A system and method for optimizing voice communications in a wireless network including an AP having a message waiting time that provides proper QoS while losing minimal communication bandwidth. The QoS may be responsive to the amount of user traffic in both the AP and neighboring APs. The method may include setting parameters for each level of QoS in response to a measure of the degree of contention for that level of QoS, and in response to a measure of the degree of contention for those levels of QoS with higher priority, and setting waiting time parameters in response to a stochastic model of contention at each level of QoS. Operational parameters might include contention window time, AIFS time, and back-off value(s), and might be modified in response to message QoS.