Minimizing energy consumption has been a major objective at all levels in sensor networks. In this paper, we present TiNA, an in-network aggregation scheme that maintains the user-specified quality of data requirement while significantly reducing the overall energy consumption. Specifically, since communication dominates power usage in sensor networks, TiNA exploits end-user temporal coherency tolerances to reduce the amount of information transmitted by individual nodes. Further, we show that TiNa, by using temporal coherency tolerances, can allow for better quality of data when the time given to perform readings is too short for all data to be propagated up through the network. We compare our proposed scheme against an existing in-network aggregation scheme with a local sensor cache. We present experimental results measuring both power savings and also the quality of data for group-by queries. These results show that TiNA can reduce power consumption by up to 60% without any loss in the quality of data and extend the life of the sensor network by up to 300%.