OBJECTIVE To evaluate the accuracy and interobserver variability of magnetic resonance cholangiography (MRC) and endoscopic retrograde cholangiography (ERC) in the diagnosis of primary sclerosing cholangitis (PSC). METHODS MRC at 1.5 T with thin- and thick-slice breath-hold technique was performed in 66 adult patients (median age, 44 y; 26 women) with an appropriate spectrum of hepatobiliary diseases. Maximum intensity projection images were reconstructed from the thin slices. ERC was performed within 48 hours of the MR examination. The reference standard of PSC diagnosis was based on a combination of clinical features and cholestatic biochemical profile with typical ERC and/or MRC abnormalities and supported by liver histology findings. Two independent reviewers who were unaware of final diagnoses analyzed all images retrospectively. RESULTS PSC was diagnosed in 39 (59%) of 66 patients. MRC provided comparable and poorer depiction than ERC of extrahepatic and intrahepatic ducts, respectively. However, the diagnostic accuracy of ERC and MRC were comparable. In the MRC detection of PSC, the average sensitivity of 2 independent readers was 80%, the specificity was 87%, and the accuracy was 83%. The corresponding values for ERC were a sensitivity of 89%, a specificity of 80%, and an accuracy of 85%. Interobserver agreement for the diagnosis of PSC was good (kappa = .61) for MRC and excellent (kappa = .81) for ERC. CONCLUSIONS PSC can be diagnosed with high accuracy and good interobserver agreement. MRC and ERC performed equally well in the diagnosis of PSC when used blinded to clinical information.