The high concordance for autism symptoms in monozygotic twin-pairs compared to di-zygotic twins and/or non-twin sib-ships suggests a high genetic determinism in autism. Those results have hypothesized multi-factorial determinism in accordance with family studies and mathematical models. However, linkage and association or candidate gene strategies have failed to-date to identify clearly involved mechanisms. Mental retardation (MR) is known as frequently associated to autism. Multiplex XLMR pedigrees have been reported with only one mutated patient having autism and MR: different X-located MR genes have been shown to be involved (NLGN4, MECP2, OPHN1, ZNF674 and FRAXA) which does not suggest that they could be "autism genes". Tuberous sclerosis studies and report of numerous autosomal domains shown deleted in MR-autistic subjects suggest that several autosomal dominant (AD) genes could be also involved in MR with autism. Whereas multiplex AD-MR families are rare, AD de novo mutations could explain numerous sporadic situations of non-specific MR and of autism with MR, in accordance with twin studies. Finally, we hypothesize that in those autistic subjects with mendelian MR, the XL-MR or AD-MR gene (G1) would pave the way for a second Mendelian factor (G2) responsible for autism symptoms.