A systematic, prospective observer-rated study was carried out to determine the prevalence of late luteal phase dysphoric disorder (premenstrual syndrome) in women with autism. A group of women with autism and learning disability (n = 26) was compared with a group of women with a non-autism learning disability (n = 36) matched for age, in-patient status, intelligence, marital status, parity, behavioural problems and ethnicity. Observers rated DSM-IV symptoms of late luteal phase dysphoric disorder every day from each subject over three consecutive menstrual cycles. Using a premenstrual increase in DSM-IV symptoms of >or= 30% as evidence of fulfilment of diagnostic criteria, the prevalence of late luteal phase dysphoric disorder was 92% in the autism group compared with 11% in the control group. This difference was highly statistically significant. The principal conclusion from this study is that there is a marked increase in premenstrual syndrome in women with autism compared with matched controls.