Our objectives were to determine the reproducibility, or interobserver agreement, of transvaginal sonographic imaging of the uterus in patients with abnormal uterine bleeding and to identify the effect of observer experience. Transvaginal ultrasound findings of 235 patients with abnormal uterine bleeding were recorded systematically on videotape. Recordings were reviewed by three observers who had different levels of experience, and who were asked to judge the endometrium/uterine cavity and myometrium separately as being normal, abnormal or inconclusive, according to predefined criteria. Reproducibility was expressed by the observed rates of interobserver agreement and by kappa statistics. The differences in agreement between observer pairs were analyzed by means of McNemar's chi 2 test. The observed rates of agreement for the judgement of the endometrium/uterine cavity varied from 0.85 to 0.89, with a kappa value ranging from 0.70 to 0.78 between observers. The judgment of the myometrium resulted in agreement rates of 0.86-0.91 and kappa values of 0.67-0.80. Although the effect of experience was evident, the differences in agreement between observers were not significant (p > 0.01). The reproducibility of the results of transvaginal sonography of the uterus in patients with abnormal uterine bleeding was good. Observations of the endometrium/ uterine cavity with a normal appearance were the most highly reproducible, with the smallest effect of observer experience. This may reduce the need for invasive diagnostic procedures in patients with abnormal uterine bleeding.