Personality scales based on the five-factor model, especially the Big-Five Scale of personality trait adjectives (Wada, 1996), are commonly used in Japan. In this study a short form of the Big-Five Scale was constructed. To avoid changes in the capacity dimension caused by the decrease in the number of items, item selection was conducted after Item Response Theory (IRT) scales were constructed for all the items. In Study 1 data was collected from 2 099 participants. A Generalized Partial Credit Model was applied to the IRT model, and items were selected using the slope and location parameters for each item. Cronbach's alpha showed that the short form, as well as the five sub-scales, had sufficient reliability as a personality test. In Study 2, we determined correlations with the NEO-FFI and tested the concurrent validity of the short form. The results indicate that the short form of Big-Five Scale demonstrates sufficient reliability and validity despite the reduced number of items.