[en] Two mechanisms for the spreading of a tracer pulse are discussed. Stratified flow and diffusion in the rock matrix are normally not accounted for when the hydrodynamic dispersivity is evaluated from tracer tests in the ground. It is shown that where there is stratified flow, the spreading of a tracer pulse cannot be described by the Fickian diffusion-dispersion approach. If a dispersion coefficient were evaluated from such an experiment, the dispersion coefficient would grow proportionally to the observation distance. When a tracer flowing with the water in the channels of a geologic medium can diffuse into the porous matrix of the solid, a tracer pulse will spread. The spreading due to this mechanism cannot be described by Fickian dispersion. In a special case when the time for matrix penetration is long, the observed tracer pulse will have an infinitely long tail. If the conventional moment method is used to determine a dispersion coefficient in such cases, the results will depend on the detection limit of the tracer to a very large extent. (Auth.)