When using quartz OSL to date loess samples from the Chinese Loess Plateau, it has been reported that the agreement between OSL ages and the independent ages is limited to the samples younger than ∼70 ka with a corresponding De of ∼230 Gy, and a sample with an expected age of 780 ka was dated to 107 ka, corresponding to 403 Gy. The growth curves of these samples do not saturate at doses of 700 Gy, and a linear growth part was observed for doses higher than 200 Gy. However, the maximum measured age of ∼100 ka imply that the De determined using this linear part of a growth curve could be problematic, or that the quartz OSL signal is not as stable as previously thought and has a barrier age of ∼100 ka. In the current study, we examine the reasons for the age underestimation. We examined the shape of growth curves, anomalous fading, thermal stability, etc. The results show that, for the loess samples examined, quartz OSL does not fade anomalously, and the barrier age of ∼100 ka is due to the fact that the OSL signals are less thermally stable, the lifetime of 0.311 Ma at 20°C obtained is much smaller than those for quartz samples from other regions such as Australia (∼100 Ma).