the data that we decide to collect and the interpretations that we impose on these data. It's natural for us to demand: What is autism's biological basis? What are the genes that cause autism? What distinguishes autistic brains and autistic cognition from typical brains and typical cognition? Implicit in this rhetoric are the assumptions that autism has a singular biological basis, that autism susceptibility is conferred by individual genetic variants, and that autistic neural and congitive processes differ categorically from typical ones. These assumptions make sense within a twentieth-century scientific frame in which singular causal factors (single genes, focal brain lesions, circumscribed cognitive deficits) manifest as disease states that are unitary and entirely distinct from healthy states, and in which contributing factors are monotonic in their relation to disease outcome and linear in their interactions with each other. An accumulation of evidence points, rather, to autism as a nonlinear, network-level perturbation of the same developmental processes that underlie normal human cognitive diversity. Thus the more closely we examine autism, the more we find ourselves gazing into a mirror.

A current model posits that autism arises when differences in local neural connections within small regions of the brain perturb the development of long-range connections between widely separated areas, altering the brain's ability to integrate information from many sources at once. Many different biological causes can contribute to this perturbation of neural connectivity - in much the same way that many different mechanical changes can cause an engine to slow or to speed up - and each of these contributing biological causes can induce its own signature set of consequences within and beyond autism's core symptoms. It's commonly said amongst caregivers and therapists that "If you've seen one person with autism, you've seen ONE person with autism." People with autism are at least as diverse as people without autism, and the diagnosis shouldn't mask such differences. All too often, potentially treatable autism-associated conditions such as attention deficit, gastrointestinal disturbance, or sleep disorders are dismissed as "part of the autism," and people with autism never receive the full evaluation and treatment that they deserve. This diversity in biological symptoms is matched by a diversity of cognitive profiles and learning styles: some people on the autism spectrum think in pictures, others think in sounds; some can force eye contact, others find direct gaze too distracting and anxiety-inducing; each individual's attention focuses intensely on special interests that are idiosyncratic and often unique. Clinical and therapeutic strategies are most likely to succeed when they are individualised and client-centred, taking advantage of each client's skills and interests as conduits to motivate and to deliver therapeutic interventions and educational content.

As finite creatures trying to comprehend an infinite world, human beings operate by drawing boundaries on that world, replacing an infinity of sense objects with a finite, manageable number of perceptual and cognitive categories. Through this process of construal we gain the power to make inferences with a single thought about entire classes of objects and events, but we also deprive ourselves of the power to experience these objects and events as they are. For trees we see forest — or for this tree with its own pattern of branches and leaves which will never come again, we see only a tree. What we end up 'seeing' or remembering is not the iconic thing-in-itself but rather a perceptual schema or script into which the thing has been inexactly projected or pigeonholed; thus our attempts to narrate the world also falsify the world. This narrative representational capacity to construe abstract symbol in place of iconic percept depends fundamentally on the brain's ability rapidly and flexibly to link sense impressions with categories: narrative connectivity demands neural connectivity. In the brains of people with autism spectrum conditions, this underlying neural integration is disrupted, rendering separate brain regions - and separate perceptual and cognitive subsystems - unusually autonomous. People with autism then develop a cognitive style based on access to raw percepts, often at the expense of immediate and rapid access to abstract representations. The resulting compulsion to attend to things in themselves, rather than things construed, underlies autistic superiorities in perceptual disembedding, exact memory, and detail-oriented analysis of deterministic, rule-based systems, and also autistic deficits in social exchange and theory-of-mind (suppressing one's own true perspective so as to construe the perspective of a social partner), language (rapidly and flexibly substituting actual objects and events with construed lexical and syntactic representations), and behavioural flexibility (mapping construed experiences to contextually appropriate behavioural responses). Both in terms of cognition and in terms of neurophysiology, these characteristics of autism in many aspects present not a categorical difference but rather an extreme of human cognitive diversity: different cultures, different sexes, different people, and even the same person in different behavioural and cognitive contexts, exercise different degrees of construal underlain by different degrees of integrative or autonomous brain function. This continuity between autistic and non-autistic variations in brain function ought perhaps to come as no surprise, given that so many of the individual genetic variants associated with autism occur commonly within the population. Thus the science of autism shows us reflections of ourselves, shows us how the contest between fractionated percept and unified construal is a property of the finite and limited physical body and brain within which the self is manifest, and allows us to glimpse the promise of transcending this contest at the boundary between the mortal and the eternal.

The straightforward way to study autism is to confront the obvious deficits: What genes or what parts of the brain underlie autistic behavioural impairments? This approach falls short in at least two ways. First, in focusing on autistic deficits it fails to account for autistic superiorities; in fact, one and the same physiological and cognitive alteration may explain both autistic weaknesses in social communication and behavioural flexibility, and autistic strengths in perception of detail, exact memory, and understanding of explicit and immediate rules. Second, this approach from deficits assumes that autistic behaviours are produced directly by biological differences in the genes and/or in the brain; in fact, many of the things that people with autism do are most productively construed as adaptive responses, the reactions of a normal human mind maturing within an abnormal perceptual and cognitive environment. We all adapt within the constraints of our bodies and brains, and we all develop strategies that capitalise on our strengths and minimise our weaknesses. A child temporarily deprived of sight in one eye, for instance, will have learnt to rely on the intact eye; to force development of the bad eye, the good eye must be covered with a patch. Likewise, a person who has suffered a stroke will learn to rely on the good side of the body, and must be forced therapeutically to exercise the bad side. The situation in autism is less visible because it involves strengths and weaknesses not simply in overt sensory or motor skills, but in perceptual and cognitive skills: people with autism spectrum conditions learn to rely the capacities at which their brains excel, at the expense of skills that don't come as easily to them. Every human brain has a balance between local connections within single regions of the brain, and long-range connections between distant regions. More short-range connections may make individual parts of the brain very efficient at at working autonomously and independently, but correspondingly difficult to coordinate with each other. Autism seems the extreme case of such an altered balance, in which separate brain systems operate independently and in parallel, but with little or no coordination. This re-conceptualisation of autism as an extreme case of normal human cognitive variation opens the way toward new therapies for autism, and to a new understanding of ourselves: The more closely we examine autism, cognitively, genetically, and physiologically, the more we find ourselves gazing into a mirror.