In wireless sensor networks, sensor nodes are usually battery-operated computing and sensing devices, hence operations are limited by the initially equipped batteries, which are hard to be recharged or replaced. In this sense, saving energy consumption becomes significant. As most energy waste is from the always-on wireless interface, S-MAC is suggested to reduce energy consumption by introducing periodic listen/sleep scheme. However, when designing the listening time, S-MAC fails to consider the traffic distribution factor. In this article, an optimization for this scheme is proposed based on the distribution model. Evaluations show that the optimized S-MAC achieves considerable improvement in energy and latency.