In-network sensor query processing systems are used for power-efficient sensory data acquisition and aggregation in wireless sensor networks. Due to the cross-layer design of these systems and the resource-limited and noisy nature of WSNs, it is challenging to study the end-to-end performance of these systems in a realistic setting. In this work, we design and implement a benchmark, Bisque, for this purpose. We identify the components of an SUT (System Under Test), set the network topology, database schema and population, select a dozen of queries, as well as fix the performance metrics and scaling factors for the benchmark. We apply Bisque to TinyDB and its variations with different network routing protocols and aggregation techniques on our WSN emulation platform, VMNet. Our initial results show both the strengths and the limitations of current-generation WSN query processing systems.