Effective methods for measuring maternal mortality in developing countries are important, particularly in assessing interventions aiming for safer motherhood. Here the performance of different approaches is compared in the same setting. Estimates of maternal mortality in a rural Ethiopian community are reported, made by direct observation, a case-control approach, and the sisterhood method. Adjusted estimates of MMR using these methods ranged between 440 and 665 per 100,000 live births. The advantages and disadvantages of the different approaches are compared, both for operational feasibility and outcome.