Andrew Shaner: UCLA Semel Institute for Psychiatry and Biobehavioral Sciences; Deputy Chief of Psychiatry and Mental Health, VA Greater Los Angeles Healthcare System

In many species, siblings compete for parental care and feeding, while parents must allocate scarce resources to those offspring most likely to survive and reproduce. This could cause offspring to evolve traits that advertise health, and thereby attract parental resources. For example, experimental evidence suggests that bright orange filaments covering the heads of North American coot chicks may have evolved for this fitness-advertising purpose. Suppose that the ability of infants and very young children to charm their parents evolved as a parentally selected fitness indicator. Young children would vary greatly in their ability to charm parents, that variation would correlate with underlying fitness, and autism could be the low-fitness extreme of this variation.
This general version of our hypothesis can explain why autism begins in childhood, why it is highly heritable, why the responsible genes have been so hard to find, why it is more common in boys and more severe in girls and why it is associated with environmental hazards, developmental abnormalities and increased mortality. Among its predictions is that autism will be more common in populations with historically high rates of genetic polyandry.
In addition to the general hypothesis, suppose that a key component of charm involves infant social behaviors that prolong breast feeding and thereby delay conception of a younger sibling. If true, this would explain why autism impairs social abilities so early and so profoundly. It would also predict that (1) within populations, age at onset of autism will parallel age at onset of weaning, (2) autism will be associated with scarce environmental resources and early weaning, (3) delaying weaning will protect against autism (4) close relatives will show higher variance in infant social ability (including its anatomical and neurophysiological bases), and in subsequent birth interval, and (5) infant social ability will correlate positively with both underlying fitness and parental resource allocation (e.g., intensity and duration of breast-feeding).
http://bec.ucla.edu/papers/Shaner08.pdf