Species abundance pattern plays a key role in understanding community structure. In this study, the species abundance patterns of tree, shrub and herb layers in a oak (Quercus liaotungensis) forest in Dongling Mountain, Beijing, China, were studied respectively with five niche models, including broken stick model (BSM), niche preemption model (NPM), overlapping niche model (ONM), random assortment model (RAM) and dominance preemption model (DPM). The models that well describing trees abundance patterns have a trend of RAM or NPM to DPM from the low to the high altitude. Most of the shrubs abundance patterns can be well fitted by both NPM and BSM, however the result of NPM is always better. While BSM is the only best fitting model for the herbs. It is obvious that the niche models are suitable in expressing the significant variation of tree species abundance along altitude, but they could not work well in those shrub and herb layers. Other models, such as the neutral theory, might be more suitable to explain the abundance distribution of the understory, which is worth to be considered in further studies.