BACKGROUND/AIMS: Capsule endoscopy (CE) is a new method enabling noninvasive diagnosis of small bowel diseases. There have been few studies examining the possibility of interobserver variation according to proficiency. We evaluated the interobserver variability between expert and novice for reviewing CE images. METHODS: Among patients who were taken CE from June 2003 to July 2003, twenty patients were randomly selected. Captured images were assessed by an expert and a novice separately. The expert has experience of more than 150 CE interpretation and the novice only had experience in performing EGD and colonoscopy. The novice had trained on interpretation of the CE with 5 cases before this study. Interobserver agreement was evaluated using kappa coefficient. RESULT: CE findings were divided into normal/ abnormal groups. Abnormal groups were classified into small focal, large focal, multiple diffuse, structural deformity groups. Compared with the expert, the novice missed 2 cases (Meckel's diverticulum and angiodysplasia). There was no disagreement in other cases. CONCLUSION: There is 90% (18/20) interobserver agreement between the expert and the novice for the interpetation of findings CE (k=0.737). Normal findings and diffuse large lesions tended to have higher concordance, whereas small focal lesion and structural deformities were more likely to be a source of disagreement.