Roundabouts have been used successfully in cities throughout the world and are wildly used in China. The critical gap is a required parameter for calculating the capacity of roundabouts and deciding the needs for stop signs. While many studies on gap acceptance characteristics at roundabouts have been car- ried out in America and Europe, few have been done in China. This study was conducted to calibrate the critical gap in a roundabout in China. The differences between critical gaps on different days (workdays and weekends) and during different times (rush hour and non rush hour) were checked. The study was per- formed at a three-lane roundabout formed by a two-way and four-lane local road and a six-lane divided ar- tery. The data was collected at this site by a video camera. The evaluations included the cumulative prob- ability distributions of the acceptable and rejected gaps which resulted in critical gaps for all cases. The re- sults show that the critical gap during the rush hour is shorter than after the rush hour, but the difference is not large and the cumulative probability distributions can be fit by the logit functional form with parameters estimated using linear regression. Roundabouts have been used successfully in Cities throughout the world and are wildly used in China. The critical gap is a required parameter for calculating the capacity of roundabouts and deciding the needs for stop signs. While many acceptance on gap acceptance characteristics at roundabouts have been car This study was conducted to calibrate the critical gap in a roundabout in China. The differences between critical gaps on different days (workdays and weekends) and during different times (rush hour and non rush hour) were checked. The study was per- formed at a three-lane roundabout formed by a two-way and four-lane local road and a six-lane divided ar- tery. The data was collected at this site by a video camera. The evaluations included the cumulative prob- ability distributions of the acceptable and rejected gaps which resulted in critical gaps for all cases. The re- sults show that the critical gap during the rush hour is shorter than after the rush hour, but the difference is not large and the cumulative probability distributions can be fit by the logit functional form with parameters estimated using linear regression.