To evaluate the accuracy of MRCP in diagnosing choledocholithiasis considering Endoscopic Retrograde Cholangiopancreatography (ERCP) as the gold standard. To compare the results achieved during the first two years of use (1999-2000) of Magnetic Resonance Cholangiopancreatography (MRCP) in patients with suspected choledocholithiasis with those achieved during the following two years (2001-2002) in order to establish the repeatability and objectivity of MRCP results.One hundred and seventy consecutive patients underwent MRCP followed by ERCP within 72 h. In 22/170 (13%) patients ERCP was unsuccessful for different reasons. MRCP was performed using a 1.5 T magnet with both multi-slice HASTE sequences and thick-slice projection technique. Choledocholithiasis was diagnosed in the presence of signal void images in the dependent portion of the duct surrounded by hyperintense bile and detected at least in two projections. The MRCP results, read independently from the ERCP results, were compared in two different and subsequent periods.ERCP confirmed choledocholithiasis in 87 patients. In these cases the results of MRCP were the following: 78 true positives, 53 true negatives, 7 false positives, and 9 false negatives. The sensitivity, specificity and accuracy were 90%, 88% and 89% respectively. After the exclusion of stones with diameters smaller than 6 mm, the sensitivity, specificity and accuracy were 100%, 99% and 99%, respectively. MRCP accuracy was related to the size of the stones. There was no significant statistical difference between the results obtained in the first two-year period and those obtained in the second period.MRCP is sufficiently accurate to replace ERCP in patients with suspected choledocholithiasis. The results are related to the size of stones. The use of well-defined radiological signs allows good diagnostic accuracy independent of the learning curve.