The arrival time of data received by a mobile station in IEEE 802.16e is stochastic. When the arrival rate is low, the shorter initial sleep time (Tmin) and the maximal sleep time (Tmax) will lead to the mobile station being awakened frequently, which increases the average listening times and consumes more energy. On the other hand, when the arrival rate is high, longer Tmin and Tmax lead to longer response times. To address these problems, a novel adaptive energy saving mechanism is proposed. Firstly, Tmin and its corresponding Tmax are initialized. The mumber of Tmax which the mobile station lasted in a sleep mode is denoted as J. The realtime Tmin is defined as R(Tmin) when the mobile station is in a sleep mode. R(Tmin) is determined by J which representd the Tmax of the former sleep mode. As a result, Tmin is adjusted adaptively, and the average listening times are reduced. A Markov chain is adopted to analyze the above adaptive energy saving mechanism. The simulation results demonstrate the effectiveness of the proposed energy saving mechanism.