From October 1990 to January 1992, 36 patients were operated in order to perform an hepatectomy for hepatic metastases (n = 33) or primary malignant tumors (= 3). The authors evaluated the reliability of preoperative ultrasonography (US) by comparing US data to intraoperative findings without consideration of others preoperative examinations. Preoperative US detected 65 out the 78 nodules founded intra-operatively; 13 nodules (4 to 30 mm) and a micronodular dissemination were not detected (sensitivity 82% for the whole lesions). Probable causes of false negative cases (size, location and echogenicity of the nodules) are discussed. Sixty three out of the 65 nodules detected were correctly located in the hepatic segmentation by preoperative US. Hepatectomy was performed in 30 of the 36 patients (84%). The resection previously planned by preoperative US was realized in 24 patients, 5 hepatic resections (14%) had to be enlarged and one completed by a lymphadenectomy; in 6 cases (16.5%) the hepatectomy was not possible because of intraoperative findings of lesions previously not detected by preoperative examinations including US (3 with not removable hepatic nodules, 3 with lymph nodes metastases). This study confirms the reliability of US for anatomical location of hepatic lesions. US sensitivity for the detection is satisfactory. However the number of nodules was underestimated in one third of the patients. The hepatic resection previously planned was effectively realized in only 67% of the cases, in one case out of 6, the resection had to be enlarged and in one case out of 6, the resection was not possible.