Woah, I really need to drag up that whitepaper I did about 5 years ago. Here's the basic premise of it, until I can find it again: Autism is a pretty rough communication disorder, or so it's approached. It seems to have a pretty strong genetic component, and perhaps some minor environmental components. Much research needs to be done on both, I suspect it's a kind of genetic predisposition plus stress model thing like so many other disorders- or at least stress in just how bad it can get. Plus once it starts the stressors elevate, so you have a sort of exponential possibility on damage. This is all marshmallow-y stuff though. Stresses can come in many forms, noted by this study is fever: increased rate of chemical process catalyzation in parts of the brain. I did not propose such a thing, but it certainly is interesting.
What's more interesting, to me, is the location of the brain studied, and how that links back to my original proposition. The locus coeruleus is thought to be part of the system we have for sorting out sensory information, and this is where the trains collide: Take a moment to look around while driving down the street (safely, of course). You're very good at picking out which information to register; which cars are near you, what color the traffic light is, what your speedometer says. It's really amazing how we sort data so quickly, down to a handful of items worthy of direct attention. This is millions of years of evolution at work to help you decide that the tiger is much more importaint than the leaves in the trees.
Suppose for a moment you lost this ability- no way to tell which is more important. The texture of the asphalt, the sway of the grass, it's all just the same as the brake lights of the vehicle directly in front of you. There's really no way a person could drive like this. I suspect this is a large component in autism. Just as a tone deaf person has trouble picking a single voice out of a noisy room, observations made it seem quite likely that a severely autistic child has great difficulty grasping a single word out of the slurry of information that is overwhelming their brains at any time. Makes it pretty hard to learn language properly, or make friendships. In time a person would simply start to regard most of the sensory data as the common white noise, like the ringing in my ears, and internalize all else.
Result? Depends on the person. Some will find certain moments to try and press their sensory intake to. Physical textures, wiggling objects, etc. Others will try and bring external things into their world in unique ways, like synaesthetes (recently, famously, Daniel Tennet). All of these things are very common to autistic persons.
Mehler's study seems to point to this exactly, but with the right amount of caution- it's a hint at what could be structures of autism. And it's fortunately in places that can be dealt with. Unfortunately, as per Lenangred, Chompsky, etc, children who did not learn language early enough in life may never quite get there anyway. Developmental damages are a tough nut.
The hope is that a neurobiological theory can result in some help for those already suffering, and help prevent suffering altogether for those to come.