The spread of advantageous transgenic genes from crop plants into wild and weedy relatives is a potential ecological problem. The available theory indicates that the spread of a gene, over space and time, will depend in part on the relative selective advantage of that gene, and in part on gene flow, or the probability distribution of gene movement from source to target plants within a single generation. Risk assessment will require knowledge of both. This paper describes an empirical investigation of the effect of spatial dispersion on gene flow in Brassica campestris, utilizing a system of wild genotypes for target plants, and agronomic genotypes with a marker gene for source plants. I have found that gene flow can be reliably modeled with an exponential probability density function, and that the mean of that distribution can vary with the spatial dispersion of plants. Two—dimensional arrays of plants had lower mean gene dispersal distance than did one—dimensional arrays. In contrast, the degree of clumping and the size of patches of plants in my experiments had no effect on gene flow. However, I found an increase in mean gene dispersal distance with increased distance between clumps of plants, or between individual plants in one year. These results imply that large isolation distances accompanied by a rare long—distance pollination event will ultimately increase the risk of gene spread.