A wealth of evidence from brain physiology, anatomy, and genetics suggests that autism involves abnormal connectons between neurones within localised regions of the brain, and that this abnormal local connectivity is a trait that can run in families. Long-range connections between widely separated neural networks normally develop as a consequence of patterned neural activity within those networks: so if local neural activity is perturbed then, in susceptible family members, disruptions in long-range connections can arise and produce autism spectrum conditions. This disruption of long-range connections gives rise to a brain that operates more as a collection of autonomous sub-processors than as an integrated unit under central control, and thus to a cognitive style featuring excellent perception of small, verbatim sensory details, and of the world as one sees it oneself, but often a complementary lack of attention to context, abstraction, and the world as others may see it. Many distinct biological differences can perturb local and long-range neural connectivities in this way, and therefore autism from a biological standpoint seems not one condition, but rather a wealth of genetic and environmental influences that combine with each other in different measures in different individuals. This difficulty of pinning autism to a single biological cause is one reason that the only therapies proven successful are behavioural, not biological. Although drugs can address autistic symptoms, current drugs ought never to be used as a sole or main therapy: the proper role for drugs is to quiet interfering symptoms so that an individual gains access to behavioural and educational therapies. As autism becomes fractionated into distinct biological causes, pharmacological and other biological therapies will be developed to address these root causes rather than only the symptoms; for instance, drugs to treat autism secondary to Fragile X syndrome will likely become available within the next few years.