Seventy-two asymptomatic, postmenopausal women treated with tamoxifen for breast cancer were studied prospectively with vaginal ultrasonography, followed by endometrial sampling. Seventy-one patients demonstrated an endometrial echo of more than 5 mm, and one displayed an endometrial echo of less than 5 mm. No patient with an endometrial echo of less than 5 mm displayed any endometrial pathology. Different endometrial pathologic conditions were found when the endometrial echo was over 5 mm. When classifying those patients whose ultrasonic endometrial widths were wider than 5 mm, on the basis of different endometrial histologic findings, no obvious correlation was found between the various pathologic endometrial findings and endometrial thickness. Thus, a "thicker" endometrial sonographic image did not necessarily correlate with pathologic endometrial findings. When ultrasonic endometrial thickness of 5 mm was considered the upper limit of normal, the sensitivity of ultrasonography in correlating to positive histologic findings was 91% and the specificity was 96%. These findings suggest that there is good correlation between endometrial width measured by ultrasonographic assessment and histologic findings.