The degree of steatosis in liver biopsies is usually assessed by a morphological semiquantitative approach in which the histopathologist uses a four-graded scale: 0-3 or none, slight, moderate and severe. Scores 1-3 are considered to correspond to fat deposition in <33, 33-66 and >66% of the hepatocytes. There is a considerable inter- and intra-individual variation in such scoring methods and a more standardized and quantitative approach is preferable. In the present study, we compare the semiquantitative technique with the stereological point counting method in the assessment of hepatic steatosis. A total of 75 archived liver needle biopsies were used. They were selected according to the original routine diagnosis of slight, moderate or severe steatosis. In all, 10 randomly selected images from each biopsy were digitized into a computer, a point grid lattice was superimposed and the number of hits on fat globules was counted. A pathologist scored the specimens in a four-graded scale as described above. The mean liver biopsy area (volume) with fat in hepatocytes was 2.2% for grade 1, 9.2% for grade 2 and 23.1% for grade 3. The kappa value for the semiquantitative estimates was 0.71 for the unweigthed kappa and 0.87 for weighted kappa. The intraclass correlation coefficient (ICC) was 0.99 for images counted twice and 0.95 when two sets of images were captured from the same biopsy. These ICCs indicate excellent agreement and above that of the semiquantitative estimates. In conclusion, the area/volume of fat content of the hepatocytes is greatly overemphasized in semiquantitative estimation. Furthermore, the point counting technique has a better reproducibility than visual evaluation and should be preferred in estimates of liver steatosis in scientific studies and in clinical contexts when the amount of steatosis is important for treatment and prognosis, such as liver transplantation.