The molecular pathogenesis of ASD (autism spectrum disorder), one of the heritable neurodevelopmental disorders, is not well understood, although over 15 autistic-susceptible gene loci have been extensively studied. A major issue is whether the proteins that these candidate genes encode are involved in general function and signal transduction. Several mutations in genes encoding synaptic adhesion molecules such as neuroligin, neurexin, CNTNAP (contactin-associated protein) and CADM1 (cell-adhesion molecule 1) found in ASD suggest that impaired synaptic function is the underlying pathogenesis. However, knockout mouse models of these mutations do not show all of the autism-related symptoms, suggesting that gain-of-function in addition to loss-of-function arising from these mutations may be associated with ASD pathogenesis. Another finding is that family members with a given mutation frequently do not manifest autistic symptoms, which possibly may be because of gender effects, dominance theory and environmental factors, including hormones and stress. Thus epigenetic factors complicate our understanding of the relationship between these mutated genes and ASD pathogenesis. We focus in the present review on findings that ER (endoplasmic reticulum) stress arising from these mutations causes a trafficking disorder of synaptic receptors, such as GABA (gamma-aminobutyric acid) B-receptors, and leads to their impaired synaptic function and signal transduction. In the present review we propose a hypothesis that ASD pathogenesis is linked not only to loss-of-function but also to gain-of-function, with an ER stress response to unfolded proteins under the influence of epigenetic factors.