In this paper, we present SURCH, a novel decentralized algorithm for efficient processing of queries generated in sensor networks. Unlike existing techniques, SURCH is fully distributed and does not require the existence or construction of a communication infrastructure. It exploits the broadcast nature of wireless communication to optimize query propagation and evaluation. In SURCH, partial results are aggregated en route while the query spreads through the network. The key features of SURCH include its ability to avoid unnecessary communication, balanced node workload, and resilience to node failures. Performance results illustrate that SURCH outperforms alternative techniques for a variety of aggregation and selection queries.