This paper presents a distributed faulty node detection approach for wireless sensor networks that employ aggregation of measurement data on their way toward the sink. The approach is based on the idea of commanding nodes to go into sleep mode in accordance with a schedule that tries to keep the network connected and whose objective is to detect faulty nodes through a process of elimination. The algorithm is run at the sink which monitors the received readings from active nodes and looks for deviations from the expected range, and uses them as triggering events. The implementation was developed using the NS2 network simulator, where performance results were collected, and showed the reliability of the approach and its ability to keep overhead traffic within limits.