OBJECTIVE The aim of this study was to evaluate the interobserver agreement of cirrhosis symptomatology. METHODS Eight patients with cirrhosis (without ascites) and 23 patients with chronic liver disease without cirrhosis were included. Four observers (2 juniors and 2 seniors, with 1 specialist and 1 general radiologist in each group) independently evaluated video recordings. Agreement of 18 variables was calculated by using kappa index (k), intraclass correlation coefficient (r) and agreement proportion. Observer bias and poorly trained observers were also investigated. RESULTS Liver: surface: k = 0.68, homogeneity: k = 0.25, span: r = 0.87, caudate lobe: r = 0.4 to 0.57 according to 3 dimensions, right lobe length: k = 0.33. Veins: portal diameter: r = 0.73, splenic diameter: r = 0.75; umbilical vein: presence: k = 0.19, diameter: r = 0.5; left gastric vein: presence: k = 0, diameter: r = 0.18; spleno-renal shunt: k = 0.41. Spleen length: r = 0.93. Presence of cirrhosis: k = 0.53. The mean agreement for the 18 variables was not significantly different between juniors (0.53 +/- 0.32) and seniors (0.40 +/- 0.38) but was significantly better in specialists (0.61 +/- 0.22) than in general radiologists (0.43 +/- 0.40; P < 0.05). CONCLUSIONS Interobserver agreement of ultrasonography is good to excellent for the main characteristics of the liver, the main splanchnic veins and spleen but poor to fair for collateral vessels and the caudate lobe. Agreement for diagnosis of cirrhosis is good but should be improved by only selecting reliable signs or experienced echographists.