It is expected that solute dispersivity is determined by the physical properties of the transport medium. However, experimental investigations to date have not identified such a relationship. In this work, we examine the relation between soil physical parameters, experimental parameters and dispersivity. Experimental data were compiled for 291 repacked, saturated homogeneous laboratory column experiments reported in the literature. Using multiple stepwise regression and classification and regression tree analysis, the relationships between soil physical parameters (sand, silt and clay content, bulk density, water content), experimental parameters (pore-water velocity, column length and diameter), and transport parameters (dispersion coefficient and dispersivity) were examined. Regression tree analysis showed that clay content was the most important factor controlling dispersivity, followed by column diameter. Columns of length less than approximately 10 cm generally resulted in greater dispersivities than longer columns. A scale effect appeared significant with increasing column diameter, but was not apparent with column length. These results suggest that it may only be possible to relate dispersivity to intrinsic soil properties by explicitly accounting for the experimental design, including column geometry, inlet dead volume and soil packing.