In 2004, taking stock of behavioural, neurophysiological, neuroanatomical and genetic findings on autism, we proposed the thesis that autism arises as an abnormality of neural connectivity, in which synaptic hyper- or hypo-connectivity produces a pathological decrease in entropy, or information capacity, within local neural networks. This abnormal developmental starting point for local neural information processing then interacts with normal programmes of activity-dependent brain and cognitive development, producing more and more abnormal long-range information transfer between brain regions and between cognitive subsystems. The result is a piecemeal cognitive style in which information processing tends to occur locally rather than in anatomically and functionally distributed, interacting networks of brain regions — a style that is low in perceptual construal and psychological distance, detailed rather than contextual, unimodal rather than integrative, iconic rather than symbolic, egocentric rather than allocentric. In the intervening decade, results at all levels of investigation have largely supported this notion of autism as a difference in neural connectivity, the major empirical modification to the theory being a refinement from simple dis-connectivity to a more nuanced pattern of developmental dys-connectivity. Our physiological research with non-autistic siblings of children with autism spectrum conditions has demonstrated that whereas autistic and non-autistic siblings share a prolonged style of attention-related prefrontal activation which correlates with autistic social communicative traits, only the autism-spectrum siblings manifest large degrees of functional dysconnectivity between brain regions — suggesting that familial liability factors need not always result in autism, and that behavioural or pharmacological interventions on the horizon have the potential to prevent or even to remediate the secondary, activity-dependent development of brain dysconnectivity. To be most effective, evidence-based behavioural treatments must work with — rather than against — this dysconnected autistic cognitive style, and must address fundamental capacities of cognitive and motor control and sensory perceptual modulation, as well as autism's more immediately obvious deficits in social interaction.