Purpose: Many diagnostic modalities for neonatal cholestasis have shown features that are helpful, however until recently none of them are not pathognomonic of biliary atresia (BA). We carried out a prospective study of infants with cholestatic jaundice with the aim of establishing an efficient method of diagnosing BA. Methods: Twenty-seven consecutive infants with cholestatic jaundice were enrolled in this study and ranged from 7 to 152 days in age (mean; $51.2{pm}34.2$ days). Gastroduodenoscopy was carried out using a fiberscope (Olympus N30). All the babies were fasted for at least 4 hours before the procedure and 20 ml of 10% dextrose solution was given at the time of endoscopy. The endoscopic examination focused on the 5 minutes observation of the evidence of biliary secretion. If there was lack of the evidence of the biliary secretion, endoscopy was removed and repeated the examination with some pause. Results: There are lack of the evidence of biliary secretion in all infants with BA. In non-BA group, 8 out of the 10 infants showed biliary secretion on the first trial, however one (Alagille syndrome) of the two infants without evidence of biliary secretion, finally exhibited biliary secretion on the second trial. The above observations resulted in the diagnostic accuracy of 96.3% with 100.0% sensitivity and 90.0% specificity. Conclusion: In light of the results from our relatively small study, endoscopy is a convenient, and relative inexpensive procedure. we strongly support the use of endoscopy for the diagnosis of BA in the screening and evaluation of infantile cholestasis.