A local magnitude scale is developed for Tanzania, East Africa, using data collected by the 1994/1995 Tanzania Broadband Seismic Experiment. The waveform data from 1189 local and regional earthquakes located within East Africa were corrected for instrument response and convolved with the nominal Wood-Anderson torsion seismograph response appropriate for the original definition of local magnitude. A total of 24,710 maximum S-wave amplitudes were incorporated into an iterative regression for individual event local magnitudes, 38 horizontal component station factors, and 2 linear distance-dependent factors in the log A0 term of the equation for ML. The resulting distance correction, log A0, is given by -log A0 = 0.776 log(r/17) + 0.000902(r − 17) + 2.0, where r is hypocentral distance in kilometers. This distance correction yields much less ground-motion attenuation than observed for southern California and is similar to that observed for local S waves and regional Lg waves propagating in eastern North America. Normalizing the distance correction using the standard constraint of 1-mm ground motion at 100 km for a magnitude 3.0 earthquake results in a bias of nearly half of a magnitude unit between mb and ML, with ML being larger. Normalizing using the constraint of 10 mm of motion at 17 km for ML 3.0 removes the bias in magnitude measures and indicates that structure within Tanzania is relatively high Q. The seismicity rate of Tanzania for 1994/1995 was examined using the Gutenberg-Richter seismicity distribution and is seen to follow the relation log N = 4.63 − 0.84 ML, where N is the number of earthquakes per year of local magnitude ML or greater. The catalog of events used in this study is seen to be complete to magnitude 2 to 2.5.