A new clustering approach, termed Distributed Energy Efficient clustering Protocol (DEEP), is proposed for wireless sensor networks. Using a non-iterative cluster formation operation, the protocol spends an extremely low overhead energy compared to the existing protocols and terminates faster than the energy-expensive iterative processes. The distributed head election algorithm guarantees that the periodically-elected leaders have the highest residual energy among their members in each data reporting cycle, effectively balancing the energy consumption among sensors. The DEEP also accounts for sensor limitations as well as practical concerns such as wireless collision that has not been considered by the existing clustering protocols. It intelligently exploits the overhearing ability of sensors to alleviate the loss of clustering control packets due to collisions. The proposed protocol is very effective in forming well-distributed clusters that ensure the required load balancing, the connectivity of clusters, and the minimum data communication energy during the data collection stage. In addition, the DEEP does not make any advanced assumptions about the required number of clusters, the network density, the energy consumption pattern of sensors or their clock synchronization and capabilities. For a thorough evaluation, we have compared the performance of DEEP with an existing clustering protocol. The simulation results show the effectiveness of DEEP in reducing the energy expenditure besides assuring other desirable features. We have also examined its performance in prolonging the network lifetime in the context of a practical routing protocol.