Background. We sought to determine whether there was empirical support for the diagnostic thresholds of DSM-IV bulimia nervosa (BN) and whether an empirically derived typology resembled the diagnostic categories of DSM-IV.

Methods. Detailed information about bulimic behaviours were assessed via personal interview in a population-based sample of 1897 Caucasian female twins. We assessed the lifetime prevalence of the component bulimic behaviours and DSM-IV and DSM-III-R BN. Latent class analysis of nine separate bulimic symptoms was used to develop an empirical typology of bulimic behaviour.

Results. Although the lifetime prevalences of bingeing (23·6%) and vomiting (4·8%) were relatively common, DSM-IV BN was distinctly uncommon (0·5%). The criterion that specified the frequency and duration of bingeing and vomiting was an important limiting condition. Analysis of alternative thresholds found little support for the DSM-IV thresholds requiring an average of twice per week for 3 months. Latent class analysis yielded an interpretable four class solution that had little overlap with the DSM-IV typology.

Conclusions. As in other studies of unselected samples of women, the lifetime presence of bulimic behaviours are relatively high. Our results suggest that the DSM-IV approach to categorizing bulimic behaviour inadequately captures the spectrum of lifetime bulimic behaviours in the general population.