We study four distributed techniques for computing the area of a region in a sensor network. Area calculation is a fundamental sensor network primitive, and distributed, in-network approaches prove more scalable than centralized collection in terms of energy consumption. The four techniques—Delaunay triangulations, Voronoi diagrams, and two new, simpler algorithms, inverse neighborhood and inverse neighborhood with location—vary in computational complexity, communication cost, and information required from the sensor network. We conclude that when sensors know their physical locations, our simple and efficient inverse-neighborhood approach performs comparably to more systematic, but more expensive, computational geometry algorithms. We also analyze the effects of radio range and deployment density on accuracy, and show that topologies derived from real testbeds behave quite differently from commonly seen random topologies with unit disk connectivity.