For sensor networks, throughput may not be the most important metric to optimize in medium access control. For many applications, periodic reports are desirable suggesting the need for a time division (TDMA) access scheme. However for many reasons, including nonuniformity of deployment and the large number of sensor nodes anticipated, TDMA is impractical. In this paper, we explore scheduled persistence for medium access control in sensor networks. A continuum of approaches from simple randomized -persistent schedules at one extreme to topology-transparent schedules based on Steiner systems at the other are considered. We investigate the probability of obtaining a collision-free slot before a specified time (number of slots) and show that while the expected throughput of these approaches is the same, their variance is strikingly different. The schemes are also remarkably robust to high density. Furthermore, when schedules are chosen at random for each frame, scheduled persistence offers an interesting alternative for medium access control in sensor networks.