Eleven tests were conducted to study the dispersion patterns of a radiotracer in five natural stream channels and in one canal. The radiotracer, gold-198, was injected as a line source. The patterns of dispersion that were observed in these channels were compared with patterns predicted by the theoretical model for one-dimensional flow developed by G. I. Taylor (1954). Analysis of the relation between time and concentration of the tracer at several sections in each of the six reaches shows that his theoretical model is not adequate to describe the dispersion patterns actually observed. Dispersion coefficients determined from the test data, though constant as theoretically predicted, are about four to 35 times greater than those predicted by Taylor.