This study examined the differential contribution of pre- and perinatal risks in narrowly versus broadly defined autism spectrum disorder (ASD) and across core symptom domains, IQ and co-morbid problems. Children with a DSM-IV diagnosis of autistic disorder (AD) (n = 121) or pervasive developmental disorder not otherwise specified (PDD-NOS) (n = 75) were compared to a typical control sample (n = 311). Diagnoses were based on extensive assessments between 12 and 49 months of age (M = 33.3, SD = 6.4) and re-evaluated at 43-98 months (M = 68.1, SD = 10.7) in 70% of the cases. Compared with controls, cases with ASD were more likely to be firstborn and show a suboptimal condition after birth. Case mothers reported more infections and more stress during pregnancy. Although the ASD subgroups showed mostly overlapping risks, cases with PDD-NOS differed from those with AD by higher exposure to smoking during pregnancy (SDP) and by a negative association of smoking with IQ, regardless of confounders. SDP appears to contribute more to broadly defined (PDD-NOS) than to narrowly defined ASD (AD). Findings suggest differences in etiological contributors between ASD phenotypes.