A prospective study was undertaken to critically evaluate the endocervical curettage. The study group consisted of 261 patients with colposcopically identifiable cervical lesions that had not been previously biopsied and were not macroscopically suspicious for invasive cancer. In 210 patients, the entire lesion, transformation zone, and squamocolumnar junction were visible (satisfactory colposcopy), and in 51 patients, the full extent of the lesion was not visualized colposcopically (unsatisfactory colposcopy). We used a method of examination not previously described. Colposcopy was performed before and after the endocervical curettage, but before the ectocervical biopsy. This method of examination allowed us to identify potentially contaminated endocervical curettages and to test the null hypothesis that a positive endocervical curettage in patients with satisfactory colposcopy is not the result of contamination. Twenty of 210 endocervical curettages in patients with satisfactory colposcopy were potentially contaminated, as suggested by an ectocervical lesion disrupted by the endocervical curettage. No patient with a positive endocervical curettage had an intact ectocervical lesion. Therefore, the null hypothesis was rejected (P less than .005). Of the 51 patients with unsatisfactory colposcopy, 39 (76%) underwent cone biopsy. Invasive cancer was diagnosed by cone biopsy in two patients when both ectocervical biopsy and endocervical curettage demonstrated only dysplasia. We also confirmed that patients with unsatisfactory colposcopy occasionally benefit from having an endocervical curettage. However, basing the treatment plan for any lesion short of invasive cancer solely on the results of colposcopy, ectocervical biopsy, and/or endocervical curettage could lead to significant undertreatment of some patients.