Extending the life of a sensor network while maintaining an acceptable level of accuracy continues to be a critical challenge in long-lived applications requiring months and years of continuous operation. In this paper, we present an approach to address this challenge, namely selective querying based on the transinformation value of nodes relative to the query being executed. The approach is distributed whereby decisions are made by individual nodes and cluster heads based on information locally available. Simulation results establish the feasibility of this approach and show significant gains in the lifetime of the network.