A series of low-level tracer experiments conducted in metropolitan St. Louis is described. Values of dispersion parameters, calculated from the tracer data, are related to readily measured or derived meteorological indices of turbulence. The results are presented graphically as families of best-fit curves in terms of downwind distance and travel time. Results are compared with those of previous dispersion experiments conducted over relatively uncomplicated terrain in open country. In terms of the meteorological indices, crosswind dispersion is better described as a function of downwind distance, whereas that in the vertical is about as well described by travel time as by downwind distance. It is concluded that for low-level point sources, the urban area affects crosswind dispersion primarily by enhancing the initial size (i.e., close to the source) of the plume. As the plume becomes much larger than the size of eddies created by the local obstructions, the dispersion closely converges to that associated with flow over open country. In the vertical, significantly enhanced dispersion as well as an enlarged initial cloud dimension occur; the enhancement in the rate of dispersion over that in open country appears somewhat greater for stable than unstable meteorological conditions.