Computer-based testing (CBT) has been used in Japan since 2002 to assess medical students' basic and clinical medical knowledge, based on the model core-curriculum, before they start clinical clerkships. For effective CBT, multiplechoice questions must accurately assess the knowledge of students. Questions for CBT are submitted by all medical schools in Japan. However, only 40% of questions are chosen for CBT and used at random; the other 60% of questions are rejected because of poor quality. Toimprove the ability of medical staff to devise questions, workshops were held at 30 medical schools. The acceptance rate of questions from schools where workshops were held was significantly increased. The workshops were extremely effbctive for improving the quality of questions.