Data from longitudinal dispersion studies by radioactive tracer, conducted by the United States Geological Survey during 1959-1961, are reanalyzed using more recent techniques. Data from 10 tests in 6 streams are examined. Methods available in the literature for calculating dispersion coefficients from tracer data are reviewed, and different methods applied to the same data are found to yield different results. A new method (routing procedure) is proposed, which uses a computer program to determine whether a suggested dispersion coefficient for a given stream predicts downstream tracer passage from upstream data. If the prediction is not adequate, the coefficient is adjusted until the coefficient giving the most accurate prediction is found. The dispersion coefficients obtained by the routing procedure are used to test a theory for predicting dispersion coefficients, proposed by the writer in a separate paper. The predicted and observed coefficients agree within 30% in half of the tests, and in the remainder (including some highly nonuniform streams) within a factor of 4.