Attenuation imaging is a novel, ultrasound-based technique to objectively detect and quantify liver steatosis. In this study, we evaluated the performance and inter-observer variability of attenuation imaging and compared it to a known quantification method of liver fat, the hepatorenal index (HRI). Two observers measured attenuation coefficients (AC) in an attenuation phantom, 20 healthy volunteers and 27 patients scheduled for biopsy for suspected diffuse liver disease. Results were compared with the HRI and histological findings. Both observers were blinded to the results of the biopsy and the measurements of the other observer. Our results showed that patients with moderate (S2, 33-66%) and severe fatty infiltration of the liver (S3, >66%) showed significantly higher ACs in comparison to patients with a liver fat fraction of less than 33% (S0/1). There was no significant difference in AC-values of patients with fatty infiltration of less than 5% (S0) and 5-32% (S1). In the Receiver Operating Characteristic (ROC)-analysis, the area under the curve (AUC)-values for the detection of moderate and severe steatosis were excellent at 0.98. Cut-off values were 0.64 dB/cm/MHz for the detection of S2- and 0.68 dB/cm/MHz for the detection of S3-steatosis. The inter-observer agreement of attenuation imaging was very good with an intraclass correlation coefficient (ICC) of 0.92 in patient and 0.96 in phantom measurements. The ICC decreased with depth in the phantom measurements. In summary, attenuation imaging showed very good inter-observer agreement and is a promising tool for the detection and quantification of moderate and severe hepatic steatosis.