Whereas I've attempted to answer some questions I pose definitively, I will not being doing so with this question. I intend merely to forward some speculation that is informed by a variety of sources. As a learning theorist, I've always been fascinated by imprinting. Imprinting is a window into the evolutionary history of the species as it expresses itself during the development of the individual. Imprinting has been most studied by learning theorists in birds (e.g., Lorenz, 1937). When the hatchling pecks their way out of the egg, the first moving object, typically a parent, will result in the hatchling moving towards that object. The hatchling (or more precisely the hatchlings of the species) has been prepared by natural selection to respond to the moving parent. Imprinting establishes the parent as very important, a potent reinforcer, and a significant cue.
Part of why we can say this is that imprinting sometimes doesn't happen or imprinting goes awry in some way, like the parent is not around and another moving object produces imprinting. Though there are many instances in which the young imprinted to him, Lorenz was fond of trains. He provided for small electric trains to be present when the young hatched. Those that encountered the trains upon hatching followed the trains. When imprinting goes awry in this manner, the hatchling moves toward the imprinted stimulus as hatchlings that imprint to their parents move towards the parent. Some other experimenters have shown that, after imprinting, the young will emit various responses to remain close to the imprinted stimulus. One experiment even showed that walking backwards is quickly learned if that brings the imprinted stimulus closer.
Though we can certainly question whether there are close parallels between imprinting in birds and social learning in humans, it is not that far afield to consider the possibility. Hoffman (1996) explored the development of social attachment in birds in a somewhat technical but engaging book titled, Amorous turkeys and addicted ducklings: A search for the causes of social attachment. Human infants are also prepared by natural selection. There are a number of reflexes that are critical to survival present at or before birth. Breathing, sucking, swallowing, and rooting are just a few of these primitive reflexes that I like to refer to as prepared responses. I've worked quite a lot with feeding and I've spent a good bit of time studying swallowing. The swallowing of the infant changes very quickly. Initially the suck is very strong and the substance flowing from the breast is more dense. However, shortly after the first few feedings, the milk comes in less dense. Sucking becomes less forceful/effortful partly due to the change in density and is likely in tune to the changing consequences following expression of milk from the breast. Reflexive responses bring behavior important for survival into contact with environmental contingencies. This is one way in which natural selection prepares the species to meet a dynamic and ever-changing environment. This is somewhat simplified and I do not intend to equate breastfeeding with "typical" development or imply that there is any relationship between breastfeeding and the subsequent development of an ASD. The main point is to show how early responses lead to more complex behavior as development occurs.
That said, one of the earliest indicators that an ASD may be present is atypicality, at birth, in primitive reflexes, like rooting and sucking. As I mentioned above, imprinting establishes the significance of a cue and if that cue is not imprinted to, it does not have this same significance. Is it possible that something is going wrong in social learning that is akin to an imprinting error? Well, one thing we know about individuals with autism relative to people without it is that social cues do not hold the same significance for people with ASDs. The Yale Child Study Group has produced a number of interesting studies on this topic. They have noted that high-functioning persons with an ASD do not attend to social cues in the same manner as other do. They also have recently conducted research that suggests that children, who are subsequently diagnosed with an ASD, as young as 6 months old will look at mouths rather than eyes. I could take a rather lengthy side track of speculating about why mouths and not eyes but I'll save that for another post.
Social learning may be set off course by a decreased salience of social cues (though we should note that social deficits are not readily apparent until at least 6 months of age). Is this due to defective stimulus control in early learning? I fully acknowledge that this is post hoc analysis.
But that is only part of the story of imprinting. During learning, physiological changes are occurring and some interesting developments have been breaking recently that might be related to social learning. Some research has suggested that abnormalities in some hormones, oxytocin and vasopressin in particular, may be related to certain neurobehavioral disorders, including ASDs (e.g., Tobin et al., 2010 in Nature). Moreover, one study suggests that providing oxytocin to high-functioning individuals with an ASD produced more affiliative behavior including higher levels of looking at faces when free access to them was provided (Andari et al., 2010 in the Proceedings of the National Academy of Sciences). Such hormones have been implicated in social behavior in numerous contexts, including bonding during infancy. While certain cues important to social development are present, it might be the case that abnormalities in the hormones present in the body might result in learning not being "stamped in" so to speak, in the same way for individuals that will be diagnosed with an ASD.