Sensing is a fundamental function in wireless sensor networks. Researchers have built WSN platforms with a wide spectrum of sensors, ranging from simple thermostats to micro power impulse radars. Traditional signal processing algorithms, however, often prove too complex for energy-and-cost-effective WSN nodes. In this work, we propose a distributed approach for event classification in wireless sensor networks. This approach is based on the assumption that events to be detected can be characterized by a set of features where each can be measured by a specific kind of sensors. The paradigm is composed of two phases: a training phase and a classification phase. In the training phase, for each event type E and a feature F, a set of values is determined. Then, in each node with a sensor corresponding to a feature F we maintain a 2-dimensional array where columns represent event types and rows represent the divisions in the readings’ range of the corresponding sensor. In the classification phase, the network detects and classifies events in a distributed fashion using a voting-like technique in which individual nodes contribute to the classification. Our algorithms are validated through extensive simulations and analysis.