Read the most advanced science of autism causes here. Bypass the commenterati and go direct to the science. Don't waste your time at the sites which pretend "no-one" knows what causes (or what sometimes cures) autism.
This is a website relating to the unchallenged theory of autism, IQ and genius, Personality and Individual Differences 14:459-482 (1993) by Robin P Clarke (the antiinnatia theory). An update review paper is being prepared for publication. Meanwhile you can download the original 1993 publication (presentationally revised) here, and the original 1993 publication (author's reprint) here . (the journal site version is here: http://dx.doi.org/10.1016/0191-8869(93)90316-U, but without added charts of social class and you may have to pay Elsevier $31.)

Abnormalities of the "resting network" and of language lateralisation

In my most recent draft of the update review of the theory, I wrote:

Pisula (2010) lists numerous areas of the brain which are known to be normally (and by implication innately) associated with specified psychological functions, and which function abnormally in autism. She notes that these findings cannot be adequately accounted for in terms of “theory of mind” or “executive dysfunction” or “lack of central coherence”. But they all rather obviously fall very clearly within the concept of innatons being affected by excessive antiinnatia. So Pisula’s review can be re-read in retrospect as even further testimony to the empirical soundness of the antiinnatia theory...."

I'd now add to that list two more instances of a similar kind.

Firstly the "resting network", which in neurotypicals activates while not concentrating on a task, and becomes inactive when task-engaged. Whereas in autistics it tends to just have a similar level of activity in both circumstances. Ref: Failing to deactivate: Resting functional abnormalities in autism. Daniel P. Kennedy, Elizabeth Redcay and Eric Courchesne

Secondly, here is Dr Courchesne speaking before the ASF's 2010 meeting:
“We discovered that autistic infants and toddlers displayed a pronounced abnormality of language activation and cortical development.” “At each age studied from infancy to young childhood, most autistic subjects had greater activation on the incorrect side, namely, the right temporal cortex, compared to the left side and this incorrect activation pattern did not change or “normalize” even by 3 or 4 years of age. The abnormal pattern was strong in a substantial percentage of autistic infants and toddlers....".

Head size, simplex/multiplex autism, and IQ

A number of reports have been published indicating that head size in the first year of life has been found to be several percent greater in autistics, such as this latest one.
Media report linked here.

This is important and substantially competent research by this team. But there's some room for improvement.

Like most researchers they presume that autism is a "disorder" for which they are trying to find out what has "gone wrong". And so even when something good is found it has to be labelled as "overgrowth" of the brain.

And yet a remarkably similar progression of "overgrowth" has been found to be characteristic of high-IQ children (Gale, O'Callaghan, Bredow, Martyn 2010). And the 1982 published 1993 antiinnatia theory explained that whatever in low doses causes high IQ in higher doses causes autism.

You can see a very good presentation to the IACC about this "overgrowth", by the founder of this line of evidence, Eric Courchesne (between minutes 130-162 plus discussion to 174). He explains that it involves substantially higher numbers of neurons particularly in prefrontal. And that it is characteristic of "simplex" autism rather than of "multiplex" autism (defined as autism cases occurring more than once in a family). And furthermore, the advance over normal growth comes to level off after the first couple of years and indeed may be followed by a decline to below normal.

I've not looked into this research in great depth, but would provisionally suggest the following as reasonable explanations of these findings.

Multiplex autism could usually be due to a family sharing an indoor mercury vapor pollution problem from (mainly maternal) dental amalgams. EC notes that multiplex is much more common, which is in line with my conclusion that dental mercury- induced autism now accounts for 80% or more of all autism.

So it would be the simplex autism that is more a matter of too many antiinnatia genes, that is genes for high IQ, and which would also be genes favouring large heads and more neurons, which is in line with these observations. (The larger heads and more neurons would not cause the higher IQ but would be correlated with it.)

The subsequent decline would be readily explained by two processes. Firstly anyone with even fleeting acquaintance with autism can see that it soon leads to the autistic experiencing a substantially understimulating environment compared to neurotypicals. Such an understimulating environment can reasonably be expected to lead to the brain waning away somewhat. And meanwhile, for those autism cases caused by mercury, the mercury would continue to accumulate due to the defective detoxification, and the brain would consequently lose neurons due to that neurotoxin accumulation.

Adult-onset autism

In the update review of the antiinnatia autism theory I explain why dental amalgam mercury vapor causes autism in infants but not in adults. Namely because beyond a certain age the behaviours associated with the innatons become established as learnt habits, skills, knowledge, as the elaborated connectivity of the brain, and as crystallised intelligence in place of fluid intelligence.

There might be thought to be a valid objection to this in terms of three case reports of autism beginning at later ages, 11, 14, and 31 years.

It is notable that all three cases resulted from herpes encephalitis. In respect of the 31-year-old, he was in a coma for more than 14 days, and from having been a university lecturer was permanently transformed to having a mental age of less than 2 years. We can thus see that it was only in the context of an extreme level of brain damage that the characteristics of autism resulted. So we can interpret his case as being one where even the established habits, skills and so on were wiped out, by a condition substantially more severe than (for instance) mercury poisoning. (Not that mercury poisoning should be considered non-serious itself.)

Autism did not begin only in the 1930s

In two weeks time there will be published a new book:
The Age of Autism: Mercury, Medicine, and a Man-made Epidemic
By Dan Olmsted, Mark Blaxill

I've not read it (yet) but I can anticipate that I would agree with the authors that autism has increased and involved mercury but would be unpersuaded by their reckoning that vaccinations have been a major factor in the increase.

Here I shall just comment on this first sentence of a preview extract circulated by Safeminds:
"We believe that autism was newly discovered in the 1930s for the simple reason that it was new."
But this first sentence can be shown to be mistaken. Dr Down of Down Syndrome fame had already in the 19th century given good descriptions of both infantile autism and regressive autism.
Reference: Down, J.L. Mental Affections of Childhood and Youth, 1887 originally, re-issued as Classics in Developmental Medicine, No. 5, 1990 Mac Keith Press, London

It was discussed in the 2006 Awares conference. This is a link to the paper by Darrold Treffert.

I also suggest that the concept of the "holy fool" such as portrayed in the play and opera Boris Godunov corresponds with mild autism: a person who speaks the truth that others cannot. And bear in mind that the rare autistic individuals would be likely to have had a hard time surviving in earlier ages; even mildly autisticky people might have had a hard time sufficient to prevent them producing children.

Furthermore, there is the important consideration that a rare condition such as autism pre-1970 would only come to the attention of a person once there is a sufficient level of cosmopolitanism, due to urbanisation and transport, enabling one person to "survey" a sufficiently large sample of people. Down made his observation in 19th century England after the railways were established in the world's first modernised country. Thereafter, the US and Austria started to catch up industrially and so autism came to attention there too. That's not to say that the personal links to mercury stated in the AoA book cannot be also part of the story.

Vitamin D, sunshine, skin color, migration, and autism

Others have pointed to evidence suggesting that autism is associated with migration to less sunny countries, especially by darker-skinned people. For instance according to Autism, ethnicity and maternal immigration. Keen DV, Reid FD, Arnone D.:
"Maternal immigration is associated with substantial increased risk of autism-spectrum disorders with differential risk according to different region of birth and possibly ethnicity."
There's the notable "epidemic" among Somalian immigrants. This could conceivably have something to do with (I guess) higher selenium levels in Somalian diet (due to lower rainfall, different geology, or the high fish content of the local diet). People moving to a diet of less selenium than they had evolved for could be more vulnerable to the antiinnatia effect of mercury, in absence of less antidoting selenium.

But while I don't rule out that possibility, a more likely process appears to involve vitamin D deficiency. Indeed a theory of causation by vitamin D deficiency has already been published in Medical Hypotheses (before that journal's editorial decisions recently became subject to non-editorial interventions by the owner, the publisher of supposedly scientific journals Elsevier).

This vitamin D deficiency thesis would tie in solidly with the antiinnatia theory, for the reason explained by the popular health writer Joseph Mercola:
"In fact, there’s compelling evidence that vitamin D is in fact KEY for proper gene expression." "Each cell in your body has its own ‘DNA library’ that contains information needed to deal with virtually every kind of stimulus it may encounter, and the master key to enter this library is activated vitamin D."
In other words, Mercola is here stating that vitamin D deficiency is something of an antiinnatia factor. So of course it would tend to cause autism (at critical developmental periods). Indeed, I indicated in my 1993 paper (and indeed all versions from 1982 onwards) that deficiencies of nutrients could obviously be antiinnatia factors.

Questions now arise as to how much contribution to antiinnatia (autism etc) is made by vitamin D deficiency, and whether the relationship with mercury, antiinnatia genes, etc is additive or synergistic.

Of course using vitamin D deficiency to raise one's infant's IQ would have certain major downsides, probably even worse than dosing them with extra mercury vapor; so farbeit from me to recommend it as a sort of "smart" drug.

P.S.: Vitamin D deficiency as an antiinnatia factor could also be the basis of recent observations of seasonal variation of autism incidence (by birthdate).

One might next predict that if D deficiency is indeed an antiinnatia factor, then IQ should also be season-of-birth dependent. It looks like that may not be the case. But then quite possible reasons for that could be that in the populations studied there was not much sunlight exposure anyway, and schoolchildren were routinely supplied with cod liver oil and orange juice (as in the uk for many years) such as would prevent substantial deficiency.

Anecdotes or Cures?

Some of the information that organisations such as the UK "National Autistic Society", "Research Autism", and Cambridge's "Autism Research Centre" aren't capable of mentioning is at this link and this one.
This is a link to a video of Polly Tommey explaining about the autism catastrophe (I don't share her ideas re mmr).

Fetal testosterone, “extreme-male-brain” and “developmental instability”

A conception of autism as extreme-male-brain (EMB) has attracted much publicity, with many persons being led to assume that it is the only remotely meritable contemporary understanding of autism.

There is indeed good reason to believe that fetal testosterone (FT) affects the development of the fetus, such that the brain remains thereafter more “male”; and this then manifests in more tendency towards “systematising” (as involved in science or engineering) and less towards empathising or emotional sensibility.

A questionable extension from this is the notion that FT increases some “autistic traits” and that in extreme those traits amount to autism, with autism being understood as being effectively identical to EMB. This “autism = EMB” thesis depends on overlooking the substantial evidence which does not fit with it. Autism (pre-increase) had strong associations with high social class and high parental IQ [1], and it is far from clear why extreme-male-brain would have. Likewise unclear is how it could credibly account for the symmetry data or the physical stigmata [1], or why it would involve such un-male but classic autism characteristics as shyness, hand-flapping/posturing, echolalia of whole sentences, lack of dizziness after spinning, intense resistance to change, toe-walking, etc [1].

EMB also struggles to explain the famous increase of autism, invoking at best a notion of a hypothesised increase of assortative mating of geeks. It is difficult to see any credible calculation of how assortative mating could have so rapidly increased autism tenfold within 20 years. And it would suggest that the autism increase in Silicon Valley would be conspicuously far greater still, whereas in practice the increase seems much the same everywhere. And EMB also fails to account for the stark change of ratio of age of onset.

On the one hand those numerous facts clash with the autism-as-EMB and geek assortative mating conceptions, while on the other hand there is the fully satisfactory alternative explanation presented here and in the 1993 paper.

These considerations show that EMB has inadequate merit as a candidate for being the central theoretical concept of autism. It can however be seen to be a part of the story, as I will now explain.

Innate programming has a more substantial role in the behaviour of female mammals than of males, for pregnancy management and nurturing (for instance empathy, theory of mind, communication). So the biologically optimum level of antiinnatia is lower for females. (In addition they would tend to have stronger genetic endowments of these predispositions, more resistant to antiinnatia factors.) Meanwhile, “systematising” is what brains do as a matter of default in the absence of specific pre-programmed reactions being evoked. Consequently the relatively “blank slate” mind of high antiinnatia tends to look like “male brain” in some respects. Or they even tend to actually be the same thing. Males would have a higher optimal level of antiinnatia, and so the characteristically male hormone testosterone would advantageously tend to raise antiinnatia somewhat (or in other words FT would tend to be something of an antiinnatia factor).

If FT were the sole or principal non-environmental antiinnatia factor, then women would be concentrated at the low end of the scales of IQ, health, body symmetry and beauty, while men would be concentrated at the high end of those scales. But that rather obviously is not the case, and that tells us that FT cannot be more than a relatively minor antiinnatia factor.

See also the explanation in the sex differences section of my original paper.

The associated concept of “developmental instability” presumes that organisms have “correct”, “intended”, “normal” courses and outcomes of development from which “maldevelopment” deviates. But they do not. There is no blueprint in DNA; rather, development just happens blindly and aimlessly in interaction with varying environment, just as natural selection does. And already in the 1993 paper I had indicated two respects in which even erraticness (‘instability’) of phenotypic outcome can be biologically advantageous (i.e. ‘intended’).

The correct concept—the oxygen to this phlogiston of developmental science—is antiinnatia, now more evidentially-suppported than ever.