In this work, we show how to use a leaky bucket counter (LBC) as a sophisticated threshold mechanism for detecting events in wireless sensor networks. After introducing the LBC and elaborating on various special cases for different possibilities of event detection, we present a case study. Using varying parameters, we compare the performance of the LBC approach to that of a moving average approach and a simple threshold-only mechanism. These mechanisms are of comparable computational complexity and have similar resource demands. The comparison underlines the differences in how old measurements influence the actual detection outcome in different ways. We also explain under which conditions an LBC is suited for event detection and when it is not.