The purpose of this work was to study intraobserver and interobserver variation in the interpretation of colposcopic images of cervical intraepithelial neoplasia (CIN). Twenty-three experienced colposcopists were asked to assess colposcopic images presented on slides and to select the biopsy site. Eleven cases were independently interpreted twice with an interval of 2-3 months by all observers. No information about the cytological classification was available. In each case the "majority assessment" was considered as the standard, being "no CIN" in 2 cases, CIN I in 4 cases, CIN II in 3 cases, and CIN III in 2 cases. Intraobserver concordance was 66.7%, the kappa value was 0.54. Interobserver agreement was found to be 52.4 and 51.0% in the first and second sessions, respectively, while the mean kappa values were 0.41 and 0.33, respectively. In selecting the site for biopsy, 77.4% of all observers agreed while the same site was selected in 85.3% of cases by the individual colposcopist in the two sessions. Overall, CIN I and II interpretations revealed lower levels of agreement than no CIN or CIN III interpretations. It is concluded that observer variability in interpreting colposcopic images and selecting the site for biopsy is in the same range as observer variation in other subjective diagnostic tests such as cytology and histopathology. This variation should be taken into account in the colposcopical management of patients with abnormal cytology.