One thousand and ten unselected London state schoolgirls were screened by questionnaire to identify an 'at risk' cohort displaying abnormal eating attitudes and two control cohorts, one with probable general psychiatric morbidity, one without. Members of all cohorts were assessed at interview for the presence of eating disorder and for putative risk factors implicated in the development of anorexia nervosa. A prevalence rate of 0.99% was detected for clinical eating disorder and 1.78% for the partial syndrome of eating disorder. Factors specifically associated with abnormal eating attitudes were identified, in particular, current or past overweight, history of amenorrhoea and perceived stress in school and social life. Some commonly accepted risk factors for eating disorders were discovered to be associations with general psychiatric morbidity. These were perceived parental pressure to eat more, taking exercise to lose weight, perceived stress at home and reporting a family history of anxiety or depression. Other well reported putative risk factors for eating disorder, including social class, birth order, age at menarche, obsessional personality and weight related career choice were not associated specifically with abnormal eating attitudes in schoolgirls. These findings represent cross-sectional data at entry into a prospective epidemiological study.