Desynchronization is a novel primitive for sensor networks: it implies that nodes perfectly interleave periodic events to occur in a round-robin schedule. This primitive can be used to evenly distribute sampling burden in a group of nodes, schedule sleep cycles, or organize a collision-free TDMA schedule for transmitting wireless messages. Here we present Desync a biologically-inspired self-maintaining algorithm for desynchronization in a single-hop network. We present (1) theoretical results showing convergence, (2) experimental results on TinyOS-based Telos sensor motes, and (3) a Desync based TDMA protocol. Desync-TDMA addresses two weaknesses of traditional TDMA: it does not require a global clock and it automatically adjusts to the number of participating nodes, so that bandwidth is always fully utilized. Experimental results show a reduction in message loss under high contention from approximately 58% to less than 1%, as well as a 25% increase in throughput over the default Telos MAC protocol.