First, we demonstrate that not only does data compression help reduce the communication latency in a sensor network (as one would expect) but that it can also be used to gain energy efficiency since data compression reduces the energy spent transmitting data–which is one of the most energyhungry tasks in a sensor network. Through simulations we show that data compression can bring more energy efficiency to a network than does recently proposed combinations of routing and data aggregation. Second, we propose a new viewpoint for data compression in sensor networks that measures distortion based on the compression scheme’s impact on statistical inference tasks (eg, detection, estimation, recognition, etc.) and makes trades in a Rate-Energy-Accuracy design space. The usefulness of this idea is demonstrated through a particular sensor network application: object location in sensorgathered images.