Xu and Tenenbaum (2007a, 2007b) applied the Bayesian model to explain the impact of differences in exemplification onwords learning, and they achieved milestones. It remains unexplored if there are differences when native language andculture are changed. Taking the same method as the original research, we added test after a long time interval, and usebetween-subject design to eliminate the practice effect. The results of Chinese adults and children show that: (1) TheBayesian model has stability over time and culture. (2) When the objects in the same category differ greatly from eachother, the Bayesian model’s predictive power on children’s results is significantly reduced. (3) Since the low-level wordsin Chinese vocabulary are often composed of high-level words and adjectives, Chinese easier to generalize. (4) Results ofChinese subjects reflect more instinct rather than logical reasoning stylewhich is differ from westerners.