Little is known about the etiology of bulimia nervosa and the attitudes associated with it. We have undertaken a study of selected (45 pairs) and unselected (106 pairs) female twins to elucidate the broad causes of individual differences in these behaviours and attitudes. The selected sample was chosen on the basis of at least one of the twin pair having a lifetime incidence of bulimia nervosa. Biometrical model fitting, which corrected for the biased twin correlations of the ascertained group, was used to investigate the genetic and environmental risk factors contributing to the development of bulimia nervosa. The best-fitting model showed that individual variation was best explained by additive genetic influences (62%) and nonshared environmental influences (38%). The proportion of genetic variance affecting individual variation in the ascertained group and the random group was not found to be significantly different. In summary, it is suggested that it may not be necessary to supplement a randomly selected sample with an ascertained sample when investigating the liability to a low-prevalence psychiatric disorder if a continuous measure of that disorder is available.